Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
EFFECTS OF WATERSHED LAND USE AND GEOMORPHOLOGY ON STREAM BASEFLOWS IN THE
SOUTHERN BLUE RIDGE MOUNTAINS
by
KATIE PRICE
(Under the Direction of Albert J. Parker and C. Rhett Jackson)
ABSTRACT
The current understanding in watershed hydrology does not provide insight into prediction of low‐flow
hydrologic response to land‐use change in developing regions like the Blue Ridge of north Georgia and
western North Carolina. To address this problem, three separate but complementary studies were
conducted. In the first study, the hydraulic characteristics of soils underlying forest, pasture and
turfgrass (lawn) land use were compared. Forest soils were shown to have substantially greater
hydraulic conductivity, greater water holding capacity, and lower bulk density than lawn and pasture
soils. These differences in soil characteristics associated with land‐use change are of sufficient
magnitude to alter watershed hydrology and decrease baseflows. In the second study, 35 watersheds
representing the range of human impact in the study area were instrumented for continuous discharge
measurement over a 16‐month period, and multiple regression analysis was used to relate watershed
geomorphic and land‐use characteristics to baseflow metrics. Watersheds with greater forest cover
were shown to have higher baseflow, counteracting the theory that forest cover reduces baseflows, due
to the high evapotranspiration rates of trees. Watershed geomorphic characteristics of drainage
density, slope variability, and colluvium were also shown to have significant impact on stream
baseflows. In the third study, the distributed, GIS‐based hydrologic model WetSpa was used to simulate
30 years of streamflow for four watersheds, under eight land use scenarios. More‐developed
watersheds were associated with greater baseflow magnitude, countering a great deal of empirical data
that shows the opposite relationship. Less‐developed watersheds demonstrated a higher proportion of
baseflow to total streamflow. The model contained embedded theoretical assumptions about the
importance of evapotranspirative losses from forest land use, without adequately accounting for the far
greater hydraulic conductivities of forest soils in contributing to subsurface storage recharge and
baseflow. Overall, the results of this dissertation demonstrate that land‐use conversion from natural
forest to pasture, low‐, and medium‐intensity development are associated with reduced baseflows, and
that factors of watershed geomorphology are also important. The modeling results should serve as a
strong caution to potential users of complex hydrologic models regarding the possible inaccuracies of
underlying assumptions within the model.
INDEX WORDS: Baseflow, Land‐Use Change, Watershed Hydrology, Soil Hydrology, Distributed
Models, Watershed Geomorphology, Human Impacts, Low Flow, Appalachian Mountains, Blue Ridge Mountains
EFFECTS OF LAND USE AND GEOMORPHOLOGY ON STREAM BASEFLOWS IN THE SOUTHERN
BLUE RIDGE MOUNTAINS
by
KATIE PRICE
B.A., The University of Georgia, 2000
M.S., The University of Georgia, 2004
A Dissertation Submitted to the Graduate Faculty of The University of Georgia in Partial Fulfillment of
the Requirements for the Degree
DOCTOR OF PHILOSOPHY
ATHENS, GEORGIA
2009
© 2009
Kate Marie Price
All Rights Reserved
EFFECTS OF LAND USE AND GEOMORPHOLOGY ON STREAM BASEFLOWS IN THE SOUTHERN
BLUE RIDGE MOUNTAINS
by
KATIE PRICE
Major Professor: Albert J. Parker C. Rhett Jackson
Committee: George A. Brook
Marguerite Madden Todd C. Rasmussen
Electronic Version Approved: Maureen Grasso Dean of the Graduate School The University of Georgia December 2009
iv
DEDICATION
This dissertation is dedicated to my parents, Kay and Charles Price, whose support, love, and
humor never cease. I am incredibly lucky to have such generous and wonderful parents, who have
always encouraged me to make the most out of life.
v
ACKNOWLEDGEMENTS
The primary financial support for my doctoral education and research was provided by the U.S.
Environmental Protection Agency Science to Achieve Results (STAR) Fellowship Program and National
Science Foundation (NSF) Doctoral Dissertation Improvement Grant BCS‐0702857. Additional support
was provided by the Coweeta LTER program (NSF Cooperative Agreement DEB‐0218001), The University
of Georgia Research Foundation, and the University of Georgia Women’s Club Scholarship Fund.
I would foremost like to thank my endlessly tolerant, loving, and supportive husband David for
bountiful encouragement and moral support over the past several years. I owe enormous thanks to my
co‐advisors, Al Parker and Rhett Jackson, for the invaluable experience and knowledge they have shared
with me. They eagerly gave very insightful and prompt feedback throughout the entire process of
planning and completing my dissertation research. The rest of my committee, George Brook,
Marguerite Madden, and Todd Rasmussen all greatly improved this work. My wonderful, sweet, and
silly family and all of my good friends have provided comic relief and support throughout my education.
The Department of Geography (faculty, staff, and students), has been a wonderful support network
throughout the course of my graduate education. Everyone’s smiling faces and generous attitudes have
been great for my morale! I would especially like to thank Audrey Hawkins for kindly keeping me and
everyone else on track, Donna Johnson for accounting support, and Jodie Guy for crisis aversion and
beaurocracy‐wrangling.
This research would not have been possible without generous assistance from many people.
Permission for site use was provided by the U.S. Forest Service, North Carolina Department of
Transportation, Jackson County Parks and Recreation Department, Friends of the Little Tennessee
Greenway, and a large number of private landowners. Jim Vose and Stephanie Laseter of the U.S. Forest
vi
Service Coweeta Hydrologic Laboratory provided streamflow and climate data. Substantial expertise
and insights were generously provided by Larry Morris, Larry West, Tom Mote, Wayne Swank, John
Chamblee, Tommy Jordan, Mu Lan, and Kelli Coleman. Field help was provided by Todd Headley,
Shelley Robertson, Clint Collins, Julia Ruth, Gregoryian Willocks, Raina Sheridan, Jason Love, Jason
Meador, Jim Kitchner, Jake McDonald, Amber Ignatius, and Ryan Ignatius. Generous resource support
was supplied by the Hydrology Laboratory in the Warnell School of Forestry and Natural Resources and
the Geomorphology Laboratory in the Department of Geography. This research was conducted in
affiliation with the Coweeta Long Term Ecological Research (LTER) Program ‐ I am truly grateful to have
had the opportunity to work with this interdisciplinary group. I would especially like to offer my sincere
thanks to the lead principal investigator of the LTER, Ted Gragson of the Department of Anthropology,
who has generously provided encouragement and resources for this project from the very beginning.
Finally, I would like to acknowledge Josh Romeis, who was one year ahead of me in my doctoral
education and a wonderful source of advice and commiseration, who tragically passed away in August
2009.
vii
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS ................................................................................................................................. v
LIST OF TABLES .............................................................................................................................................. x
LIST OF FIGURES ............................................................................................................................................ xi
CHAPTER
1 INTRODUCTION ......................................................................................................................... 1
2 INFLUENCES OF WATERSHED LAND USE AND GEOMORPHOLOGY ON STREAM BASEFLOWS:
A REVIEW .............................................................................................................................. 5
2.1 Introduction .................................................................................................................... 7
2.2 Geomorphic controls on baseflow ................................................................................. 9
2.3 Effects of human land use on baseflow........................................................................ 15
2.4 Summary and conclusions ............................................................................................ 24
2.5 References .................................................................................................................... 25
3 VARIATION OF SURFICIAL SOIL HYDRAULIC PROPERTIES ACROSS LAND USES IN THE
SOUTHERN BLUE RIDGE MOUNTAINS, USA ....................................................................... 40
3.1 Introduction .................................................................................................................. 42
3.2 Study area ..................................................................................................................... 46
3.3 Methods ....................................................................................................................... 47
3.4 Results .......................................................................................................................... 51
3.5 Discussion ..................................................................................................................... 54
3.6 Conclusions ................................................................................................................... 60
viii
3.7 Acknowledgements ...................................................................................................... 60
3.8 References .................................................................................................................... 61
4 EFFECTS OF LAND USE AND GEOMORPHOLOGY ON STREAM BASEFLOWS IN THE SOUTHERN
BLUE RIDGE MOUNTAINS OF GEORGIA AND NORTH CAROLINA, USA .............................. 79
4.1 Introduction and Background ....................................................................................... 81
4.2 Study Area .................................................................................................................... 84
4.3 Methods ....................................................................................................................... 85
4.4 Results .......................................................................................................................... 91
4.5 Discussion ..................................................................................................................... 95
4.6 Conclusions ................................................................................................................. 106
4.7 References .................................................................................................................. 108
5 A TEST OF A DISTRIBUTED, GIS‐BASED HYDROLOGIC MODEL FOR EVALUATING BASEFLOW
RESPONSE TO LAND USE CHANGE IN THE SOUTHERN BLUE RIDGE MOUNTAINS OF
NORTH CAROLINA ............................................................................................................ 132
5.1 Introduction ................................................................................................................ 134
5.2 Model Description ...................................................................................................... 136
5.3 Methods ..................................................................................................................... 139
5.4 Results ........................................................................................................................ 144
5.5 Discussion ................................................................................................................... 145
5.6 Conclusions ................................................................................................................. 150
5.7 References .................................................................................................................. 151
6 SUMMARY AND CONCLUSIONS ............................................................................................. 168
APPENDICES .............................................................................................................................................. 172
A Discharge Rating Curves ........................................................................................................ 172
ix
B Hydrographs .......................................................................................................................... 179
C WetSpa Parameterization and Flow Routing ........................................................................ 198
x
LIST OF TABLES
Page
Table 2.1: Anthropogenic impacts on baseflow ......................................................................................... 37
Table 2.2: Recharge response to various effects of urbanization .............................................................. 38
Table 2.3: Summary of studies assessing the response of baseflow and recharge to urbanization .......... 39
Table 3.1: Land use – Macon Co. and Jackson Co., NC ............................................................................... 67
Table 3.2: Site characteristics: Land use, soil series, elevation, and aspect of soil sampling sites ............. 68
Table 3.3: National Cooperative Soil Survey Official Series Descriptions ................................................... 69
Table 3.4: Hydraulic properties of upper (0‐7.5 cm) and lower (7.5‐15 cm) soil cores .............................. 70
Table 3.5: Soil physical characteristics by categories of parent material and land use .............................. 71
Table 3.6: Paired locations (adjacent locations with different land uses) .................................................. 72
Table 4.1: General topographic and land use characteristics of study watersheds ................................. 116
Table 4.2: Explanation of watershed characteristics considered for use in regression modeling ........... 117
Table 4.3: Correlations among independent variables included in regression analysis ........................... 119
Table 4.4: Precipitation stations ............................................................................................................... 120
Table 4.5: Best models for each dependent variable ............................................................................... 121
Table 4.6: Comparison of paired watershed flows ................................................................................... 122
Table 4.7: Correlations between land use and baseflow metrics ............................................................. 123
Table 5.1: 2006 land use of study watersheds and explanation of reclassification of NLCD land use
scheme .................................................................................................................................... 155
Table 5.2: Flow summary values for varied land use scenarios ................................................................ 156
xi
LIST OF FIGURES
Page
Figure 3.1: Study area and soil sampling locations: Macon Co. and Jackson Co., NC ................................ 73
Figure 3.2: Soil physical characteristics by parent material and land use .................................................. 74
Figure 3.3: Particle size distributions of forest, lawn, and pasture soils .................................................... 75
Figure 3.4: Comparison of saturated hydraulic conductivity (Ksat) measurements by field and laboratory
methods. ................................................................................................................................... 76
Figure 3.5: Saturated hydraulic conductivity among sites .......................................................................... 77
Figure 3.6: Comparison of soil saturated hydraulic conductivities with precipitation intensities occurring
in western North Carolina ......................................................................................................... 78
Figure 4.1: Study area and monitored watersheds .................................................................................. 124
Figure 4.2: Representative example of rating curves developed using different rating curve fitting
methods .................................................................................................................................. 125
Figure 4.3: Hydrologic conditions during the study period ...................................................................... 126
Figure 4.4: Examples of Bayesian power law rating curves used in this study ......................................... 127
Figure 4.5: Interpolations of precipitation for the three time periods in this study ................................ 128
Figure 4.6: Ranges of values for baseflow metrics across all study watersheds ...................................... 129
Figure 4.7: Difference of mean baseflows between lower‐ and higher‐ forest cover watersheds ......... 130
Figure 4.8: Examples of varied recharge response to Tropical Storm Fay ................................................ 131
Figure 5.1: Study areas and watersheds used for streamflow simulation ............................................... 157
Figure 5.2: Summary of WetSpa spatial parameterization ....................................................................... 158
Figure 5.3: Simulated land use scenarios .................................................................................................. 159
xii
Figure 5.4 Comparison of simulated vs. observed streamflow ................................................................ 160
Figure 5.5: Flow duration curves for the varied land use scenarios ......................................................... 162
Figure 5.6: Recurrence intervals of 7‐day annual low flows ..................................................................... 164
Figure 5.7: Streamflow summary metrics for each of the progressive land use scenarios ...................... 166
Figure 5.8: Streamflow summary metrics for each of the spatial land use scenarios .............................. 167
1
CHAPTER 1
INTRODUCTION
Baseflow is the portion of streamflow that is sustained between precipitation events, fed to
stream channels by subsurface pathways. Baseflow is influenced by natural factors such as climate,
geology, relief, soils, and vegetation. Human impacts on the landscape may modify some or all of these
factors, in turn affecting baseflow timing and quantity. Understanding baseflow is of great importance,
as these flows are critical to issues of water quality, water supply, and aquatic habitat. Baseflow has
been shown to be strongly influenced by watershed characteristics of geomorphology and land use, but
the relative influences of these factors has remained unresolved. Within a given setting of climate and
geology, watershed geomorphology affects baseflow by influencing the amount, type, and distribution
of subsurface storage reservoirs, and channel network characteristics that relate to the rates of removal
of stored water from the catchment. Upon this natural template, human land use also exerts strong
influence on baseflow. Compared with natural vegetation, human land use alters many aspects of the
hydrologic system. Additions of impervious surface and soil compaction accompany most forms of
human land use, thereby decreasing infiltration of rainwater and recharge of the subsurface storage
reservoirs that sustain baseflows. Additionally, changes in land cover result in changes in
evapotranspiration rates, altering catchment water budgets. A thorough review of literature addressing
watershed influences on baseflow is presented in Chapter 2.
The primary objective of this research was to assess the relative impacts of watershed
geomorphology and land use on baseflows in the southern Blue Ridge Mountains, within the upper Little
Tennessee River system, upstream of Lake Fontana (2628 km2). This region is characterized by
2
pronounced topographic relief, with valley floors of 500 masl and peaks nearing 2000 masl. Land use in
the region contains significant areas of protected National Forest, but there is rapid growth and
development occurring in unprotected portions of the region. This area provides an ideal setting for
addressing linkages between surface characteristics and baseflow for several key reasons: 1) The
mountainous relief in this area is associated with pronounced topographic variability, allowing
comparison of diverse morphometric settings. 2) The region is underlain by crystalline bedrock,
avoiding complicated hydrology associated with porous or soluble sedimentary terrain. 3) There
exists an acute need for heightened understanding of stream response to human impact in this
rapidly developing region, especially due to the presence of many threatened aquatic species. 4)
The presence of the Coweeta Hydrologic Laboratory and Long Term Ecological Research Station
(LTER) in the within the study area allows for a large quantity and variety of related background
data and connects this research to ongoing monitoring and assessment of human impacts on the
landscape.
The main objective of assessing the roles of human land use and watershed geomorphology
on stream baseflows in the southern Blue Ridge was accomplished via three separate but
complementary specific research questions:
1) How does human land use alter the hydrologic characteristics of soils, and what are the
implications of these impacts on watershed baseflow processes?
2) What is the observed range of baseflow variability among southern Blue Ridge watersheds,
and which watershed geomorphic and land use characteristics are linked to this variability?
3) How will stream baseflows respond to human impact that exceeds current levels of
development, and do differences in watershed topography elicit differences in baseflow
response to land‐use change?
3
The research conducted to address these three questions are presented as separate manuscripts in
this dissertation (Chapters 3‐5).
Chapter 3 addresses the influence of human land use on soil hydraulic characteristics. Soils
under three land use classes (forest, pasture, and turfgrass) in equivalent slope and parent material
were compared. Forest soils were shown to have significantly lower bulk density and greater
hydraulic conductivities and water holding capacities than pasture or lawn soils. Differences
between the hydraulic characteristics under forest vs. nonforest soils were of a magnitude to carry
significant implications for watershed function. Chapter 4 presents an empirical assessment of the
overall variability in baseflow among watersheds with varied geomorphic and land use
characteristics. Overall, geomorphic characteristics exerted a greater influence on stream
baseflows than land use in this region. Watershed geomorphic characteristics of drainage density,
slope variability, and colluvium were shown to significantly influence baseflows. Watersheds with
greater forest cover were associated with higher baseflow, despite the substantial water use of
mature trees. This underscores the importance of high infiltration and subsurface recharge in
undisturbed soils. Chapter 5 contains the results of GIS‐based, fully distributed modeling of stream
baseflow response to land‐use change in three watersheds of varied topography. The WetSpa
model was used to explore baseflow response to varied land use change scenarios among four
watersheds. Simulated streamflow over a 30‐year period, under eight land use scenarios, showed
increasing baseflow magnitudes under land use scenarios. These results contradict the empirical
data, and result from the structure of the model, which assigns higher evapotranspiration rates to
forest land cover, without adequately accounting for the much higher infiltration rates in forest
soils, compared with nonforest soils. These results serve as a strong caution regarding the use of
4
hydrologic models. Finally, Chapter 6 presents the overarching summary points and conclusions of
the dissertation.
5
CHAPTER 2
INFLUENCES OF WATERSHED LAND USE AND GEOMORPHOLOGY ON STREAM BASEFLOWS: A REVIEW1
________ 1K. Price. To be submitted to Progress in Physical Geography
6
ABSTRACT
Baseflow is the portion of streamflow that is sustained between precipitation events, fed to stream
channels by subsurface pathways. Understanding baseflow is of great importance, as these flows are
critical to issues of water quality, water supply, and aquatic habitat. There is a large body of literature
investigating the relationships between watershed characteristics and stream baseflow, and this paper
seeks to synthesize the findings of case studies throughout the world. This review emphasizes literature
research covering the relationships between baseflow and watershed geomorphology and land use
(emphasizing forest, agriculture, and urbanization), which are key controls on baseflow through their
influence on infiltration, rates of water removal from the catchment, and subsurface storage properties.
The literature shows that there is much that remains to be resolved toward a solid understanding of
how watershed properties influence baseflow. While it is clear that watershed topography and
geomorphology influence baseflow, there is no consensus on which geomorphic parameters are most
closely linked to subsurface storage and baseflow. Similarly, there is little consistency among studies
investigating the effects of land‐use change on baseflow. Many studies associate higher watershed
forest cover with lower baseflows, attributed to high evapotranspiration rates of forests, while other
studies indicate increased baseflow with higher watershed forest cover due to higher infiltration and
recharge of subsurface storage. The demonstrated effects of agriculture and urbanization are also
inconsistent, due to varied additions of imported water and extremely variable natural background
conditions. This review underscores the need for more research that addresses multiple aspects of the
watershed system in explaining baseflows, and for methodological consistency to allow for more fruitful
comparisons across case studies.
7
2.1. Introduction
Baseflow is influenced by natural factors such as climate, geology, relief, soils, and vegetation.
Human impacts on the landscape may modify some or all of these factors, in turn affecting baseflow
timing and quantity. The need for a greater understanding of streamflow response to external change
has been recognized for decades, but previous research has tended to emphasize flood response to
increased human pressures on the landscape (e.g., Knox, 2001; Choi, 2004). In this respect, the literature
is under‐represented in studies investigating baseflow response to human impact. A scientific
understanding of watershed processes and baseflow is critical to effective water policy and
management. Population growth is associated with increasing demands on freshwater resources for
industry, agriculture, and human consumption, and water shortages are not uncommon in the United
States, even in humid regions (Hornbeck et al., 1993). A firmer grasp on the controls of baseflow is
pivotal in issues of contaminant dilution (Novotny and Olem, 1994; Jordan et al., 1997; Barnes and
Kalita, 2001), stream ecology (Klein, 1979; Boulton, 2003; Konrad and Booth, 2005), and adequate water
supply to population centers (Hornbeck et al., 1993, Illinois EPA, 2002). Human waste allocation
requires accurate estimation of baseflow discharge (Smakhtin, 2001), and contaminants that enter
stream systems via soil or groundwater storage are most highly concentrated during baseflow. These
factors carry negative implications for stream biota and human consumption if baseflows are reduced
(Novotny and Olem, 1994; Barnes and Kalita, 2001; Dewson et al., 2007). Reduced baseflow is also
associated with reduced stream width, warmer temperatures, lower dissolved oxygen, and higher
nutrient concentrations that may promote excessive growth of habitat‐choking algae. These conditions
are often fatally stressful for sensitive, endemic species, and low water levels in streams have been
associated with decreases in richness of aquatic macroinvertebrate and fish species (Boulton, 2003;
Mote et al., 2004).
8
2.1.1 Baseflow overview ‐ Within the literature, there is inconsistent terminology usage, with “baseflow”
and “low flow” commonly used interchangeably to denote streamflow occurring between precipitation
and/or snowmelt events, resulting from sustained subsurface inputs to the stream channel. These and
other terms are also inconsistently differentiated within the literature to specify the lowest annual
streamflow within a watershed or region. In this review, the term “baseflow” will be used generally to
represent streamflow fed from deep subsurface and delayed shallow subsurface storage between
precipitation and/or snowmelt events (Ward and Robinson, 1990), with “annual low flow” specifying dry
season minimum flow (Smakhtin, 2001). Several sources emphasize that “baseflow” is not synonymous
with groundwater flow, as it includes water transmitted from shallow unsaturated storage in addition to
significant contributions from phreatic storage (Anderson and Burt, 1980; Ward and Robinson, 1990;
Buttle, 1998; Brutsaert, 2005). Baseflow is also derived from the drainage of near‐surface valley soils
and riparian zones, as water concentrates in these areas during and following precipitation events
(Smakhtin, 2001; Brutsaert, 2005). Baseflow is naturally influenced by a wide range of factors
(Brutsaert, 2005):
• Basin physiographic characteristics
• Distribution of storage in river channels and groundwater aquifers
• Evapotranspiration(ET) from stream banks and throughout the catchment
• Geomorphology of the landscape and stream network
• Configuration and nature of the riparian aquifers and near surface soils
Many of these factors may be altered with human impact on the landscape, and it thus becomes critical
to understand not only the relationships between basin physical properties and stream baseflow, but
also the ways in which anthropogenic impact affects these physical properties.
9
2.1.2 Limits and scope of this review ‐ This review emphasizes literature covering geomorphic and
anthropogenic effects on baseflow in humid regions of the world, avoiding, for example, arid and polar
settings. Additionally, karst environments (or other settings with disconnected groundwater and surface
water systems) are not covered, and there is bias toward representation of upland systems with shallow
groundwater storage. While climate is a clear factor in determining stream baseflow, via direct
influence on precipitation input and evaporation and indirect influence via vegetation and drainage
network response, the effects of climate on baseflow are not addressed. Furthermore, this review
highlights directions (as opposed to magnitudes) of baseflow response to human impact, because of
inconsistent methodologies and metrics in the studies presented.
A section on geomorphic controls on baseflow discharge will cover the influences of basin
geology, surface topography, subsurface topography, and soils. This section is followed by an overview
of anthropogenic effects on baseflow, with emphases on forest removal, agriculture, and urbanization,
because of the large body of research on those topics.
2.2. Geomorphic controls on baseflow
2.2.1 Bedrock geology ‐ Catchment geology substantially affects baseflow‐generating processes
(Farvolden, 1963; Freeze, 1972; Smakhtin, 2001; Tague and Grant, 2004; Neff et al., 2005). In regions
underlain by permeable, soluble, or highly fractured bedrock, groundwater storage volumes within the
bedrock itself may be highly significant, and the connectivity to the surface water network may be
extremely complex. In contrast, areas underlain by crystalline or massive bedrock with minor fracturing
may not store significant quantities of water and thus contribute to relatively short water residence
times. In karst environments, a losing effect on baseflow has been observed (White, 1977), due to the
often very high storage capacities in limestone and dolomite solution cavities. The idea that the
hydrogeologic properties of catchment bedrock influence storage, residence times, and baseflow is
10
straightforward, but catchment geology also indirectly affects basin hydrology in its influence on
drainage network structure. Easily eroded bedrock lends itself more readily to channel formation and
pedogenesis, both affecting storage capacities and rates of water transmission (Farvolden, 1963).
2.2.2 Surface topography – Thorough assessment of basin topography is often missing from watershed
analyses. Surface topography is a key control on baseflow, both directly and indirectly, and the influence
of topography is most pronounced in relatively high relief settings (Tetzlaff et al., 2009). Exceptions exist
in karst or highly porous settings, such as volcanic or glacial terrain, where water can move freely in the
subsurface below surface drainage divides (Devito et al., 2005). Topographic gradients control the rate
at which soil water moves downslope, thereby determining whether stormwater is flushed to the
channel network or retained in the soil post‐event. The effect of land‐use change on streamflow may be
mitigated or amplified by basin surface and/or subsurface topography, and ideally these factors should
be considered in assessment of stream response to human impact (Dubé et al., 1995; Iroumé et al.,
2005). Little is known regarding which specific topographic variables are most useful for predicting
baseflow and/or explaining baseflow variability response to land use change, but many metrics have
been demonstrated as beneficial components of hydrologic models.
Metrics of surface topography in hydrologic modeling are often reduced to single indices, with
Beven and Kirkby’s (1979) topographic index (TI) the most common. TI is computed as ln(α/tan β),
where α = specific contributing area to a given site, and β = the local slope angle at that site. TI
increases as contributing area increases and slope angle decreases. Increasing drainage area should
increase groundwater contributions, and decreasing slope angle should reduce the rate of groundwater
transmission, assuming that surface topography approximates the hydraulic gradient for shallow
groundwater systems (Buttle et al., 2001). Troch et al. (1993) reported that the TOPMODEL approach
using TI and soil transmissivity yielded accurate depths to shallow water tables. However, many studies
11
that test predicted versus observed water table depths, streamflows, or other related factors using this
approach have reported limited success (Burt and Butcher, 1985; Jordan, 1994; Moore and Thompson,
1996; Rodhe et al., 2006; Buttle et al., 2001). Furthermore, the index is so highly generalized that basin
mean TI values may not vary greatly within a study region (McGuire et al., 2005), limiting its use in cross‐
site comparisons. The lack of total success of such an approach does not by any means negate the
importance of surface topography in the storage and transmission of baseflow, although some of these
authors arrive at that conclusion. The lack of success is at least partially due to the insufficiency of the
index in characterizing basin topography. Though obviously simplistic, TI is readily computed from
digital terrain data and incorporated into spatial models, and is thus widely used in popular applications,
such as TOPMODEL (Beven and Kirkby, 1979), SWAT (Anderson et al., 1993), and WetSpa (Liu et al.,
2003).
Several studies have demonstrated that parameters expressing catchment geometry (e.g,
hypsometric integral, metrics expressing degree of stream network development, and indices of
flowpath length and gradient) are beneficial in prediction and analysis of baseflow and related factors
(Farvolden, 1963; Woods et al., 1997; McGuire et al, 2005). Among many influences addressed,
Farvolden (1963) found potential discharge (a flow component related to baseflow) to be most strongly
correlated to basin geometry in a mountainous region of Nevada. Woods et al. (1997) devised a
subsurface flow index based on surface topography, which the authors report to efficiently describe the
time‐varying spatial pattern in subsurface runoff generation, ideal for use in steep forested catchments
in humid climates. Corroborating the idea that catchment‐scale flow path distribution is largely a
function of catchment geometry (Kirchner et al., 2001; Lindgren et al., 2004), McGuire et al. (2005)
found strong correlations between catchment terrain indices representing flow path distance and
gradient to the stream network in the Oregon Cascades. Santhi et al. (2008) found topographic relief to
be a predictor of baseflow index (proportion of total streamflow as baseflow) on a regional scale.
12
However, dimensionless topographic parameters were shown to have no relationship with baseflow
index in southeastern Australia (Lacey and Grayson, 1997). Drainage density, or the length of stream
network per unit watershed area, has been shown to have a negative relationship to baseflow in many
settings (Farvolden, 1963; Gregory and Walling, 1968; Marani et al., 2001; Warner et al., 2003). Higher
drainage density is synonymous with greater contact area between subsurface storage and stream
channels. Presumably, this greater contact area facilitates removal of water and reduced baseflows
during drier times of year.
In addition to its influence on subsurface flowpath distribution and transit times, surface
topography also relates to the distribution of shallow storage. Surface topographic characteristics may
express the amount of alluvial bottomland and floodplain storage (Brown et al., 2005), and the presence
and extent of colluvium available for subsurface water storage. Alluvial aquifers are understood to be a
key source of streamflow in many settings (Larkin and Sharp, 1992). In theory, the presence and extent
of alluvial valleys is closely linked with baseflow quantity, though few studies have directly addressed
this relationship (Brown et al, 2005; Soulsby et al., 2006). Schilling (2009) showed that amounts of
groundwater recharge were highly dependent on topographic position, with the greatest quantities of
recharge observed in alluvial zones. Using geochemical and isotopic tracers, Tetzlaff and Soulsby (2008)
demonstrated that the upper 54% of a large river catchment in Scotland supplied 71% of the system
baseflow, and that the groundwater of the lower slopes of montane headwaters (where colluvium
deposits occur) provide a major source of baseflow to the river system. Colluvium has also been shown
to be an important shallow reservoir in the Cascades (Galster and Leprade, 1991; Shulz et al., 2008).
2.2.3 Subsurface topography and soil characteristics‐ Subsurface topography, in addition to surface
relief, exerts strong influence on water storage and throughflow pathways, and thus influences
baseflow. Throughflow processes require a confining layer through which water cannot easily infiltrate,
13
thereby initiating lateral subsurface flow (Hutchinson and Moore, 2000). It is these confining layers that
prevent continued infiltration of water, thereby allowing shallow storage contributions to baseflow. In
hydrologic modeling, topographic indices to estimate soil moisture properties and rates of throughflow
are generally limited to metrics of surface topography, despite the influence of confining layers on
flowpaths. Many studies have indicated substantial influence of subsurface topography on hillslope
hydrology and soil moisture characteristics (e.g. McDonnell et al., 1996; Gburek and Folmar, 1999;
Hutchinson and Moore, 2000; Chaplot and Walter, 2003; Chaplot et al., 2004). During or immediately
following storm or snowmelt events, when water table elevations are relatively high, the soil moisture
surface is more likely to parallel the surface topography than that of the confining layer (Hutchinson and
Moore, 2000). However, the influence of subsurface topography is of particular importance during
relatively low moisture conditions, when the topography of the confining layer may be the predominant
control on moisture retention, and, thus, an important factor for baseflow. However, no known studies
have specifically addressed the influence of subsurface topographic characteristics on stream baseflows.
Subsurface strata that induce throughflow are widely varied, but are most often associated with
pedogenically unaltered parent material. Dense bedrock (Hursh and Fletcher, 1943), impermeable
saprolite (Chaplot et al., 2004), heavily compacted till (Hutchinson and Moore, 2000; Reuter and Bell,
2003), and hydraulically restrictive loess layers (O’Geen et al., 2004) have all been demonstrated to
influence soil and hillslope hydrology. Additionally, pedogenic features such as claypans (Wilkison and
Blevins, 1999), well‐developed argillic horizons (Perillo et al., 1999), fragipans (Parlange et al., 1989), and
relatively dense mineral horizon development beneath loose organic layers (Kim et al., 2005) have been
shown to limit vertical infiltration, although the effect is generally not adequately widespread to
significantly impact meso‐ or macroscale hydrology. Pedogenic features generally fail to function as true
confining layers, primarily due to macropore and preferential flow path development across the
hydraulically restrictive horizon. Soil pipeflow has been demonstrated as extremely significant in humid
14
areas (Bryan and Jones, 1997). As macropores and pipes serve as conduits for a substantial share of soil
water, networks that perforate less‐permeable layers undermine their confining potential. Such
networks commonly disrupt potential pedogenic confining layers, such as argillic horizons. Tree root
growth, animal burrowing, and other bioturbation processes affect soil horizons to a much greater
extent than seen with parent material confining layers such as bedrock, saprolite, or compacted till.
Wilkison and Blevins (1999) used chemical tracers to demonstrate vertical preferential flow paths
through a claypan to outweigh lateral throughflow above the claypan. Similarly, Perillo et al. (1999)
identified vertical preferential flow pathways created by decayed roots through a well‐developed argillic
horizon that partially induced lateral flow. Thus, it seems that extreme circumstances are required for
pedogenic features to serve as broadly‐influential confining layers. These circumstances seem
particularly unlikely to be met in humid forested environments, where biological activity is abundant
and disruptive to hydraulically resistant horizons. Thus, it is assumed that lithologic contacts underlying
soil, such as the soil/bedrock or saprolite/bedrock interface (Hatcher, 1998; McDonnell et al., 1996), are
more important in governing subsurface flow and contributions to baseflow than pedogenic features in
the soil itself.
2.2.4 Combined influences of topography and soils ‐ Soil properties influence the distribution of water
storage, but correlations between soil properties and topography typically hinder isolation of the
influence of soil characteristics on water storage and baseflow. Primarily, variation in soil texture plays a
significant role in the rate of moisture loss due to surface or subsurface topographic gradients (Dodd
and Lauenroth, 1997; Yeakley et al., 1998). Spatially dependent soil texture and hydraulic conductivity is
often present in settings with substantial subsurface flow (Wilson et al., 1989). Soil textural variability
undoubtedly affects the rate at which water is transmitted through subsurface pathways, particularly
during unsaturated conditions (Famiglietti et al., 1998). Spatial variability of soil moisture is most
15
pronounced during unsaturated conditions between storm events (Van Ommen et al., 1989; Hutchinson
and Moore, 2000; Sidle et al., 2000; Kim et al., 2005), and such variability is partially attributable to soil
texture. However, determining the strength of this influence is complicated by the theoretical
correlations between topography and soil texture. Systematic downslope variation in soil texture
commonly occurs, as the result of decreasing slope and corresponding slowed rates of water movement
from ridge to toeslope positions (Schaetzl and Anderson, 2005). Thus, correlations between soil texture
and hillslope position are likely to exist, with finer particle size, thicker soils, and low slope gradients
combining their influences to encourage soil moisture retention. Conversely, steep upper slopes are
likely characterized by coarser, less developed, and thinner soils, thereby more rapidly transmitting
water. Furthermore, soil hydrology is strongly affected by spatial variability of soil moisture, which may
be predominantly controlled by surface and/or subsurface topography (Woods et al., 1997). From this
perspective, isolating the influence of soil characteristics from topography is problematic.
2.3. Effects of human land use on baseflow
Widespread vegetation change and soil disturbance accompany most forms of land‐use change,
and such impacts are often sufficient to alter the timing and quantity of baseflow. Additionally, human
impact may involve direct water removal or inputs to streams or catchments. Table 2.1 summarizes
baseflow response to several common forms of human impact. Extreme impact (e.g., urbanization) may
be associated with a total rearrangement of surface and subsurface pathways, in addition to changes in
soil properties, vegetation, etc. This section on anthropogenic controls on baseflow addresses patterns
observed with forest removal, urbanization, and agriculture.
2.3.1 Forest removal ‐ There is an apparent conflict in the literature regarding baseflow response to
forest removal (Johnson, 1998; Smakhtin, 2001; Brujinzeel, 2004; Brown et al., 2005). Globally inclusive
16
literature investigating the role of basin forest cover on flow in headwater stream catchments (i.e., < 2
km2) indicates an increase in mean annual flow in response to removal of basin vegetation (examples of
reviews: Hibbert, 1967; Bosch and Hewlett, 1982; Swank et al., 1988; Sahin and Hall, 1996; Jones and
Post, 2004; Brown et al., 2005), with many studies specifically indicating increases in baseflow (Harr et
al., 1982; Keppeler and Ziemer, 1990; Hicks et al., 1991; Smith, 1991). This relationship is attributed to
greater interception and evapotranspiration rates associated with forest cover, and a solid consensus
exists regarding the greater water use of mature trees compared with other vegetation types (Bosch and
Hewlett, 1982; Calder, 1990; McCulloch et al., 1993). In some cases, these results have been interpreted
as a potentially dangerous suggestion that watershed management approaches could include
deforestation to increase water yield for public use (Brooks et al., 1991; Chang, 2003).
Despite the prevailing literature for headwater streams, there is a sound theoretical basis for
decreased baseflow in response to forest removal, as forest cover is associated with higher infiltration
and recharge of basin subsurface storage. Additionally, the high water holding capacity of forest soils
allows for high water storage and slow drainage, theoretically sustaining higher baseflows (Ohnuki et al.,
2008; Price et al., in review). The negative relationship between watershed forest cover and baseflow
volume for headwater streams presumably results from experimentation methods where the surface
infiltration characteristics are not drastically altered, thus isolating evapotranspiration changes as the
key influence on recharge and baseflow (Bruijnzeel, 2004; Brown et al., 2005). In fact, some studies
investigating permanent land‐use change have shown decreased baseflow from conversion of forest to
nonforest land use (e.g., Bruijnzeel, 2004; Line and White, 2007). Data from 30 streams in the
Piedmont and Blue Ridge provinces of the southern Appalachian Highlands indicate a significant positive
relationship between basin forest cover and baseflow discharge (Price and Jackson, 2007). Due to the
relative lack of soil disturbance associated with paired‐catchment studies emphasizing forest regrowth,
17
very few studies in the experimental forestry literature demonstrate increases in baseflow with greater
forest cover (Smakhtin, 2001; Brown et al., 2005).
The above forestry experimentation studies relating forest removal to increased streamflow
assess response in very small basins, rarely larger than 2 km2, and generally one to two orders of
magnitude smaller. It is unclear whether or not these relationships demonstrate the same direction and
magnitude in larger watersheds (Blöschl, 2001; Sivapalan, 2003; Shaman et al., 2004). Studies relating
forest cover to baseflow in larger systems are extremely rare, and there is no empirical basis for
extrapolating the results of small experimental catchment studies to larger, more heterogeneous basins
(Pilgrim et al., 1982; Smakhtin, 2001; Costa et al., 2003), whereas there is evidence that upscaling from
hillslope/small catchment studies to larger basins leads to erroneous interpretation (Farvolden, 1963;
Sivapalan, 2003; Soulsby et al., 2004). The few examples assessing baseflow response at larger scales
have demonstrated mixed results (e.g. Wilk et al., 2001; Costa et al., 2003), indicating a need for further
investigation. It is probable that increasing topographic complexity with increasing scale may allow for
water storage units (e.g., alluvial bottomlands) that do not exist in the small headwater catchments
emphasized in the forestry experimentation literature. Such storage could potentially offset ET losses
associated with forest cover and sustain higher baseflow in forested settings where sufficient infiltration
occurs, creating a positive relationship between forest cover and baseflow. Additionally, non‐forest land
use in larger basins may be associated with more intensive soil compaction and losses, impervious
surface, etc., that decrease infiltration, recharge, and baseflow beyond the impacts associated with
forest hydrology experimentation (Brujinzeel, 2004; Price et al., in review).
In addition to spatial scale, there are issues of temporal scale. The nature of paired catchment
studies prevalent in the experimental forestry literature generally involves the application of a
treatment and measurement of a response; this is followed by return of the vegetation to pre‐treatment
conditions (Swank et al., 1988). Sub‐ and exurban land uses are generally far more permanent than
18
forestry experimentation, and a complete understanding of long‐term baseflow response to forest
removal is lacking.
2.3.2 Urbanization ‐ Urbanization involves a wide range of impacts, and specific stream response
depends on many factors (Doyle, 2000). Anthropogenic impacts on watershed hydrology accompanying
urbanization involve widespread and drastic re‐organization of surface and subsurface pathways, and
frequently are complicated by importation of water from other watersheds or previously disconnected
groundwater reservoirs. Following urbanization, water is more quickly flushed through catchments due
to reduced hydraulic resistance of land surfaces and channels associated with impervious surface
coverage, channelization, and subsurface storm drainage networks. Intuitively, it follows that
accelerating water removal from stream systems would be linked with corresponding decreases in
recharge and baseflow in urban systems. This assumption dominated hydrologic understanding of
urban impacts for decades, largely due to the influence of Leopold’s (1968) widely cited urban hydrology
guidebook (Brandes et al., 2005). In this benchmark publication, management implications center on
baseflow reduction associated with urbanization, based more on theory than observed trends. While
the assumption that increased impervious surface decreases infiltration, recharge, and ultimately
baseflow is theoretically solid, Leopold’s conceptual model has proven to be overly simplistic and is not
well‐supported by published data (Ferguson and Suckling, 1990). While event flows do consistently
increase and result in faster recession to baseflow with increased impervious surface (Ferguson and
Suckling, 1990; Konrad, 2003; Brandes et al., 2005; Burns et al., 2005), the intuitive corollary of baseflow
decline does not behave quite as neatly, as a result of additional urban effects on subsurface recharge.
The complete picture of hydrologic response to urbanization is extremely complex, with some factors
acting to reduce recharge and others to increase recharge (Table 2.2).
19
Assumptions that urbanization decreases baseflow are generally based on reduced recharge due
to increased impervious surface, which is indeed a dominant factor in urban hydrology. Impervious
surface coverage in urban basins drastically exceeds that of basins experiencing other land‐use types.
Road networks, parking lots, rooftops, etc., all contribute to increased impervious percentages, with
individual cities demonstrating different degrees of greenspace to offset the impacts of impervious
surface. While impervious coverage undoubtedly has an enormous effect on urban hydrology, it is
unrealistic to view urban systems in a surface‐based framework as is commonly applied to systems
experiencing lower‐intensity impacts. In more moderately impacted settings, surface hydrology remains
dominated by natural processes (e.g., evapotranspiration, soil hydrology) following landscape change.
In most urban settings, however, water is completely redistributed to accommodate human activities
and prevent flood damage. Water is routed across the surface and through the subsurface via ditching,
storm drains, water mains, wastewater sewers, and other means, completely altering the rates and
paths of water transmission through urban basins. Such re‐working of the hydrologic system precludes
explanation of baseflow response to urban land use solely in terms of the effects of vegetation removal
and increased impervious surface (Lerner, 2002; Meyer, 2005), although such simplification is still
commonplace.
A major additional complication occurs in urban systems: virtually all major cities import water
(Lerner, 2002). The importation of water may include pumping from deep groundwater that is
otherwise disconnected from the surface water system, piping of water from other watersheds, and/or
withdrawal of water from downstream reservoirs. This water is redistributed throughout cities via pipe
networks that have been consistently demonstrated to lose substantial quantities of water (Lerner,
2002). Lerner (1986) reports water main leakage rates of 20‐25% to be common, with rates reaching as
high as 50%. Wastewater sewer systems have also been observed to leak substantial amounts of water,
which often originates outside the drainage basin. Such leakage, along with surface inputs of imported
20
water (e.g., septic drainage, lawn/garden watering, and other forms of outdoor domestic water usage)
may enter subsurface storage and can significantly offset or overshadow storage losses due to other
urbanization effects. Sustained baseflow with urbanization has also been attributed to ET reduction
associated with vegetation removal (e.g., Appleyard et al., 1999; Rose and Peters, 2001). However, the
role of ET in urban systems remains largely unresolved. For example, Oke (1979) demonstrated ET rates
to remain steady despite decreased vegetation cover in Vancouver, B.C., due to advection of heat from
non‐vegetated surfaces. While such processes may act significantly in suburban areas or cities with
abundant vegetation, they cannot be assumed to dominate in all urban areas.
All of the factors addressed above may be expressed to varying degrees in different cities or
regions, resulting in inconsistent hydrologic response to urbanization throughout the world (Table 2.3).
It seems that there is no truly predictable response of annual low flow, proportion of baseflow to total
streamflow, or groundwater recharge to urbanization, as demonstrated by the case studies outlined
below. Of the studies reviewed that directly address annual low flow response to urbanization, none
demonstrated a pronounced decrease in discharge (e.g. Harris and Rantz, 1964; Rose and Peters, 2001;
Konrad and Booth, 2002). Harris and Rantz (1964) attribute increased annual low flow to distribution
and leakage of imported water, an insight issued decades before most hydrologists accepted such a
source to be significant. Rose and Peters (2001) attribute the lack of annual low flow response in
Atlanta, Georgia, to an offsetting of the effects of impervious surface by reduced ET associated with
vegetation removal. Finally, Konrad and Booth (2002) interpret inconsistent annual low flow response
in the Puget Sound basin to varying degrees of development, implying that in some cases a development
threshold necessary to induce response has not been reached.
The response of baseflow proportion shows a weak tendency toward decline among the case
studies reviewed. Streams in Pennsylvania, New York, Georgia, and Orgeon, all demonstrated baseflow
reduction associated with urbanization (Leopold, 1968; Simmons and Reynolds, 1982; Rose and Peters,
21
2001; Chang, 2007). In all cases, the authors attribute observed declines to recharge loss associated
with impervious surface coverage, and Simmons and Reynolds (1982) additionally cite the removal of
wastewater from stream basins. In contrast, streams in Harlow, Great Britain, and southern New York
demonstrated baseflow increases with urbanization, presumably due to distribution and leakage of
imported water (Hollis, 1977; Burns et al., 2005). The wide variety of factors controlling baseflow
discharge and system response to urbanization likely explains the disagreement among these studies. A
lack of consistent results or no response was observed in the majority of the reviewed studies
addressing baseflow (Beran and Gustard, 1977; Ferguson and Suckling, 1990; Brandes et al., 2005;
Konrad and Booth, 2005). Explanations for the lack of clear trends include effects from pronounced
seasonality in the Pacific Northwest (Konrad and Booth, 2005), marked variability of background
conditions and specific impacts in the Mid‐Atlantic region (Brandes et al., 2005), and the offsetting of
rapid transmission of stormwater by distribution and leakage of imported water (Ferguson and Suckling,
1990).
Additional case studies were reviewed that address recharge to subsurface storage, as this is
inextricably linked with baseflow. Results from these studies generally indicate a more consistent
response to urbanization than seen with annual low flow or baseflow proportion. Four of the studies
reviewed, conducted in Caracas, Venezuela; Perth, Australia; Wolverhampton, UK; and northeastern
Illinois demonstrate increased recharge with urbanization (Appleyard et al., 1999; Hooker et al., 1999;
Seiler and Alvarado‐Rivas, 1999; Meyer, 2005). In all of these cases, recharge increases are attributed to
distribution of imported water and/or infrastructure leakage, with Appleyard et al. (1999) additionally
citing reduced ET as a factor. Decreases in recharge were observed in Long Island, New York (Koszalska,
1975); Atlanta, Georgia (Rose and Peters, 2001); and the Kleine Nete basin in Belgium (Dams et al.,
2008) attributed to export of wastewater in New York and reduced infiltration in the latter two studies.
Two studies in southern New York failed to demonstrate a clear direction of response to urbanization
22
(Ku et al., 1992; Burns et al., 2005). It is noteworthy that a larger percentage of recharge studies
demonstrated increase than was seen in the baseflow studies. The fact that increases in recharge were
slightly more common than increases in baseflow may indicate that urban manipulation detectibly
complicates the pathways between subsurface recharge and channel flow. However, the only study
that explicitly addressed both baseflow and recharge demonstrated the same direction of response in
both components (Rose and Peters, 2001), which suggests that the discrepancies seen among recharge
and baseflow studies may simply be further evidence of lack of consistent response to urbanization in
different settings.
Interpretation of baseflow response to urbanization is further complicated by several
considerations. Comparison of urban response across cities and regions is problematic, based on
differences in natural hydrologic background variability, unique infrastructure systems, and varied
management approaches. Research design and choice of parameters assessed is not universally
consistent, clouding cross‐study comparison. Investigators tend to seek clear trends in response to
urbanization, and in the process may overlook complex patterns associated with geographic variability
in physical setting, a point reinforced by more comprehensive analyses (e.g., Ferguson and Suckling,
1990; Rose and Peters, 2001; Konrad and Booth, 2005). Relatively intense, long‐term urbanization has
been the focus of most urban hydrology research, and far less is known about the impacts of lower‐
density or carefully mediated urban development. Land‐use activities associated with moderate impact
or episodic disturbance may not result in detectible stream response, given other background sources of
hydrologic variability (Konrad and Booth, 2002). The conceptual model outlined by Leopold (1968) does
not include consideration of these and other factors, and it unfortunately appears that baseflow
response to urbanization cannot be predicted by a highly simplified set of parameters.
23
2.3.3 Agriculture ‐ As seen with urbanization, baseflow response to agricultural land use may be positive
or negative, depending on management practices. First, there is the obvious confounding factor of
irrigation (Dow et al., 2007; He et al., 2009). If crops are irrigated from the surface water resources
linked to the stream network, increased ET may reduce baseflows. However, if irrigation water is drawn
from disconnected groundwater resources or from outside the drainage basin, increases in baseflow
may occur. Furthermore, varied management practices are associated with a wide range in associated
soil impacts (e.g., conventional tillage practices vs. no‐till and conservation tillage), differing temporal
patterns to intensive cropping (e.g., perennial vs. seasonal cultivation), and whether or not crop residue
or other soil cover are used during the fallow season (Kent, 1999). Drainage tiling may also have strong
impacts on baseflow in agricultural areas (Schilling and Helmers, 2008).
Accordingly, studies investigating baseflow response to agricultural land use have demonstrated
mixed results. Schilling and Libra (2003) showed that many Iowa rivers have seen increases in annual
baseflow and baseflow index, and additional work has shown that these increases were significantly
related to increasing row crop intensity (Schilling, 2005). Increases in baseflow over the past 60 years
within the upper Mississippi River basin have been attributed to reductions in ET associated with
conversion from perennial to seasonal cultivation (Ling and Slack, 2005; Zhang and Schilling, 2006), and
changes in tillage practices (Potter, 1991; Kent, 1999). Using rainfall simulation experiments, Rasiah
and Kay (1995) showed that minimum tillage practices were associated with lower overland flow and
increased infiltration compared with conventional tillage of corn crops in Canada. Charlier et al. (2008)
showed that greater overland flow in agricultural areas of Guadeloupe reduced recharge and decreased
baseflows. Decreased agricultural land use in Georgia and Wisconsin has been linked with increased
baseflows attributed to higher infiltration rates (Knox et al., 2001; Juckem et al., 2008), while large scale
conversion of forest to agricultural land in Thailand demonstrated no significant changes in baseflow
(Wilk et al., 2001). Overall, the literature addressing baseflow response to agricultural influence
24
demonstrates two main trends: Watersheds that have been under agricultural land use for extended
periods show baseflow increases in response to improved cropping and tillage practices. However, the
variety of management practices, variable uses and sources of irrigation, and other background sources
of variability prevent any clear influence of agricultural land use on baseflows when compared with
other land uses.
2.4. Summary and conclusions
Within a given setting of climate and geology, watershed topography and geomorphology
influence baseflow by affecting the storage properties and rates of water transmission within a
catchment. The influence of factors of slope, relief, and drainage density are particularly noteworthy.
However, it remains unclear whether these factors are themselves strong drivers of baseflow, or
whether they instead correlate to other aquifer properties that more directly control baseflow. More
research is needed to understand the role of subsurface topography on baseflow, and very little is
known about water storage in varied geomorphic units (e.g., colluvial deposits and alluvial bottomlands)
and their linkages to baseflow.
Research investigating anthropogenic controls on baseflow has tended to disproportionately
emphasize forestry experimentation and urbanization, and within these studies the natural background
controls on baseflow are often downplayed or ignored. Several recent studies emphasize the
importance of considering changes in soil hydrology when assessing streamflow response to land‐use
change. Very little is known about baseflow response to land‐use change in larger, more complex
systems, or in settings affected by development of moderate intensity, information which is essential for
effective water resources protection and management. It is increasingly clear that the results of forestry
experimentation studies demonstrating baseflow increase with forest removal should not be
extrapolated to more complex system with long‐term land‐use change and extensive soil disturbance. It
25
is difficult to draw overarching conclusions regarding the influence of watershed characteristics on
baseflow from the existing body of literature, given the enormous diversity of natural background
conditions, watershed parameters, and baseflow metrics among case studies. This highlights a clear
need for more studies investigating the relative influences of watershed geomorphology and land use
within a given natural template, and for efforts to be made toward developing consistent methodologies
for watershed characterization and baseflow quantification.
2.5. References
Anderson, J.G., Allen, P.M. and Bernhardt, G., 1993. A comprehensive surface‐groundwater flow model. Journal of Hydrology, 142: 47‐69. Anderson, M.G. and Burt, T.P., 1980. Interpretation of recession flow. Journal of Hydrology, 46: 89‐101. Appleyard, S.J., Davidson, W.A. and Commander, D.P., 1999. The effects of urban development on the utilisation of groundwater resources in Perth, Western Australia. In: J. Chilton (Editor), Groundwater in the urban environment ‐ selected city profiles. A.A. Balkerma, Rotterdam, pp. 97‐104. Barnes, P.L. and Kalita, P.K., 2001. Watershed monitoring to address contamination source issues and remediation of the contamination impairments. Water Science and Technology, 44(7): 51‐56. Beran, M.A. and Gustard, A., 1977. A study into the low‐flow characteristics of British rivers. Journal of Hydrology, 35: 147‐157. Beven, K.J. and Kirkby, M.J., 1979. A physically based, variable contributing area model of basin hydrology. Hydrological Sciences Bulletin, 24(1): 43‐69. Blodgett, J.C., Walters, J.R. and Borchers, J.W., 1992. Streamflow gains and losses and selected flow characteristics of Cottonwood Creek, north central California, 1982‐1985. U.S. Geological Survey Water‐Resources Investigation Report 92‐4009. 19 pp. Blöschl, G., 2001. Scaling in hydrology. Hydrological Processes, 15: 709‐711. Bosch, J.M. and Hewlett, J.D., 1982. A review of catchment experiments to determine the effect of vegetation changes on water yield and evapotranspiration. Journal of Hydrology, 55: 3‐23.
26
Boulton, A.J., 2003. Parallels and contrasts in the effects of drought on stream macroinvertebrate assemblages. Freshwater Biology, 48: 1173‐1185. Brandes, D., Cavallo, G.J. and Nilson, M.L., 2005. Base flow trends in urbanizing watersheds of the Delaware River basin. Journal of the American Water Resources Association: 1277‐1291. Brooks, K.N., Ffolliot, P.F., Gregersen, H. M., and Thames, J. L., 1991. Hydrology and the Management of Watersheds. Iowa State University Press, Ames, IA, 402 pp. Brown, A. E., Zhang, L., McMahon, T. A., Western, A. W., and Vertessy, R. A., 2005. A review of paired catchment studies for determining changes in water yield resulting from alterations in vegetation. Journal of Hydrology 310: 28‐61. Bruijnzeel, L. A., 2004. Hydrological functions of tropical forests: not seeing the soil for the trees? Agriculture, Ecosystems, and Enviroment, 104(1): 185‐228. Brutsaert, W., 2005. Hydrology: an introduction. Cambridge University Press, Cambridge, 605 pp. Bryan, R.B. and Jones, J.A.A., 1997. The significance of soil piping processes: inventory and prospect. Geomorphology, 20(3‐4): 209‐218. Burns, D., Vitvar, T., McDonnell, J., Hassett, J., Duncan, J. and Kendall, C., 2005. Effects of suburban development on runoff generation in the Croton River basin, New York, USA. Journal of Hydrology, 311: 266‐281. Burt, T.P. and Butcher, D.P., 1985. Topographic controls of soil moisture distribution. Journal of Soil Science, 36: 469‐486. Buttle, J.M., 1998. Fundamentals of small catchment hydrology. In: C. Kendall and J.J. McDonnell (Editors), Isotope tracers in catchment hydrology. Elsevier, Amsterdam, pp. 1‐49. Buttle, J.M., Hazlett, P.W., Murray, C.D., Creed, I.F., Jeffries, D.S. and Semkin, R., 2001. Prediction of groundwater characteristics in forested and harvested basins during spring snowmelt using a topographic index. Hydrological Processes, 15: 3389‐3407. Calder, L.R., 1990. Evaporation in the uplands. Wiley, Chichester, UK, 166 pp. Chang, M., 2003. Forest Hydrology: an Introduction to Water and Forests. CRC Press, Boca Raton, FL, 373 pp. Chang, H., 2007. Comparative streamflow characteristics in urbanizing basins in the Portland Metropolitan Area, Oregon, USA. Hydrological Processes, 21: 211‐222.
27
Chaplot, V. and Walter, C., 2003. Subsurface topography to enhance the prediction of the spatial distribution of soil wetness. Hydrological Processes, 17(13): 2567‐2580. Chaplot, V., Walter, C., Curmi, P., Lagacherie, P. and King, D., 2004. Using the topography of the saprolite upper boundary to improve the spatial prediction of the soil hydromorphic index. Geoderma, 123(3‐4): 343‐354. Charlier, J. B., Catlan, P., Moussa, R., and Voltz, M., 2008. Hydrological behavior and modeling of a volcanic tropical cultivated catchment. Hydrological Processes 22(22): 4355‐4370. Choi, W., 2004. Climate change, urbanisation, and hydrological impacts. International Journal of Global Environmental Issues, 4(4): 267‐286. Costa, M.H., Botta, A. and Cardille, J.A., 2003. Effects of large‐scale changes in land cover on the discharge of the Tocantins River, Southeastern Amazonia. Journal of Hydrology, 283: 206‐217. Dams, J., Woldeamlak, S. T., and Batelaan, O. 2008. Predicting land‐use change and its impact on the groundwater system of the Kleine Nete catchment, Belgium. Hydrology and Earth System Sciences 12: 1369‐1385. Davies, B.R., O'Keefe, J.H. and Snaddon, C.D., 1993. A synthesis of the ecological functioning, conservation, and management of South African river ecosystems. Water Research Commission Report 62/93, Pretoria, South Africa. 232 pp. Devito, K., Creed, I., Gan, T., Mendoza, C., Petrone, R., Silins, U., and Smerdon, B. 2005. A framework for broad‐scale classification of hydrologic response units on the Boreal Plain: is topography the last thing to consider? Hydrological Processes 19: 1705‐1714. Dewson, Z. S., James, A. B. W., Death, R. G., 2007. Journal of the North American Benthological Society, 26(4): 754‐766. Dodd, M.B. and Lauenroth, W.K., 1997. The influence of soil texture on the soil water dynamics and vegetation structure of a shortgrass steppe ecosystem. Plant Ecology, 133(1): 13‐28. Dow, C. L., 2007. Assessing regional land‐use/cover influences on New Jersey Pinelands streamflow through hydrograph analysis. Hydrological Processes 21(2): 211‐222. Doyle, M.W., Harbor, J.M., Rich, C.F. and Spacie, A., 2000. Examining the effects of urbanization on streams using indicators of geomorphic stability. Physical Geography, 21(2): 155‐181.
28
Dubé, S., Plamondon, A.P. and Rothwell, R.L., 1995. Watering‐up after clear‐cutting on forested wetlands of the St. Lawrence Lowlands. Water Resources Research, 31: 1741‐1750. Famiglietti, J.S., Rudnicki, J.W. and Rodell, M., 1998. Variability in surface moisture content along a hillslope transect: Rattlesnake Hill, Texas. Journal of Hydrology, 210(1‐4): 259‐281. Farvolden, R.N., 1963. Geologic controls on ground‐water storage and base flow. Journal of Hydrology, 1: 219‐249. Ferguson, B.K. and Suckling, P.W., 1990. Changing rainfall‐runoff relationships in the urbanizing Peachtree Creek watershed, Atlanta, Georgia. Water Resources Bulletin, 26(2): 313‐322. Freeze, R.A., 1972. Role of subsurface flow in generating surface runoff: 1. Base flow contributions to channel flow. Water Resources Research, 8(3): 609‐623. Gburek, W.J. and Folmar, G.J., 1999. Flow and chemical contributions to streamflow in an upland watershed: a baseflow survey. Journal of Hydrology, 217(1‐2). Galster, R. W. and Leprade, W. T. 1991. Geology of Seattle, Washington, United States of America. Bulletin of the Association of Engineering Hydrologists 28(3): 235‐302. Gustard, A., 1989. Compensation flows in the UK: a hydrological review. Regulated Rivers Resource Management, 3(1‐4): 49‐59. Gustard, A. and Wesselink, A.J., 1993. Impact of land‐use change on water resources: Balquhidder catchments. Journal of Hydrology, 145: 389‐401. Harr, R.D., Levno, A. and Mersereau, R., 1982. Streamflow changes after logging 130‐year‐old douglas fir in two small watersheds. Water Resources Research, 18(3): 644‐647. Harris, E.E. and Rantz, S.E., 1964. Effect of urban growth on streamflow regime of Permanente Creek, Santa Clara County, California. U.S. Geological Survey Water‐Supply Paper 1591‐B, Washington, D.C. Hatcher, R.D., 1988. Bedrock geology and regional geologic setting of Coweeta Hydrologic Laboratory in the eastern Blue Ridge. In: W.T. Swank and D.A. Crossley, Jr. (Editors), Forest Hydrology and Ecology at Coweeta. Springer‐Verlag, New York, pp. 81‐92. He, B., Wang, Yi., Takase, K., Mouri, G., and Razafindrabe, B. H. N. 2009. Estimating land use impacts on regional scale urban water balance and groundwater recharge. Water Resources Management 23: 1863‐1873.
29
Hibbert, A.R., 1967. Forest treatment effects on water yield. In: W.E. Sopper and H.W. Lull (Editors), Forest Hydrology. Pergamon, Oxford, UK, pp. 527‐543. Hicks, B.J., Beschta, R.L. and Harr, R.D., 1991. Long‐term changes in streamflow following logging in western Oregon and associated fisheries implications. Water Resources Bulletin, 27(2): 217‐226. Hollis, G.E., 1977. Water yield changes after the urbanization of the Canon's Brook Catchment, Harlow, England. Hydrological Sciences Bulletin, 22: 61‐75. Hooker, P.J., McBridge, D., Brown, M.J., Lawrence, A.R. and Gooddy, D.C., 1999. An integrated hydrological case study of a post‐industrial city in the West Midlands of England. In: J. Chilton (Editor), Groundwater in the urban environment ‐ selected city profiles. A.A. Balkerma, Rotterdam, pp. 145‐150. Hornbeck, J.W., Adams, M.B., Corbett, E.S., Verry, E.S. and Lynch, J.A., 1993. Long‐term impacts of forest treatment on water yield: a summary for northeastern USA. Journal of Hydrology, 150: 323‐344. Hursh, C.R. and Fletcher, P.W., 1943. The soil profile as a natural reservoir. Soil Science Society of America Proceedings, 7: 480‐486. Hutchinson, D.G. and Moore, R.D., 2000. Throughflow variability on a forested hillslope underlain by compacted glacial till. Hydrological Processes, 14(10): 1751‐1766. Illinois EPA, 2002. Quantity joins quality as a major water focus in Illinois. Environmental Progress, 27(1). http://www.epa.state.il.us/environmental‐progress/v27/n1/water‐focus.html. Iroumé, A., Huber, A. and Schulz, K., 2005. Summer flows in experimental catchments with different forest covers, Chile. Journal of Hydrology, 300(1‐4): 300‐313. Johnson, R., 1998. The forest cycle and low river flows: a review of UK and international studies. Forest Ecology and Management, 109: 1‐7. Jones, J.A. and Post, D.A., 2004. Seasonal and successional streamflow response to forest cutting and regrowth in the northwestern and eastern United States. Water Resources Research, 40: 1‐19. Jordan, J.P., 1994. Spatial and temporal variability of stormflow generation processes on a Swiss catchment. Journal of Hydrology, 153: 357‐382. Juckem, P. F., Hurt, R. J., Anderson, M. P., and Robertson, D. M. 2008. Effects of climate and land management on streamflow in the Driftless Area of Wisconsin. Journal of Hydrology, 355: 12‐30. Kent, C.A., 1999. The influence of changes in land cover and agricultural land management practice on baseflow in southwest Wisconsin, 1968‐1998. Ph.D. Dissertation, University of Wisconsin, Madison, WI.
30
Keppeler, E.T. and Ziemer, R.R., 1990. Logging effects on streamflow: water yield and summer low flows at Caspar Creek in northwestern California. Water Resources Research, 26(7): 1669‐1679. Kim, H.J., Sidle, R.C. and Moore, R.D., 2005. Shallow lateral flow from a forested hillslope: Influence of antecedent wetness. Catena, 60(3): 293‐306. Kirchner, J.W., Feng, X. and Neal, C., 2001. Catchment‐scale advection and dispersion as a mechanism for fractal scaling in stream tracer concentrations. Journal of Hydrology, 254: 82‐101. Klein, R.D., 1979. Urbanization and stream quality impairment. Water Resources Bulletin, 15: 948‐963. Knox, J.C., 2001. Agricultural influence on landscape sensitivity in the Upper Mississippi River Valley. Catena, 42(2‐4): 193‐224. Konrad, C.P., 2003. Effects of Urban Development on Floods. U.S. Geological Survey Fact Sheet 076‐03. 4 pp. Konrad, C.P. and Booth, D.B., 2002. Hydrologic trends associated wtih urban development for selected streams in the Puget Sound Basin, Western Washington. U.S. Geological Survey Water‐Resources Investigation Report 02‐4040, Tacoma, WA. Konrad, C.P. and Booth, D.B., 2005. Hydrologic changes in urban streams and their ecological significance. American Fisheries Society Symposium, 47: 157‐177. Koszalska, E.J., 1975. Water‐table on Long Island, New York, in March 1974 Long Island Water Resources Bulletin LIWR‐5. Kottegoda, N.T. and Natale, L., 1994. Two component log‐normal distribution of irrigation affected low flows. Journal of Hydrology, 158: 187‐199. Ku, H.F.H., Hagelin, N.W. and Buxton, H.T., 1992. Effects of Urban Storm‐Runoff Control on Ground‐Water Recharge in Nassau County, New York. Ground Water, 30(3): 507‐514. Lacey, G. C. and Grayson, R. B., 1997. Relating baseflow to catchment properties in South‐Eastern Australia. Journal of Hydrology 204: 231‐250. Larkin, R. G. and Sharp, J. M. 1992. On the relationship between river‐basin geomorphology, aquifer hydraulics, and ground‐water flow direction in alluvial aquifers. Geological Society of America Bulletin 104: 1608‐1620.
31
Leopold, L.B., 1968. Hydrology for Urban Land Planning ‐ A Guidebook on the Hydrologic Effects of Urban Land Use. U.S. Geological Survey Circular 554, Washington, D.C. 18 pp. Lerner, D.N., 1986. Leaking pipes recharge groundwater. Ground Water, 24(5): 654‐662. Lerner, D.N., 2002. Identifying and quantifying urban recharge: a review. Hydrogeology Journal, 10: 143‐152. Lindgren, G.A., Destouni, G. and Miller, A.V., 2004. Solute transport through the integrated groundwater‐stream system of a catchment. Water Resources Research, 40(W03511): 1‐13. Line, D. E. and White, N. M., 2007. Effects of development on runoff and pollutant export. Water Environment Research 79(2): 185‐190. Lins, H. F. and Slack, J. R., 2005. Seasonal and regional characteristics of US streamflow trends in the United States from 1940 to 1999. Physical Geography 26(6): 489‐501. Liu, Y.B., Pfister, L., Gebremeskel, S., DeSmedt, F. and Hoffman, L., 2003. A diffusive transport approach for flow routing in GIS‐based flood modeling. Journal of Hydrology, 238(1‐4): 91‐106. Marani, M., Eltahir, E., and Rinaldo, A. 2001. Geomorphic controls on regional base flow. Water Resources Research, 37(10): 2619‐2630. Mayden, R. L., 1987. Historical ecology and North American highland fishes: a research program in community ecology. In Matthews, W. J. and Heins, D. C. (eds.), Community and Evolutionary Ecology of North American Stream Fishes, University of Oklahoma Press, Norman, OK. McCulloch, J.S.G. and Robinson, M., 1993. History of forest hydrology. Journal of Hydrology, 150: 189‐216. McDonnell, J.J., Freer, J., Hooper, R., Kendall, C., Burns, D., Beven, K. and Peters, J., 1996. New method developed for studying flow on hillslopes. EOS, Transactions of the Americal Geophysical Union, 77: 465‐472. McGuire, K.J., McDonnell, J.J., Weiler, M., Kendall, C., McGlynn, B.L., Welker, J.M. and Seibert, J., 2005. The role of topography on catchment‐scale water residence time. Water Resources Research, 41(W05002): 1‐14. Meyer, S.C., 2002. Impacts of urbanization on base flow and recharge rates, Northeastern Illinois: summary of year 2 activities. In: Proceedings of the 12th Annual Research Conference: Research on Agricultural Chemicals and Groundwater Resources in Illinois, Makanda, IL.
32
Meyer, S.C., 2005. Analysis of base flow trends in urban streams, northeastern Illinois, USA. Hydrogeology Journal, 13(5‐6): 871‐885. Moore, R.D. and Thompson, J.C., 1996. Are water table variations in a shallow forest soil consistent with the TOPMODEL concept? Water Resources Research, 32(3): 663‐669. Mote, P.W., Parson, E., Hamlet, A.F., Keeton, W.S., Lettenmaier, D., Mantua, N., Miles, E.L., Peterson, D., Peterson, D.L., Slaughter, R. and Snover, A.K., 2004. Preparing for climatic change: The water, salmon, and forests of the Pacific Northwest. Climate Change, 61(1‐2): 45‐88. Neff, B. P., Day, S. M., Piggott, A. R. and Fuller, L. M., 2005. Base Flow in the Great Lakes Basin.U.S. Geological Survey Scientific Investigations Report, 2005‐2517. Novotny, V. and Olem, H., 1994. Water quality: prevention, identification, and management of diffuse polution. Van Nostrand Reinhold, New York City, 1054 pp. O'Geen, A.T., McDaniel, P.A., Boll, J. and Brooks, E., 2003. Hydrologic processes in valley soilscapes of the eastern Palouse Basin in northern Idaho. Soil Science, 168(12): 846‐855. Oke, T.R., 1979. Advectively‐assisted evapotranspiration from irrigated urban vegetation. Boundary Layer Meteorology, 17: 167‐173. Owen, M., 1991. Groundwater abstraction and river flows. Journal of the Institute of Water Environmental Management, 5(6): 697‐702. Parlange, M.B., Bryan, R.B., Steenhuis, T.S., Timlin, D.J. and Stagnitti, F., 1989. Subsurface flow above a fragipan horizon. Soil Science, 148(2): 77‐86. Perillo, C.A., Gupta, S.C., Nater, E.A. and Moncrief, J.F., 1999. Prevalence and initiation of preferential flow paths in a sandy loam with argillic horizon. Geoderma, 89(3‐4): 307‐331. Pilgrim, D.H., Cordery, I. and Baron, B.C., 1982. Effects of catchment size on runoff relationships. Journal of Hydrology, 58: 205‐221. Pirt, J. and Simpson, M., 1983. The estimation of river flows. Trent Water Authority, UK. 41 pp. Potter, K.W., 1991. Hydrological impacts of changing land‐management practices in a moderate‐sized agricultural catchment. Water Resources Research, 27(5): 845‐855. Price, K. and Jackson, C.R., 2007. Effects of forest conversion on baseflows in the southern Appalachians: A cross‐landscape comparison of synoptic measurements. Proceedings of the 2007
33
Georgia Water Resources Conference. <http://cms.ce.gatech.edu/gwri/uploads/proceedings/2007/2.3.4.pdf> Price, K., C. R. Jackson, and A. J. Parker, Variation of surficial soil hydraulic properties across land uses in the southern Blue Ridge Mountains, USA, submitted to J. Hydrol. Reuter, R.J. and Bell, J.C., 2003. Hillslope hydrology and soil morphology for a wetland basin in south‐central Minnesota. Soil Science Society of America Journal, 67(1): 365‐372. Rasiah, V., and Kay, B. D., 1995. Runoff and soil loss as influenced by selected stability parameters and cropping and tillage practices. Geoderma 68: 321‐329. Riggs, H.C., 1976. Effects of man on low flows. In: Proceedings of the Conference on Environment, Aspects Irrigation and Drainage, University of Ottawa. pp. 306‐314. Rodhe, A., Nyberg, L. and Bishop, K., 1996. Transit times for water in a small till catchment from a step shift in the oxygen 18 content of the water input. Water Resources Research, 32(12): 3497‐3511. Rose, S. and Peters, N.E., 2001. Effects of urbanization on streamflow in the Atlanta area (Georgia, USA): a comparative hydrological approach. Hydrological Processes, 15: 1441‐1457. Sahin, V. and Hall, M.J., 1996. The effects of afforestation and deforestation on water yields. Journal of Hydrology, 178: 293‐309. Santhi, C., Allen, P.M., Mattiahi, R. S., Arnold, J. G., and Tuppiad, P., 2008. Regional estimation of base flow for the conterminous United States by hydrologic landscape regions. Journal of Hydrology 351: 139‐153. Schaetzl, R.J. and Anderson, S., 2005. Soils: Genesis and Geomorphology. Cambridge, 817 pp. Schilling, K. E., 2005. Relation of baseflow to row crop intensity in Iowa. Agriculture, Ecosystems, and Environment, 105: 433‐438. Schilling, K. E. 2009. Investigating local variation in groundwater recharge along a topographic gradient, Walnut Creek, Iowa, USA. Hydrogeology Journal 17: 397‐407. Schilling, K. E. and Libra, R. D., 2003. Increased baseflow in Iowa over the second half of the 20th century. Journal of the American Water Resources Association, 39: 851‐860. Schilling, K. E. and Helmers, M. 2008. Effects of subsurface drainage tiles on streamflow in Iowa agricultural watersheds: Exploratory hydrograph analysis. Hydrological Processes 22(23): 4497‐4506.
34
Schulz, W. H., Lidke, D. J., and Godt, J. W. 2008. Modeling the spatial distribution of landslide‐prone colluvium and shallow groundwater on hillslopes of Seattle, WA. Earth Surface Processes and Landforms 33: 123‐141. Seiler, K.‐P. and Alvarado Rivas, J., 1999. Recharge and discharge of the Caracas aquifer, Venezuela. In: J. Chilton (Editor), Groundwater in the urban environment: selected city profiles. A.A. Balkerma, Rotterdam, pp. 233‐238. Shaman, J., Stieglitz, M. and Burns, D., 2004. Are big basins just the sum of small catchments? Hydrological Processes, 18(16): 3195‐3206. Sidle, R.C., Tsuboyama, Y., Noguchi, S., Hosoda, I., Fujieda, M. and Shimizu, T., 2000. Stormflow generation in steep forested headwaters: a linked hydrogeomorphic paradigm. Hydrological Processes, 14: 369‐385. Simmons, D.L. and Reynolds, R.J., 1982. Effects of urbanization on base flow of selected south‐shore streams, Long Island, New York. Water Resources Bulletin, 18(5): 797‐805. Sivapalan, M., 2003. Process complexity at hillslope scale, process simplicity at the watershed scale: is there a connection? Hydrological Processes, 17: 1037‐1041. Smakhtin, V.U., 2001. Low flow hydrology: a review. Journal of Hydrology, 240: 147‐186. Smith, R.E., 1991. Effect of clearfelling pines on water yield in a small eastern Transvaal catchment, South Africa. Water SA, 17(3): 217‐224. Soulsby, C., Rodgers, P.J., Petry, J., Hannah, D.M., Malcolm, I.A. and Dunn, S.M., 2004. Using tracers to upscale flow path understanding in mesoscale mountainous catchments: two examples from Scotland. Journal of Hydrology, 291: 174‐196. Soulsby, C., Tetzlaff, D., Rodgers, P., Dunn, S., and Waldron, S. 2006. Runoff processes, stream water residence times and controlling landscape characteristics in a mesoscale catchment: An initial evaluation. Journal of Hydrology 325: 197‐221. Sutherland, A. S., Meyer, J. L., and Gardiner, E. P., 2002. Effects of land cover on sediment regime and fish assemblage structure in four southern Appalachian streams. Freshwater Biology, 47: 1791‐1805. Swank, W.T., Swift, L.W., Jr. and Douglass, J.E., 1988. Streamflow changes associated with forest cutting, species conversions, and natural disturbances. In: W.T. Swank and D.A. Crossley, Jr. (Editors), Forest Hydrology and Ecology at Coweeta. Springer‐Verlag, New York, pp. 297‐312.
35
Tague, C. and Grant, G.E., 2004. A geological framework for interpreting the low‐flow regimes of Cascade streams, Willamette River Basin, Oregon. Water Resources Research, 40(W0403). Tetzlaff, D. and Soulsby, C. 2008. Sources of baseflow in larger catchments – Using tracers to develop a holistic understanding of runoff generation. Journal of Hydrology 359: 287‐302. Tetzlaff, D., Siebert, J., McGuire, K. J., Laudon, H., Burns, D. A., Dunn, S. M., and Soulsby, C. 2009. How does landscape structure influence catchment transit time across different geomorphic provinces? Hydrological Processes 23: 945‐953. Trimble, S.W., Weirich, F.H. and Hoag, B.L., 1987. Reforestation and the reduction of water yield on the southern piedmont since circa 1940. Water Resources Research, 23(3): 425‐437. Troch, P.A., Mancini, M., Paniconi, C. and Wood, E.F., 1993. Evaluation of a catchment scale water balance model. Water Resources Research, 29: 1805‐1817. Van Ommen, H.C., Hulsof, J., Van Den Heuvel, M., Dijksma, R., Hendrick, J.M.H. and Dekker, L.W., 1989. Experimental assessment of preferential flow paths in a field soil. Journal of Hydrology, 105(3‐4): 253‐262. Vogt, A. E. 2004. Responses of instream habitat and fishes to modest changes in forest cover in southeastern streams. M.S. Thesis, The University of Georgia, Athens, GA. Ward, R.C. and Robinson, M., 1990. Principles of Hydrology. McGraw Hill, Maidenhead, U.K., 365 pp. White, E.L., 1977. Sustained flow in small Appalachian watersheds. Journal of Hydrology, 32(71‐86). Wilk, J., Andersson, L. and Plermkamon, V., 2001. Hydrological impacts of forest conversion to agriculture in a large river basin in northeast Thailand. Hydrological Processes, 15(14): 2729‐2749. Wilkison, D.H. and Blevins, D.W., 1999. Observations on preferential flow and horizontal transport of nitrogen fertilizer in the unsaturated zone. Journal of Environmental Quality, 28(5): 1568‐1580. Wilson, G.V., Alfonsi, J.M. and Jardine, P.M., 1989. Spatial variability of saturated hydraulic conductivity of the subsoil of two forested watersheds. Soil Science Society of America Journal, 53(3): 679‐685. Woods, R.A., Sivapalan, M. and Robinson, J.S., 1997. Modeling the spatial variability of subsurface runoff using a topographic index. Water Resources Research, 33(5): 1061‐1073. Yeakley, J.A., Swank, W.T., Swift, L.W., Jr., Hornberger, G.M. and Shugart, H.H., 1998. Soil moisture gradients and controls on a southern Appalachian hillslope from drought through recharge. Hydrology and Earth System Sciences, 2(1): 41‐49.
36
Zhang, Y.‐K., and Schilling, K. E., 2006. Increasing streamflow and baseflow in Mississippi River since the 1940s: Effect of land use change. Journal of Hydrology, 324: 412‐422.
37
.Table 2.1 – Anthropogenic impacts on baseflow (Smakhtin, 2001)
Impact Attributed Effect Baseflow Response
Reference(s)
Groundwater abstraction lowers water tables decrease Owen, 1991 Wetland drainage accelerated removal of water from
valley bottoms decrease Riggs, 1976
Valley bottom vegetation change ET change, dependent on specific impact
increase or decrease
Swank et al., 1988; Keppeler and Ziemer, 1990
Catchment afforestation increased ET decrease Trimble et al., 1987; Gustard and Wesselink, 1993
Catchment forest harvest decreased ET increase Harr et al., 1982; Swank et al., 1988; Hicks et al.,1991
Catchment forest conversion decreased ET, increased infiltration increase or decrease
Wilk et al., 2001; Costa et al., 2003
River abstraction direct removal of water from channel decrease Kottegoda and Natale, 1994 Effluent discharge to rivers direct input of water to channel increase Pirt and Simmons, 1983 Irrigation return flow direct input of water to channel increase Blodgett et al., 1992; Dow et
al, 2007 Importation of water surface and subsurface water inputs increase Davies et al., 1993 Flow regulation channel impoundment with regulated
release increase or decrease
Gustard, 1989
38
Table 2.2– Recharge response to various effects of urbanization (Meyer, 2002)
INCREASED RECHARGE
Surface distribution of imported water (irrigation and other outdoor water use) Infrastructure leakage of imported water Stormwater detention Leakage of event water into shallow groundwater via storm sewers
DECREASED RECHARGE
Impervious surface coverage and soil compaction Rapid transmission of event water through storm sewers and modified channels Leakage of shallow groundwater into storm sewers Shallow groundwater withdrawal Removal of waste water outside of catchment
39
Table 2.3 – Summary of studies assessing the response of baseflow and recharge to urbanization
Location Response to Urbanization
Attributed mechanism(s) Reference
Atlanta, Georgia decrease reduced infiltration Rose and Peters, 2001 Coatesville, Pennsylvania decrease reduced infiltration Leopold, 1968 Long Island, New York decrease reduced infiltration + export of sewerage water Simmons and Reynolds, 1982 Portland, Oregon decrease reduced infiltration Chang, 2007 Atlanta, Georgia decrease reduced infiltration Rose and Peters, 2001 Long Island, New York decrease export of sewerage water Koszalska, 1975 Western Washington state inconsistent insufficient impact in some of the study basins Konrad and Booth, 2002 Western Washington state inconsistent seasonality effects Konrad and Booth, 2005 Delaware River Basin inconsistent varied influences among basins Brandes et al., 2005 Long Island, New York inconsistent seasonality effects Ku et al., 1992 Santa Clara County, California increase distribution and leakage of imported water Harris and Rantz, 1964 Southern New York state increase septic effluent Burns et al., 2005 Harlow, Great Britain increase Hollis, 1977 Caracas, Venezuela increase infrastructure leakage Siler and Alvarado‐Rivas, 1999 Northeastern Illinois increase distribution and leakage of imported water Meyer, 2005 Perth, Australia increase reduced ET + distribution and leakage of imported
water Appleyard et al., 1999
Wolverhampton, U.K. increase distribution and leakage of imported water Hooker et al., 1999 Atlanta, Georgia no response reduced infiltration offset by reduced summer ET Rose and Peters, 2001 Great Britain no response Beran and Gustard, 1977 Atlanta, Georgia no response reduced infliltration offset by distribution and leakage
of imported water Ferguson and Suckling, 1990
Southern New York state no response insufficient impact (suburban) Burns et al., 2005
40
CHAPTER 3
VARIATION OF SURFICIAL SOIL HYDRAULIC PROPERTIES ACROSS LAND USES IN THE SOUTHERN BLUE RIDGE MOUNTAINS, USA1
__________ 1Price, K., Jackson, C. R., and Parker, A. J. Submitted to Journal of Hydrology 9/27/2008
41
ABSTRACT
A full understanding of hydrologic response to human impact requires assessment of land‐use
impacts on key soil physical properties such as saturated hydraulic conductivity, bulk density, and
moisture retention. Such properties have been shown to affect watershed hydrology by influencing
pathways and transmission rates of precipitation to stream networks. Human land use has been shown
to influence these soil physical properties as a result of erosion, compaction, and pore structure
evolution. Our objective was to characterize soil physical properties under three land‐use classes (forest,
pasture, and managed lawn) in the southern Blue Ridge Mountains of southwestern North Carolina. A
total of 90 points were sampled (30 in each land use classes) throughout a 983 km2 study area.
Saprolitic and alluvial soils were emphasized, and sites were selected that showed consistent land use
history over a period of at least 30 years. Particle size distribution, in situ saturated hydraulic
conductivity (measured using an Amoozemeter compact constant head permeameter), bulk density, and
volumetric moisture content at field capacity were measured at each point. Forest soils demonstrated
markedly lower bulk densities and higher infiltration rates, and water holding capacities, than lawn and
pasture soils. No soil property significantly differed between pasture and lawn. Mean values for each
property were as follows (forest = F, lawn = L, pasture = P): saturated hydraulic conductivity (mmh‐1) –
F=63, L=7, P=8; bulk density (gcm‐3) – F=0.8, L=1.2, P=1.2; volumetric water holding capacity (%) –
F=72%, L=42%, P=39%. Particle size distributions did not significantly differ among land use classes or
parent materials, and the differences between the hydraulic properties of forest vs. nonforest soils were
attributed to compaction associated with land management practices. The magnitudes of differences
between forest and nonforest infiltration rates suggest that widespread conversion of forest to other
land uses in this region will be accompanied by decreased infiltration and increased overland flow,
potentially significantly altering water budgets and leading to reduced baseflows and impaired water
quality.
42
Keywords: soil hydrology, saturated hydraulic conductivity, infiltration, Amoozemeter, land use change,
Appalachian
3.1. Introduction
The inter‐related soil traits of texture, saturated hydraulic conductivity, bulk density, and
macroporosity influence hillslope and watershed hydrology (Farres, 1987; Rawls et al., 1993; Cerda,
1996). Through their effects on infiltration rates, these characteristics determine the proportion of
precipitation entering and retained in subsurface storage and the rates of transmission of water to
stream networks, thus affecting both stormflow production and baseflow maintenance (Hewlett, 1961;
Zimmerman et al., 2006; Tetzlaff et al., 2007). Land use practices have been shown to be of key
importance to soil hydrology, attributed to the effects of tillage, erosion, compaction, and pore
structure evolution (Rasiah and Kay, 1995; Harden, 2006). Such disturbances, in some cases, outweigh
genoform traits (e.g. those inherited from parent material, topographic setting, etc.) in determining soil
water movement (Schwartz et al., 2003; Zhou et al., 2008).
Compared with soils impacted by human land use, soils underlying native vegetation (e.g.,
undisturbed forest) generally feature low bulk density and high saturated hydraulic conductivity, total
porosity, and macroporosity, as a result of ample litter cover, organic inputs, root growth and decay, and
abundant burrowing fauna (Lee and Foster, 1991). In contrast, soils exposed to human impact are often
stripped of organic‐rich upper horizons and compacted by heavy equipment or livestock, increasing bulk
density and reducing infiltration rates (Celik, 2005; Li and Shao, 2006). In many cases, soils impacted by
land use change may demonstrate marked disparities from the original soil (Jiménez et al., 2006, Zhou et
al., 2008). Replacement of native vegetation with managed land cover is generally associated with
decreased rooting networks and faunal activity, thereby reducing the potential for well‐developed
macropore networks (Reiners et al.; 1994, Schwartz et al., 2003). The rooting systems of woody
43
vegetation such as forest and shrubland demonstrate substantially greater depth, diameter, dispersion,
and biomass than rooting systems of herbaceous plants or cultivated crops (Lee and Lauenroth, 1994;
Jackson et al. 1996; Messing et al. 1997). Conversion of native vegetation to managed land is also
commonly associated with decrease in litter accumulation and soil organic matter (Solomon et al., 1999;
Richter and Markewitz, 2001), which significantly influences soil water retention characteristics and soil
structure (Berglund, 1980; Buytaert et al., 2005; Harden, 2006). Such impacts have shown to be best
expressed in the upper portion of the soil column, with land‐use invariance in soil physical
characteristics repeatedly demonstrated at depths greater than approximately 15‐30 cm (e.g., Grossman
et al., 2001; Godsey and Elsenbeer, 2002; Li and Shao, 2006; Zhou et al., 2008).
Studies investigating soil physical response to land use change have heavily emphasized
comparison of cultivated cropland soils versus soils underlying native forest, shrubland, or grassland.
However, current development pressures in many regions of suburban and exurban growth do not
include conversion of native vegetation to cropland, and in fact are often associated with decline of
cultivated land (Richter and Markewitz, 2001; Gragson and Bolstad, 2006). Land use change in such
settings is likely to involve conversion of native or secondary vegetation to pasture or managed
turfgrass. As seen with cultivated soils, studies comparing soils under native vegetation to pasture or
lawn have also shown degradation of soil physical properties. A distinction is apparent between soils
underlying woody vegetation, such as forest or shrubland, versus herbaceous land cover, such as
pasture or grassland, even under varied management practices and degrees of compaction (Jiménez et
al., 2006).
Forest cover has been associated with lower bulk density and greater saturated hydraulic
conductivity than pasture in varied climates and parent materials throughout the world (Reiners et al.,
1994; Godsey and Elsenbeer, 2002; Jiménez et al., 2006; Li and Shao, 2006; Abbasi et al., 2007). Less is
known regarding soil physical response to conversion of native vegetation to turfgrass. Impacts
44
associated with residential development or the creation of golf courses, parks, ball fields, etc., typically
involve topsoil removal and/or compaction associated with grading and sod‐laying. It is probable that
observed soil changes in response to land use conversion to lawn grass are predominantly due to these
initial disturbances (Wigmosta, 1991; Hamilton and Waddington, 1999). Furthermore, the nature of
lawn grass and associated management do not encourage soil recovery post‐disturbance. Lawn grass
typically demonstrates shallow rooting depth, low organic matter accumulation, and is generally
associated with lower faunal activity than pasture or forest (Pizl and Josens, 1995). The few studies
addressing the physical properties of soils underlying lawn grass have shown exceptionally low
infiltration rates and high bulk densities, (Hamilton and Waddington, 1999; Oliveira and Merwin, 2001).
Comparison of soil physical properties across a land use gradient in Baltimore that included forest,
pasture, and managed lawn showed that soils underlying lawn grass demonstrated higher bulk density
and lower porosity than forest or pasture soils (Pouyat et al., 2007).
Cumulatively, these changes in soil physical characteristics associated with conversion of native
to managed vegetation reduce soil infiltration and storage capacities, possibly resulting in increased
overland flow and reduced subsurface storage. Along with factors such as increased road density and
impervious surface coverage, such soil changes are often important contributors to the flashier
hydrologic regimes typically associated with watersheds impacted by anthropogenic land cover change.
Decreased infiltration and increased Hortonian overland flow resulting from conversion of woody
vegetation to human land use has been demonstrated in many settings to be a direct consequence of
altered soil hydrology (Bens et al., 2007; Zimmermann et al., 2006; Ilstedt et al., 2007; Leblanc et al.,
2008). Booth et al. (2002) emphasized that the conversion of forests to lawns in urbanizing watersheds
caused substantial hydrologic change often neglected with respect to the effects of impervious surface
coverage. Given the increase in managed lawn associated with low‐ and medium‐density urban growth
45
occurring in many regions, it is of immediate importance to understand the effects of such land use
change on soil physical properties and the associated implications for watershed hydrology.
This research sought to determine the magnitudes of differences among soil physical properties
under three land uses (forest, pasture, and managed lawn) and across two parent materials (alluvium
and saprolite) in the southern Blue Ridge Mountains, a region currently experiencing pronounced
growth. Cultivated land was avoided due to its declining presence in the study area, the pronounced
cross‐site inconsistency in management practices, and the temporal variation in properties through the
cropping cycle. We attempted to avoid legacy effects from past land uses by limiting sites to those
demonstrating land use consistency for 30+ years. Particle size distribution, saturated hydraulic
conductivity, bulk density, total porosity, and water holding capacity were determined for 90 surface soil
samples all taken within one geologic unit and related to land use (forest, pasture, and lawn) and parent
material (saprolite and alluvium). Five sites were selected for each land cover/parent material
combination to represent regional variation and three points were sampled within each site to
represent local variation.
Variability in soil physical properties due to land use differences were expected to have resulted
from two mechanisms: 1) direct compaction by heavy equipment and/or livestock associated with
nonforest land uses, and 2) variation in macropore development, organic matter, and soil structure
associated with different vegetation types and associated fauna.
Thus it was expected that nonforest land use (pasture and lawn) would be associated with increased
bulk density and reduced saturated hydraulic conductivity, porosity, water holding capacity, and organic
carbon, and that soils underlying managed lawn would demonstrate greater contrast to forest soils than
would pasture soils. Understanding such differences is necessary for the development of a full
understanding of how land use change in this region is affecting watershed hydrologic processes, and is
46
additionally necessary for providing data for use in regional hydrologic modeling to forecast hydrologic
response to land‐use change.
3.2. Study area
This research was conducted in Macon and Jackson counties in southwestern North Carolina
(Figure 3.1). These counties are located within the Tallulah Falls thrust sheet of the East Flank Blue
Ridge lithotectonic belt (Robinson et al., 1992), a sub‐unit of the Blue Ridge physiographic province. All
bedrock types within the thrust sheet are crystalline, specifically consisting of intrusive igneous rocks
and varied metasedimentary assemblages, metamorphosed 350‐450 mya (Robinson et al., 1992;
Wooten et al., 2003). A sub‐area of Macon and Jackson counties underlain by a complex of biotite gneiss
and amphibolite will be emphasized in this study (Figure 3.1). Macon and Jackson counties are
characterized by moderate relief (540 to 1952 m), which is the product of bedrock weathering, fluvial
erosion, and mass wasting during the period of tectonic stability since the early Cenozoic (Leigh and
Webb, 2006). A saprolite mantle up to 30 m thick drapes the ridges and slopes throughout the study
area, and substantial deposits of colluvium are present on benches, coves, and footslopes (Hewlett,
1961; Hadley and Goldsmith, 1963; Southworth, 2003).
Soil parent materials in the study area are saprolite (75.0% by area), colluvium (19.4%), alluvium
(2.2%), and mixed colluvium/alluvium (1.3%), with the remaining 2.1% characterized by open water,
earthen fill, and mining pits (Soil Survey Staff – NRCS, 2007). Most soils are classified as udepts (60.8%
by area) forming at high elevations and in younger colluvial and alluvial landforms. Udults commonly
occur on saprolite backslopes of intermediate elevation and comprise 37.0% of the study area. Much
smaller areas of aqualfs, udalfs, aquepts, psamments, and udorthents are also present.
The climate is humid subtropical at the lowest elevations and marine humid temperate at higher
elevations, according to the Köppen classification. The 1971‐2000 average annual temperature at the
47
Coweeta Experiment Station (weather station elevation = 685.5 m), in the southern portion of the study
area, is 12.7 ◦C, with average January and July temperatures of 2.7 ◦C and 22.1 ◦C, respectively (NCDC,
2007). The 30‐year average annual precipitation is 183 cm, with a high monthly average of 20 cm
occurring in March (NCDC, 2007).
In the absence of human disturbance, regional land cover would be nearly 100% forest (Yarnell,
1998; Delcourt and Delcourt, 2004). Present‐day land use is predominantly forest, with nonforest land
cover occurring primarily as pasture and low‐density development (Table 3.1). Pronounced human
impact has occurred since Euro‐American settlement, with expansion of bottomland agriculture
occurring since the early nineteenth century (Gragson and Bolstad, 2006). The region experienced
intensive, widespread timber harvest and agriculture during the late nineteenth and early twentieth
century, followed by forest regrowth on mountain slopes (Davis, 2000). Agricultural land abandonment
and vegetation regrowth have been common since the 1960s, accompanied by population growth and
associated expansion of residential and low‐ to medium‐density urban landcover (Wear and Bolstad,
1998; Gragson and Bolstad, 2006). The largest town in the study area is Franklin, with a 2006 population
of 3618 (U.S. Census Bureau, 2007).
3.3. Methods
3.3.1 Site selection:
Ninety sites were identified for sampling, 30 within each of the three land use classes of forest,
lawn and pasture. All forest sites were under deciduous forest cover, pasture sites were open graminoid
fields either grazed by livestock or maintained for hay production, and lawn sites were managed
turfgrass. All sites were within the “biotite gneiss and amphibolite” unit on the regional 1:250 000‐scale
surficial bedrock map (Robinson et al., 1992), comprising a 983 km2 area within Macon and Jackson
Counties (Figure 3.1). Site selection was limited to an elevation range of 600‐800 m. Higher elevations
48
were not included to avoid significant cross‐site differences in temperature and precipitation associated
with elevation. Digital Light Detection and Ranging (LiDAR) coverage of each county (0.305 m vertical
resolution, 6.1 m pixel length) was obtained from the North Carolina Department of Transportation
(NCDOT, 2007a,b). From the parent materials within the study area, only saprolite and alluvium
were examined. Colluvium was excluded due to its characteristic textural heterogeneity. Sites were
evenly distributed among saprolite and alluvium, resulting in 15 sites in each land use class located
within each parent material (Table 3.2). Alluvial sites were within Dystrudepts (Rosman, Reddies, and
Cullowhee series), and saprolite sites were within Typic Hapludults (Cowee‐Evard and Evard‐Cowee
series complexes and Fannin series; Table 3.3). Digital soil coverages of Macon and Jackson counties
were obtained from the Soil Data Mart (USDA‐NRCS, 2005; USDA‐NRCS, 2007). The parent material of
each soil series in these counties was identified from the corresponding Offical Series Description (Soil
Survey Staff ‐ NRCS, 2007). The mapped series was verified by field profile descriptions. In a small
number of cases, the designated map unit was determined to be incorrect and a different series was
assigned based on profile description.
Alluvial sites were limited to flat, undissected portions of late‐prehistoric or historic terraces,
avoiding active floodplains or very old terrace surfaces. In order to avoid wide variability in insolation or
textural and morphological differences associated with hillslope position, saprolite sites were limited to
backslopes of generally south‐facing aspect (100‐260⁰) and similar gradient (15‐35%). LiDAR elevation
data were used to determine the aspect and gradient , which were field‐confirmed using a magnetic
compass and clinometer.
ArcGIS 9.2 was used to identify 30 locations meeting all site selection criteria (Figure 3.1). These
30 sites were evenly distributed among parent materials and land uses, with five sites in each parent
material/land use combination (e.g., five alluvial forest sites, five saprolite forest sites, five alluvial lawn
sites, etc.; Table 3.2). At each site, three randomly‐selected points at least 10 m apart were sampled
49
and treated as independent, resulting in a total of 90 sampling points. The minimum distance of 10 m
separating sampling points was determined following the design of similar studies (e.g. Jiménez et al.,
2006), and from the results of previous studies demonstrating the absence of spatial autocorrelation of
soil physical characteristics at distances greater than 1 m (Di et al., 1989; Lal, 1996; Webb et al., 2000;
Zhou et al., 2008). Analysis of 1:62 500‐scale 1970s aerial photography and published land use
classifications from 1992 and 2001 (USGS, 2000; USGS, 2003) confirmed consistency of land use at each
site over the past 30+ years.
Where possible, locations were identified where two land uses were adjacent within the same
soil unit. Ten such pairs of locations were identified: three pairs each of adjacent forest/lawn and
lawn/pasture locations and four pairs of adjacent forest/pasture locations. For the pair analyses, the
three sites were grouped to represent each location.
3.3.2 Field data collection methods:
At each sampling point, the organic matter was cleared from the mineral soil surface
surrounding the sampling point, and three 25‐cm deep, 5‐cm radius boreholes were dug within 50 cm of
the sampling point. A compact constant head permeameter (Amoozegar, 1989a) was used to measure
the infiltration rate in each borehole. These values were converted to saturated hydraulic conductivities
(Ksat) using the Glover Solution (Amoozegar, 1989b), and the geometric mean of the three Ksat values was
used to represent each sampling point. A ring corer was used to extract two undisturbed 331.3 cm3 core
samples per point (from depths 0‐7.5 cm and 7.5‐15 cm ) for laboratory analysis of bulk density, total
porosity, water holding capacity, and saturated hydraulic conductivity (for comparison with field Ksat).
Mineral soil bulk samples of 100 g were collected from the zones of 0‐7.5 cm and 7.5‐15 cm for particle
size analysis.
50
3.3.3 Laboratory methods:
A 50 g subsample of each bulk soil sample was crushed, passed through a 2 mm sieve, oven‐
dried, dispersed in sodium hexametaphosphate, and analyzed for particle size distribution by
hydrometer method (Gee and Bauder, 1986). Agro‐Services International performed laboratory
analyses of bulk density (ρB), water holding capacity, and saturated hydraulic conductivity (Ksat‐L). Ring
core samples of known volume were oven dried and weighed to determine ρB (Blake and Hartge, 1986).
Gravimetric water holding capacity was determined by saturating the soil core for 24 hours, following
standard methods of soil wetting for determination of the initial drainage curve (Klute, 1986). Cores
were allowed to drain for 24 hours, and gravitational moisture content at field capacity was determined
by subtracting the drained core weight from the saturated weight. The gravitational moisture content
(GMC) was converted to volumetric moisture capacity at field capacity (VMCfc) by VMCfc = GMC/ρB. Ksat‐L
was determined using the constant head method (Klute and Dirksen, 1986). Total porosity (φT) was
calculated from the ρB: φT = 1 – ρB/ρP, where ρP = particle density. Particle density was assumed to be
2.65 gcm‐3 (Danielson and Sutherland, 1986; Li and Shao, 2006).
3.3.4 Statistical Analyses:
Two‐way analysis of variance (ANOVA) was used to determine the relative roles of parent
material and land‐use class on soil hydrologic properties. One‐way ANOVA was used to evaluate
differences in soil physical characteristics as a function of land use and to test the variability among sites
within a given land use. Normality was tested using the Kolmogorov‐Smirnov test. Standard statistical
transformations (log10, inverse, and square root) were used to achieve normal distributions where
possible. In cases where such transforms failed to normalize a given parameter, the nonparametric
ANOVA on Ranks test was used. Particle size data (as fractions summing to unity for each point) were
arcsine‐square root transformed prior to statistical analyses. T‐tests or nonparametric Mann‐Whitney
51
Rank Sum tests were performed to test significance of difference between mean values of parameters
for each parent material (alluvium vs. saprolite), and to test pairwise differences between the land use
classes of forest, lawn, and pasture. Paired t‐tests or nonparametric Wilcoxon signed rank tests were
used to compare means of the upper and lower cores. For all tests, a threshold of p < 0.05 was used to
define statistical significance. All statistical analyses were performed using SigmaStat 3.5.
3.4. Results
The physical characteristics of forest soils clearly and strongly differed from pasture and lawn
soils in this study area. Forest soils demonstrated significantly lower ρB and higher φT, Ksat, and VMCfc
than soils in the other two land uses (Figure3.2). For no parameter did pasture and lawn soils
significantly differ from each other. Soil texture was very similar among parent materials and land‐use
classes, removing the need for separate analyses for the two parent materials.
3.4.1 Particle size:
The sand, silt, and clay percentages of the surface soil (average of 0‐7.5 and 7.5‐15 cm samples)
ranged from 29‐47%, 27‐39%, and 26‐32% among all 90 sites. The vast majority (83 of 90 sites) classified
as sandy loam or loam, with the remainder falling into loamy sand, silt loam, and clay loam (Figure 3.3).
On average, the bottom core (7.5‐15 cm) contained slightly less sand and silt than the upper core, and
contained an average 2% more clay (Table 3.4). Differences in silt and clay between the upper and lower
cores were statistically significant, whereas differences in sand content were not. Differences in the
sand, silt, and clay percentages among forest, lawn, and pasture soils were not significant, nor were
textural differences among the parent materials (Table 3.5). The similarity of the particle size
distributions among these samples allows comparison of the physical characteristics of ρB, φT, Ksat, and
52
VMCfc as a function of land use without concerns that systematic textural differences may be
complicating the relationships.
3.4.2 Bulk density and porosity:
The ρB of all soil sites ranged widely from 0.41 to 1.51 gcm‐3 in the upper 7.5 cm and from 0.72
to 1.66 gcm‐3 in the 7.5‐15 cm depth. Overall, the mean ρB of the upper cores was significantly lower
than the lower cores (1.10 vs. 1.30 gcm‐3, T = 6349.0, p < 0.001). The φT ranged from 42.9 to 84.5% in the
upper core and 37.2 to 72.9% in the lower core. The mean φT values of the upper and lower cores were
significantly different (58.4 vs. 50.9%, t = 5.77, p < 0.001). The ρB and φT of alluvial and saprolite soils
were highly similar and did not significantly differ at either depth (ρB from 0‐15cm: 1.18 vs. 1.22 gcm‐3, T
= 1930.5, p = .347; φT from 0‐15 cm: 55.5 vs. 53.9%, t = ‐0.87, p = 0.388).
One‐way ANOVA demonstrated that ρB and φT significantly differed among the land use classes
at both depths (Table 3.5), and pairwise comparisons indicate significant differences between forest
soils and the other two land uses at both depths. The average ρB of the upper and lower cores in forest
soils was 0.96 gcm‐3, 38% lower than lawn and pasture soils, which were essentially equal (1.33 and 1.32
gcm‐3. Correspondingly, the average φT of lawn and pasture soils did not significantly differ (49.9 vs
50.3%; T = 923.5; p = 0.906), but the φT of forest soils was significantly higher than the lawn and pasture
soils.
3.4.3 Saturated hydraulic conductivity:
A paired t‐test indicated that the mean saturated hydraulic conductivities determined by the in
situ method (Ksat) and laboratory method ( Ksat‐L) were significantly different (mean = 33 vs 55 mmh‐1; t =
6.59; p < 0.001). Because the field measurement represented the upper 25 cm, the average of the upper
and lower core laboratory measurements was used for comparison with the field measurement (Figure
53
3.4). The laboratory method indicated a larger range in conductivites than the field method (2.18 –
327.83 mmh‐1 compared with 1.06 – 197.21 mmh‐1). The field Ksat values correlated more strongly with
the ρB and VMCfc than did the laboratory measurements (r = ‐0.63 vs. ‐0.67 with ρB, r = 0.49 vs. 0.59 for
VMCfc), despite the fact that Ksat‐L, ρB, and VMCfc were all measured from the same core sample. For this
reason, field Ksat is emphasized herein. The average Ksat‐L of the upper core was nearly twice as great as
the lower core (75 vs. 35 mmh‐1; t = 6.86; p < 0.001).
Forest soils demonstrated far greater Ksat than lawn or pasture soils by both field and lab
methods (Table 3.5). The average field Ksat of the forest soils was approximately seven times greater
than the lawn and pasture soils, which were equivalent (forest = 77 mmh‐1, lawn = 11 mmh‐1, pasture =
12 mmh‐1). ANOVA results indicate significant difference in Ksat among the land uses, with pairwise
results indicating that forest soils had significantly higher Ksat than lawn and pasture soils, which did not
significantly differ (Table 3.5).
3.4.4 Water holding capacity:
VMCfc ranged from 37.7 to 74.0% in the upper core and from 32.6 to 87.8% in the lower core.
The VMCfc of the upper core was significantly greater than that of the lower core (50.7 vs. 46.2%; t =
7.36; p < 0.001). As seen with the other variables, the VMCfc of forest soils differed significantly from
pasture and lawn soils, which again did not significantly differ from each other (Table 3.5). Forest soil
VMCfc was nearly 10% greater than pasture and lawn soils (54.8 vs 45.4 and 45.2%).
3.4.5 Paired locations:
Pair comparisons corroborate the results using all 90 sites (Table 3.6). In all forest/pasture and
forest/lawn pairs, the forest soils demonstrated lower ρB and higher φT, Ksat, and VMCfc than the
nonforest soils. In no pair was the directionality of difference reversed. Differences within
54
forest/nonforest pairs were of similar magnitude to those reported above, with the greatest contrasts
between forest and pasture soils. There were very few pronounced within‐pair differences for the
lawn/pasture pairs, and there was no consistency in directionality of difference between lawn and
pasture sites for any parameter.
3.4.6 Site Variability of Ksat:
As a correlate of the other physical parameters, field Ksat was used to evaluate the assumption of
independence among the sampling points within each site. Within a given land use, the variability
among sites was not drastically greater than seen among the three points within a given site (Figure
3.5). ANOVA on ranks indicated that variability among lawn sites was not statistically significant (H =
11.38; p = 0.251). While forest and pasture sites did demonstrate statistically significant differences
among sites (forest: H = 19.065; p = 0.025; pasture: H = 19.146; p = 0.024), in both land uses the
exclusion of the site with the highest mean Ksat value resulted in a lack of statistically significant
variability among the sites. However, exclusion of these sites does not negate the statistical significance
of the ANOVA used to evaluate the differences in Ksat as a function of land use. Thus, the treatment of
all 90 sites as independent was deemed to be justifiable.
3.5. Discussion
The results of this study indicate a clear distinction between the hydraulic properties of forest
vs. lawn and pasture soils. This is especially noteworthy given the similarity of textures among the soils
included in this study. While it was hypothesized that lawn soils would show a greater distinction from
forest soils than would pasture soils, both nonforest land uses exhibited remarkably similar soil physical
characteristics. The differences observed in the Ksat, ρB, φT, and VMCfc between forest and nonforest soils
are interpreted to have resulted from a combination of land management and differences in macropore‐
55
forming biotic activity. Pasture sites have likely experienced compaction by livestock and/or heavy
equipment, and lawn site preparation generally involves compaction or removal of topsoil. Forest soils
typically demonstrate a far greater presence of woody roots and burrowing fauna, resulting in well‐
developed macropore networks (Messing et al., 1997). Such networks can have a profound impact on
soil conductivities. Several studies have indicated a distinction between soils underlying woody vs.
herbaceous vegetation, with shrubs and trees supporting much higher macroporosity and Ksat than even
native, unmanaged grassland (Godsey and Elsenbeer, 2002; Jiménez et al., 2006; Li and Shao, 2006). The
results of this study corroborate this distinction. While there are many examples of comparative studies
between native vegetation and cultivated soils and between forest and pasture soils, far less
information is available for turfgrass soils. There is a clear need for a more comprehensive
understanding of the hydrologic effects of forest conversion to turfgrass, given the suburban and
exurban development pressures facing many areas of the world.
The results of this study demonstrate pronounced magnitudes of difference between forest and
nonforest soils, and the discussion that follows compares the magnitudes of difference for the individual
parameters with differences between forest and nonforest soils in other studies.
3.5.1 Bulk Density:
In this study, nonforest soils had average ρB as much as 38% higher than forest soils. While a few
studies have demonstrated only minor differences between the ρB of forest vs. pasture or grassland soils
(e.g. Agnihotri and Yadav, 1995; Celik, 2005), a greater number of studies have indicated large
differences under such land covers. Over a range of 0‐15 cm depth, Reiners et al. (1994) reported an
average ρB of 0.69 gcm‐3 under primary forest soils, in contrast to 0.80 gcm‐3 under active pasture soils in
the Costa Rican rainforest. Harden (2006) observed ρB of soils under grass cover five times greater than
those under forest cover in the Ecuadorian parámo. Statistically significant differences of smaller
56
magnitude have been observed in the Himalayan foothills of Pakistan (Abbasi et al., 2007) and within
three of four studied soil series in Pennsylvania (Zhou et al., 2008). Livestock grazing has been shown to
directly reduce soil ρB in Argentina (Cisneros et al., 1999.
Fewer studies have investigated ρB differences between turfgrass and other land uses, but
differences have been demonstrated in several regions. Pouyat et al. (2007) found ρB to be one of the
most influential factors for statistically distinguishing forest vs. turfgrass land cover in the Baltimore
metropolitan area, despite the minor difference in magnitude (average forest ρB = 1.1 gcm‐3, average
park/golf course/residential/institutional turfrass = 1.2‐1.3 gcm‐3. Zhou et al (2008) found that
woodland soils demonstrated lower ρB than urban soils in Pennsylvania.
3.5.2 Saturated hydraulic conductivity:
On average, forest soils in this study area demonstrated Ksat values approximately seven times
greater than pasture and lawn soils. Several other studies have shown similar magnitudes of difference
between forest and pasture. Godsey and Elsenbeer (2002) and Zimmerman et al. (2006) exhibited order
of magnitude differences between the near‐surface (0‐12.5 cm) conductivities of forest and pasture soils
in Brazil (250 vs. 15 mmh‐1 and 206 vs. 26 mmh‐1, respectively). A similar magnitude of difference was
demonstrated in Peru, where grazed pasture soils were characterized by an average Ksat of 41 mmh‐1,
compared with 420 mmh‐1 observed in forest soils (Allegre and Cassel, 1996). An even greater
magnitude of difference was demonstrated in Colombia, where the average Ksat values of fine‐textured
forest and pasture soils were 143 vs. 2 mmh‐1 and Ksat values of coarse‐textured forest and pasture soils
were 159 vs. 8 mmh‐1 (Martinez and Zinck, 2004). The results of the Martinez and Zinck (2004) study are
particularly noteworthy, as they indicate the land use signature on soil Ksat is independent of soil texture.
Significantly greater Ksat of forest vs. pasture soils has also been demonstrated in the Loess Plateau of
China (Li and Shao, 2006) and in Nigeria (Ghuman et al., 1991), though not of as great a magnitude as
57
demonstrated by the southern Blue Ridge soils or the aforementioned studies. A review of 14
afforestation studies in the tropics showed an average three‐fold increase in infiltration capacity
compared with previous disturbed conditions (Ilstedt et al., 2007).
While pronounced differences in the hydraulic conductivities of forest vs. pasture soils have
been shown throughout the world, several studies have shown a lack of significant difference between
forest and pasture or grassland soils (e.g. Messing et al., 1997; Celik, 2005; Zhou et al., 2008). This
discrepancy may be a result of complications of legacy effects of prior land use. The lack of difference
could also be the result of the wide range of impacts characterizing a pasture or grassland site, e.g.,
whether or not the site has been exposed to livestock grazing or heavy equipment. Soil compaction by
such mechanisms has been shown to reduce infiltration rates and conductivities (Agnihotri and Yadav,
1995; Allegre and Cassell, 1996).
While there is a large body of literature addressing differences in Ksat between forest and
pasture sites, far fewer comparative values exist for forest and lawn sites. Zhou et al. (2008) reported a
lack of significant difference between the conductivities of lawn and forest soils in Pennsylvania. The
authors speculate the lack of statistical significance of differences may have resulted from pronounced
temporal variability and from interactions between land use and other independent variables. Although
few studies have compared lawn soils to native land uses directly, several studies have indicated very
low conductivities of lawn soils, of similar magnitude to those seen in this study (Wigmosta, 1991;
Hamilton and Waddington, 1999; Oliviera and Merwin, 2001).
3.5.3 Water holding capacity:
Despite well‐documented correlations between water holding capacity and other soil physical
parameters, there is much less evidence for land‐use dependence of water holding capacity than seen
with Ksat, ρB, or φT. This study found consistent and significantly higher VMCfc in forest than pasture and
58
lawn soils, by a factor of nearly 20%. However, several studies have demonstrated no significant
differences between the water holding capacity at field capacity of disturbed and undisturbed soils (e.g.,
Jusoff, 1989). Harden (2006) showed substantially greater water holding capacity of páramo grassland
vs. forest soils in the Ecuadorian Andes. Páramo soils were shown to have VMC three times greater than
the forest soils, despite the fact that the grassland soils demonstrated five times greater ρB. Soils in the
Loess Plateau of China had equivalent gravimetric water content at field capacity among shrubland,
forest, and grassland cover (Li and Shao, 2006).
3.5.4 Implications for altered water budgets:
The results of this study indicate pronounced reductions in Ksat with nonforest land use in the
southern Blue Ridge Mountains. This region is currently experiencing development pressures associated
primarily with exurban growth, and housing density is expected to increase dramatically in coming
decades (Cho et al., 2003; Gragson and Bolstad, 2006). Land use in Macon and Jackson counties is
predominantly forest (83% in 2001), and an increase in housing density will inevitably be associated with
a decline in forest cover. The results of this study indicate that significant changes in watershed land use
from forest to turfgrass will be associated with major alterations to watershed water budgets. The
decreased hydraulic conductivity will likely be associated with increased Hortonian overland flow. The
rainfall intensity/duration/frequency (IDF) curves from Franklin, NC confirm this likely scenario (Figure
3.6). With Ksat representing the lower bound of the soil infiltration rate, the IDF curves show that, while
the mean Ksat of forest soils is high enough to accommodate all but very infrequent and short duration
storm events, rainfall intensities commonly exceed the mean Ksat values of lawn and pasture soils.
Storms of intensities exceeding the conductivities of lawn and pasture soils also commonly persist for
significant durations.
59
Changes in the proportion of forest and nonforest land use within southern Blue Ridge
watersheds will be associated with increased overland flow and decreased times of water transmission
to stream networks. Such changes carry important implications for increased flood hazards, greater
contaminant and nutrient transport to streams, surface erosion, and increased stream temperatures.
Correspondingly, increased overland flow is associated with reduced subsurface recharge and decreased
baseflows, consequences of which include reduced water supply, increased concentration of
contaminants in streams, and impaired instream aquatic habitat. This is a particular threat in the
crystalline terrain of the southern Blue Ridge, where there is no significant bedrock aquifer supplying
baseflow to streams – in this region, the soil and saprolite mantle is the predominant source of
sustained streamflow (Hewlett, 1961; Velbel, 1985). Furthermore, as indicated by Harden (2006),
decreases in the surface infiltration capacities of soils in mountainous terrain are particularly
pronounced, because of the rapid rates of transmission of overland flow to stream systems in steep
terrain.
Land use impacts on soil hydrology have been shown to influence watershed processes in
several studies. Harden (2006) showed that human activities in the Ecuador páramo have altered the
timing and distribution of infiltration and runoff, specifically attributed to soil compaction. Increased
overland flow was observed in King County, Washington, associated with turfgrass land use, attributed
to topsoil removal and construction compaction (Wigmosta, 1991). Bens et al. (2007) determined that
soils play a critical role for water retention and overland flow prevention for flood control in the German
lowlands. It is clear from this and other studies that the flashier hydrologic regimes typically associated
with increased impervious surface in human impacted areas are likely partially due to soil alteration
associated with land use change, and that there is a serious need to address such impacts when
evaluating or predicting hydrologic response to land use change.
60
3.6. Conclusions
Soils in the southern Blue Ridge exhibit marked differences in physical characteristics under
forest and nonforest land uses (pasture and lawn). Soil particle size distributions do not differ
significantly among the parent materials or land uses, and soil hydraulic properties do not differ
significantly between alluvium and saprolite soils. Among both parent materials, forest soils had
significantly lower bulk density and higher saturated hydraulic conductivity and water holding capacity
than pasture and lawn soils, which did not significantly differ from each other. The mean saturated
hydraulic conductivity among forest sites was approximately seven times greater than in pasture and
lawn soils. Nonforest soils had hydraulic conductivities lower than the rainfall intensities of common,
long‐duration storms in the region, and nonforest soils also had reduced water holding capactities.
Accordingly, altered water budgets and increased Hortonian overland flow should be expected to
accompany continued land use change in the southern Blue Ridge. These results strongly support the
concept that soil modification is a significant driver of the watershed hydrologic changes of increased
floods and reduced baseflows observed with land use change.
3.7. Acknowledgements
Funding was provided by the U.S. Environmental Protection Agency – Science to Achieve Results
(STAR) Fellowship program, National Science Foundation (NSF) Doctoral Dissertation Improvement
Award BCS‐0702857, and The University of Georgia (UGA) Research Foundation. Additional support was
generously provided by the Coweeta LTER (NSF Cooperative Agreement DEB‐0218001), for which the
authors would especially like to thank Dr. Ted Gragson of the UGA Department of Anthropology.
Equipment was provided by the following UGA labs: the Soil Lab of the WSFNR, the Soil Morphology Lab
of the Department of Crop and Soil Science, and the Geomorphology Lab in the Department of
Geography. Field assistance was provided by Clint Collins and Julia Ruth. Insightful feedback during the
61
course of this research was provided by Dr. David S. Leigh of the UGA Department of Geography, Dr.
Lawrence A. Morris of the UGA Warnell School of Forestry and Natural Resources (WSFNR) and Dr. Larry
T. West of the Natural Resource Conservation Service. Dr. Todd Rasmussen of the UGA WSFNR and Dr.
George A. Brook and Dr. Marguerite Madden of the UGA Department of Geography reviewed a draft of
this manuscript. The authors express tremendous gratitude to the numerous private landowners
throughout Macon Co. and Jackson Co., NC, who allowed access and disturbance to their property for
the purposes of this study. Dr. James Vose of the U.S. Forest Service Coweeta Hydrologic Laboratory also
granted site use permission for this research.
3.8. References
Abbasi, M. K., Zafar, M., Khan, S. R., 2007. Influence of different land‐cover types on the changes of selected soil properties in the mountain region of Rawalakot Azad Jammu and Kashmir. Nutrient Cycling in Agroecosystems 78, 97‐110. Agnihotri, R. C., Yadav, R. C., 1995. Effects of different land uses on infiltration in ustifluvent soil susceptible to gully erosion. Hydrological Sciences 40 (3), 395‐406. Allegre, J. C., Cassel, D. K., 1996. Dynamics of soil physical properties under alternative systems to slash‐and‐burn. Agriculture, Ecosystems, and Environment 58, 39‐48. Amoozegar, A. 1989a. A compact constant‐head permeameter for measuring saturated hydraulic conductivity of the vadose zone. Soil Science Society of America Journal 53, 1356‐1361. Amoozegar, A. 1989b. Comparison of the Glover Solution with the simultaneous‐equations approach for measuring hydraulic conductivity. Soil Science Society of America Journal 53, 1362‐1367. Bens, O., Wahl, N. A., Fischer, H., Hϋttl, R. F., 2007. Water infiltration and hydraulic conductivity in sandy cambisols: impacts of forest transformation on soil hydrologigal properties. European Journal of Forest Research 126, 101‐109. Berglund, E. R., Ahyoud, A., Tayaa, M. H., 1980. Comparison of soil and infiltration properties of range and afforested sites in northern Morocco. Forest Ecology and Management 3, 295‐306. Blake, G. R., Hartge, K. H. 1986. Particle density. In: A. Klute (Ed.), Methods of Soil Analysis: Part 1 ‐ Physical and Mineralogical Methods. 2nd edition, Soil Science Society of America, Madison, WI, pp. 377‐382.
62
Bonnin, G.M., Martin, D., Lin, B., Parzybok, T., Yekta, M., Riley, D. 2004. Point precipitation frequency estimates from NOAA Atlas 14: Frankin, NC. Accessed April 11, 2008. <http://hdsc.nws.noaa.gov/hdsc/pfds/orb/nc_pfds.html> Booth, D. B., Hartley, D., Jackson, R., 2002. Forest cover, impervious surface area, and mitigation of stormwater impacts in King County, WA. Journal of the American Water Resources Association 38 (3), 835‐846. Buytaert, W., Wyseure, G., De Bièvre, B., Deckers, J., 2005. The effect of land‐use change on the hydrological behaviour of Histic Andisols in south Ecuador. Hydrological Processes 19, 3985‐3997. Celik, I., 2004. Land‐use effects on organic matter and physical properties of soils in a southern Mediterranean highland of Turkey. Soil and Tillage Research 83, 270‐277. Cerda, A., 1996. Soil aggregate stability in three Mediterranean environments. Soil Technology 9 (3), 133‐140. Cho, S., Newman, D. H., Wear, D. N., 2003. Impacts of second home development on housing prices in the southern Appalachian highlands. Review of Urban and Regional Development Studies 15 (3), 208‐225. Cisneros, J. M., Cantero, J. J., Cantero, A., 1999. Vegetation, soil hydrophysical properties, and grazing relationships in saline‐sodic soils of Central Argentina. Candadian Journal of Soil Science 79 (3), 399‐409. Danielson, R.E., Sutherland, P.L. 1986. Porosity. In: A. Klute (Ed.), Methods of Soil Analysis: Part 1 ‐ Physical and Mineralogical Methods. 2nd edition, Soil Science Society of America, Madison, WI, pp. 443‐462. Davis, D. E. 2000. Where there are Mountains: An Environmental History of Appalachia. University of Georgia Press, Athens, GA. Delcourt, P. A., Delcourt, H. R. 2004. Prehistoric Native Americans and Ecological Change: Human Ecosystems in Eastern North America since the Pleistocene. Cambridge University Press, Cambridge. Di, H.J., Kemp, R.A., Trangmar, B.B. 1989. Use of geostatistics in designing sampling strategies for soil survey. Soil Science Society of America Journal 53, 1163‐1167. Elliott, E. T., 1986. Aggregate structure and carbon, nitrogen, and phosphorous in native and cultivated soils. Soil Science Society of America Journal 50, 627‐633. Farres, P. J., 1987. The dynamics of rainsplash erosion and the role of soil aggregate stability. Catena 14, 119‐130. Gee, G. W., Bauder, J. W. 1986. Particle‐size analysis. In: A. Klute (Ed.), Methods of Soil Analysis: Part 1 ‐ Physical and Mineralogical Methods. 2nd edition, Soil Science Society of America, Madison, WI, pp. 383‐412.
63
Ghuman, B. S., Lal, R., Shearer, W., 1991. Land clearing and use in the humid Nigerian tropics: I. Soil Physical Properties. Soil Science Society of America Journal 55, 178‐183. Godsey, S., Elsenbeer, H., 2002. The soil hydrologic response to forest regrowth: a case study from southwestern Amazonia. Hydrological Processes 16, 1519‐1522. Gragson, T. L., Bolstad, P., 2006. Land use legacies and the future of southern Appalachia. Society and Natural Resources 19 (2), 175‐190. Grossman, R. B., Harns, D. S., Seybold, C. A., Herrick, J. E., 2001. Coupling use‐dependent and use‐invariant data for soil quality evaluation in the United States. Journal of Soil and Water Conservation 56, 63‐68. Hadley, J. B., Goldsmith, R.,1963. Geology of the eastern Great Smoky Mountains, North Carolina and Tennessee. U.S. Geological Survey, Professional Paper No. 349‐B. Hamilton, G. W., Waddington, D. V., 1999. Infiltration rates on residential lawns in central Pennsylvania. Journal of Soil and Water Conservation 54(3), 564‐568. Harden, C. P., 2006. Human impacts on headwater fluvial systems in the northern and central Andes. Geomorphology 79, 249‐263. Hewlett, J. D.,1961. Soil moisture as a source of base flow from steep mountain watersheds. U.S. Department of Agriculture ‐ Forest Service, Asheville, NC, Southern Research Station Paper No. 132. Ilstedt, U., Malmer, A., Verbeeten, E., Murdiyarso, D., 2007. The effect of afforestation on water infiltration in the tropics: A systematic review and meta‐analysis. Forest Ecology and Management 251, 45‐51. Jackson, R. B., Canadell, J., Ehlerlinger, J. R., Mooney, H. A., Sala, O. E., Schulze, E. D., 1996. A global analysis of root distributions for terrestrial biomes. Oeceologica 108 (3), 389‐411. Jiménez, C. C., Tejedor, M., Morillas, G., Neris, J., 2006. Infiltration rate in andisols: Effect of changes in vegetation cover (Tenerife, Spain). Journal of Soil and Water Conservation 61 (3), 153‐158. Jusoff, K., 1989. Physical soil properties associated with recreational use of a forested reserve area in Malaysia. Environmental Conservation 16 (44), 339‐342. Klute, A. 1986. Water retention: Laboratory methods. In: A. Klute (Ed.), Methods of Soil Analysis: Part 1 ‐ Physical and Mineralogical Methods. 2nd edition, Soil Science Society of America, Madison, WI, pp. 635‐662. Klute, A., Dirksen, C. 1986. Hydraulic conductivity and diffusivity: Laboratory methods. In: A. Klute (Ed.), Methods of Soil Analysis: Part 1 ‐ Physical and Mineralogical Methods. 2nd edition, Soil Science Society of America, Madison, WI, pp. 687‐734. Lal, R., 1996. Deforestation and land‐use effects on soil degradation and rehabilitation in western Nigeria. I: Soil physical and hydrological properties. Land Degradation and Development 7, 19‐45.
64
Leblanc, J. C., Gonçalves, E. R., Mohn, W. W., 2008. Global response to desiccation stress in the soil actinomycete Rhodococcus jostii RHA1. Applied Environmental Microbiology 74 (9), 2627‐2636. Lee, C. A., Lauenroth, W. K., 1994. Spatial distributions of grass and shrub root systems in the shortgrass steppe. The American Midland Naturalist 132 (1), 117‐123. Lee, K. E., Foster, R. C., 1991. Soil fauna and soil structure. Australian Journal of Soil Research 29 (6), 745‐775. Leigh, D. S., Webb, P. A., 2006. Holocene erosion, sedimentation, and stratigraphy at Raven Fork, Southern Blue Ridge Mountains. Geomorphology 78, 161‐177. Li, Y. Y., Shao, M. A., 2006. Change of soil physical properties under long‐term natural vegetation restoration in the Loess Plateau of China. Journal of Arid Environments 64, 77‐96. Martínez, L. J., Zinck, J. A., 2004. Temporal variation of soil compaction and deterioration of soil quality in pasture areas of Colombian Amazonia. Soil and Tillage Research 75 (1), 3‐18. Messing, I., Alrikkson, A., Johansson, W., 1997. Soil physical properties of afforested and arable land. Soil use and Management 13, 209‐217. NCDC, 2003. Climatography of the United States no. 84, 1971‐2000. Accessed September 12, 2007. <http://cdo.ncdc.noaa.gov/cgi‐bin/climatenormals/> NCDOT, 2007a. Jackson County ‐ Elevation Grid. Accessed September 15, 2007. <http://www.ncdot.org/it/gis> NCDOT, 2007b. Macon County ‐ Elevation Grid. Accessed September 15, 2007. <http://www.ncdot.org/it/gis> Oliviera, M. T., Merwin, I. A., 2001. Soil physical conditions in a New York orchard after eight years under different groundcover management systems. Plant and Soil 234 (2), 233‐237. Pizl, V., Josens, G., 1995. Earthworm communities along a gradient of urbanization. Environmental Pollution 90 (1), 7‐14. Pouyat, R. V., Yesilonis, I. D., Russell‐Anelli, J., Neerchal, N. K., 2007. Soil chemical and physical properties that differentiate urban land‐use and cover types. Soil Science Society of America Journal 71 (3), 1010‐1019. Rasiah, V., Kay, B. D., 1995. Runoff and soil loss as influenced by selected stability parameters and cropping and tillage practices. Geoderma 68, 321‐329. Rawls, W. J., Ahuja, L. R., Brakensiek, D., Shirmohammadi, A. 1993. Infiltration and soil water movement. In: D. Maidment (Ed.), Handbook of Hydrology.McGraw‐Hill, New York, pp. 5.1‐5.51.
65
Reiners, W. A., Bouwman, A. F., Parsons, W. F. J., Keller, M., 1994. Tropical rain forest conversion to pasture: changes in vegetation and soil properties. Ecological Applications 4 (2), 363‐377. Richter, D. D., Jr., Markewitz, D. 2001. Understanding Soil Change. Cambridge University Press, Cambridge. Robinson, G. R., Jr., Lesure, F. G., Marlowe, J. I., II, Foley, N. K., Clark, S. H.,1992. Bedrock geology and mineral resources of the Knoxville 1 degree by 2 degrees quadrangle, Tennessee, North Carolina, and South Carolina. U.S. Geological Survey, Bulletin No. 1979. Schwartz, R. C., Evett, S. R., Unger, P. W., 2003. Soil hydraulic properties of cropland compared with reestablished and native grassland. Geoderma 116, 47‐60. Soil Survey Staff ‐ NRCS, 2007. Official Soil Series Descriptions. Accessed September 15, 2007. <http://soils.usda.gov/technical/classification/osd/index.html> Solomon, D., Lehmann, J., Zech, W., 2000. Land use effects on soil organic matter properties of chromic luvisols in semi‐arid northern Tanzania: carbon, nitrogen, lignin, and carbohydrates. Agriculture, Ecosystems, and Environment 78, 203‐213. Southworth, S., Schultz, A., Denenney, D., Triplett, J.,2003. Surficial geologic map of the Great Smoky Mountains National Park. U.S. Geological Survey, Reston, Open File Report No. 03‐081. Tetzlaff, D., Soulsby, C., Waldron, S., Malcolm, I. A., Bacon, P. J., Dunn, S. M., Lilly, A., Youngson, A. F., 2007. Conceptualization of runoff processes using a geographical information system and tracers in a nested mesoscale catchment. Hydrological Processes 21, 1289‐1307. U.S. Census Bureau, 2007. 2006 Population Estimates. Accessed Ocober 28, 2007. <http://factfinder.census.gov> USDA‐NRCS, 2005. Soil Survey Geographic (SSURGO) Database for Macon County, North Carolina. Accessed September 10, 2007. <http://soildatamart.nrcs.usda.gov> USDA‐NRCS, 2007. Soil Survey Geographic (SSURGO) Database for Jackson County, North Carolina. Accessed September 10, 2007. <http://soildatamart.nrcs.usda.gov> USGS, 2000. North Carolina Land Cover Data Set: First Edition. Accessed September 20, 2007. <www.seamless.usgs.gov> USGS, 2003. North Carolina Land Cover Database Zone 57 Land Cover Layer. Accessed September 20, 2007. <www.seamless.usgs.gov> Velbel, M. A., 1985. Geochemical mass balances and weathering rates in forested watersheds of the southern Blue Ridge. American Journal of Science 285, 904‐930. Wear, D. N., Bolstad, P., 1998. Land‐use changes in the Southern Appalachian landscapes: spatial analysis and forecast evolution. Ecosystems 1 (6), 575‐594.
66
Webb, T. H., Claydon, J. J., Harris, S. R., 2000. Quantifying variability of soil physical properties within soil series to address modern land‐use issues on the Canterbury Plains, New Zealand. Australian Journal of Soil Research 38, 1115‐1129. Wigosta, M. S. (1991). Modeling and monitoring to predict spatial and temporal characteristics in small catchments. University of Washington, Seattle, WA, Ph.D. Wooten, R. M., Carter, M. W., Merschat, C. E.,2003. Geology of Gorges State Park. North Carolina Geological Survey, Information Circular No. 31. Yarnell, S. L.,1998. The southern Appalachians: a history of the landscape. U.S. Department of Agiculture Forest Service Southern Research Station, Asheville, NC, General Technical Report No. SRS‐18. Zhou, X., Lin, H. S., White, E. A., 2008. Surface soil hydraulic properties in four soil series under different land uses and their temporal changes. Catena 73 (2), 180‐188. Zimmermann, B., Elsenbeer, H., De Moraes, J. M., 2006. The influence of land‐use change on soil hydraulic properties: Implications for runoff generation. Forest Ecology and Management 222, 29‐38.
67
Table 3.1 Land use – Macon Co. and Jackson Co., NC (Source: USGS, 2003)
Land Use % Area
Open Water 0.6%
Developed 5.9% Open 5.3% Low‐intens i ty 0.5% Medium‐intens i ty 0.1% High‐intens i ty 0.0%
Forest 85.6% Deciduous 80.8% Evergreen 3.1% Mixed 1.7%
Shrub/Scrub 1.4%
Grassland/Pasture 6.3%
Cultivated Cropland .1%
Woody Wetlands .1%
68
Table 3.2 – Site characteristics: Land use, soil series, elevation, and aspect of soil sampling sites. Elevation, slope, and aspect values represent the range of the three sites at each location. Site numbers correspond to Figure 3.1.
# Name Land Use Series Elev. (m) # Name Land Use Series Elev. (m) Slope (%) Aspect (⁰)
1 Beasley Creek Forest Reddies 687‐689 16 Chapel Hill Forest Evard‐Cowee Complex 668 21‐26 170‐178
2 Caney Fork Forest Rosman 673 17 Gibson Cove Forest Evard‐Cowee Complex 673‐681 20‐30 205‐210
3 Rabbit Creek Forest Reddies 643‐644 18 Olive Hill Forest Evard‐Cowee Complex 710‐718 26‐33 100‐145
4 Tuckasegee River Forest Rosman 646‐647 19 Shope Fork Forest Fannin 701‐703 21‐26 204‐210
5 Weather Station Forest Reddies 687‐688 20 Wayah Road Forest Evard‐Cowee Complex 749‐787 33‐34 214‐245
6 Beasley Creek Lawn Reddies 691 21 Ledford Lane Lawn Evard‐Cowee Complex 668‐670 18‐32 162‐165
7 Fairgrounds Lawn Rosman 616 22 Olive Hill Lawn Evard‐Cowee Complex 701‐703 17‐21 111‐135
8 Mark Watson Park Lawn Cullowhee 614‐615 23 Speedwell Acres Lawn Evard‐Cowee Complex 697‐699 32‐35 240
9 Watauga Hazard Lawn Reddies 619‐620 24 Walnut Creek Lawn Evard‐Cowee Complex 661‐664 21‐28 140‐180
10 Weather Station Lawn Reddies 686‐687 25 WCU Lawn Cowee‐Evard Complex 674‐679 20‐24 178‐182
11 Killian Farm Pasture Rosman 658 26 Chapel Hill Pasture Evard‐Cowee Complex 661‐664 26‐29 144‐157
12 Rocky Branch Pasture Rosman 600 27 Gibson Cove Pasture Evard‐Cowee Complex 673‐675 17‐25 194‐209
13 Rabbit Creek Pasture Reddies 643 28 Speedwell Acres Pasture Evard‐Cowee Complex 691‐693 19‐22 235‐255
14 Tuckasegee River Pasture Rosman 648 29 Sunny Lane Pasture Evard‐Cowee Complex 668‐670 17‐32 183‐200
15 Watauga Hazard Pasture Reddies 619 30 Walnut Creek Pasture Evard‐Cowee Complex 673‐681 23‐33 152‐185aFor a l l al luvium s i tes : s lope = 0‐2%, aspect = n/a
ALLUVIUM SITESa SAPROLITE SITES
69
Table 3.3 – National Cooperative Soil Survey Official Series Descriptions (Source: Soil Survey Staff, 2007)
Series Taxonomic Class Typical Texture Official DescriptionAlluvium
Cullowhee Coarse‐loamy over sandy or sandy‐skeleta l , mixed, superactive, mesic Fluvaquentic Dystrudepts
Fine Sandy Loam Somewhat poorly drained, moderately rapidly permeable soi l s on floodpla ins in the Southern Appalachian Mounta ins . They formed in recent al luvium that i s loamy in the upper part and i s moderately deep to sandy strata that conta in more than 35 percent by volume rock fragments . They are very deep to bedrock. Slope ranges from 0 to 3%.
Reddies Coarse‐loamy over sandy or sandy‐skeleta l , mixed, superactive, mesic Oxyaquic Dystrudepts
Fine Sandy Loam Moderately wel l drained, moderately rapidly permeable soi ls on floodpla ins in the Blue Ridge. They formed in recent al luvium that i s loamy in the upper part and i s moderately deep to sandy strata conta ining more than 35 percent by volume gravel and/or cobbles . Slope ranges from 0 to 3%.
Rosman Coarse‐loamy, mixed, superactive, mesic Fluventic Humic Dystrudepts
Loam Very deep, wel l drained to moderately wel l drained, moderately rapidly permeable soi l s on floodpla ins in the Southern Appalachian Mountains . They formed in loamyal luvium. Slopes range from 0 to 3%.
Saprolite
Cowee Fine ‐loamy, parasesquic, mes ic Typic Hapludul ts
Gravel ly Sandy Loam Moderately deep, wel l drained, moderately permeable soi l s on ridges and s ide s lopes of the Blue Ridge . They formed in res iduum affected by soi l creep in the upper part, and weathered from fels ic to mafic, igneous and high‐grade
hi k Sl f 2 95%Evard Fine ‐loamy, parasesquic,
mes ic Typic Hapludul tsSandy Loam Very deep, wel l drained, moderately permeable soi l s on
ridges and s ides lopes of the Blue Ridge. They formed in res iduum affected by soi l creep in the upper part
Fannin Fine ‐loamy, paramicaceous , mes ic Typic Hapludul ts
Loam Very deep, wel l drained soi l s on gently s loping to very s teep ridges and s ide s lopes of the Blue Ridge. They formed in res iduum that i s affected by soi l creep in the upper part, and i s weathered from high‐grade metamorphic rocks that are high in mica content such as mica gneiss and mica schis t. Slopes are 6 to 95%.
70
Table 3.4 – Hydraulic properties of upper (0‐7.5 cm) and lower (7.5‐15 cm) soil cores
0‐7.5 cma 7.5‐15 cma
⎯x (σ) ⎯x (σ) t or Z* p trans .
Ksat‐L (mm h‐1)c 42 (3) 20 (3) 6.59 <0.001 log10
ρB (g cm‐3) 1.10 (.25) 1.30 (.23) ‐12.47 <0.001 ‐
φT (%)c 57.5 (1.17) 50.2 (1.18) 12.83 <0.001 log10
VMCfc (%)c 50.1 (1.15) 45.5 (1.19) 7.36 <0.001 log10
Sandd (%) 57.0 (11.8) 56.5 (13.2) 1.15* 0.251 asr
Siltd (%) 28.7 (8.2) 27.5 (8.4) 3.79* <0.001 asr
Clayd (%) 14.4 (5.2) 16.1 (6.4) ‐5.31* <0.001 asran=90; b178 degrees of freedom; symbols defined in section 3.3t = paired t‐test stati s ti c; *Z = Wilcoxon Signed Rank test s tatis ticc Geometric mean/standard deviation; das percent of <2mm masstrans . = transformation used to achieve normal dis tributionasr = arcs ine square root transformation
Diff. of Meansb
71
Table 3.5 – Soil physical characteristics by categories of parent material and land use
Alluviuma Saprolitea Forestc Lawnc Pasturec
t or T* p t or T* p t or T* p
Ksat (mm h‐1)f 0‐25 cm 15† (4)† 15† (4)† 0.07 0.941 log10 63† (2)† 7† (3)† 8† (3)† 55.26 <0.001 9.43 <0.001 8.74 <0.001 0.69 0.493 log10
ρB (g cm‐3) 0‐7.5 cm 1.09 (0.25) 1.11 (0.3) 1941.5* 0.395 ‐ 0.83 (0.17) 1.23 (0.14) 1.24 (0.17) 63.42 <0.001 9.60 <0.001 9.90 <0.001 0.29 0.771 ‐
7.5‐15 cm 1.27 (0.25) 1.33 (0.2) 1921.5* 0.311 ‐ 1.08 (0.21) 1.43 (0.12) 1.39 (0.17) 36.70** <0.001 536.0* <0.001 591.0* <0.001 958.5* 0.525 ‐
Average 1.18 (0.23) 1.22 (0.2) 1930.5* 0.347 ‐ 0.96 (0.16) 1.33 (0.12) 1.32 (0.16) 60.99 <0.001 9.71 <0.001 9.40 <0.001 0.30 0.765 ‐
φT (%) 0‐7.5 cm 59.0 (9.3) 57.8 (9.5) ‐0.69 0.492 1/x 68.7 (6.5) 53.6 (5.4) 53.1 (6.3) 50.85** <0.001 494.0* <0.001 504.0* <0.001 873.0* 0.539 ‐
7.5‐15 cm 51.9 (9.2) 49.9 (7.9) ‐0.95 0.344 1/x 59.1 (7.8) 46.2 (4.7) 47.5 (6.2) 31.86 <0.001 7.25 <0.001 6.52 <0.001 0.73 0.468 1/x
Average 55.5 (8.8) 53.9 (8.3) ‐0.87 0.388 1/x 63.8 (6.2) 49.9 (4.5) 50.3 (5.9) 48.78** <0.001 491.5* <0.001 523.0* <0.001 923.5* 0.906 ‐
VMCfc (%) 0‐7.5 cm 53.0 (8.4) 48.3 (5.7) ‐2.99 0.004 1/x 56.2 (7.7) 48.5 (5.1) 47.3 (6.2) 17.24 <0.001 5.87 <0.001 5.38 <0.001 0.48 0.629 ‐
7.5‐15 cm 47.2 (8.9) 45.3 (8.7) ‐0.95 0.345 1/x 53.3 (9.5) 42.3 (5.0) 43.2 (6.5) 21.20 <0.001 5.41 <0.001 4.68 <0.001 0.73 0.470 ‐
Average 50.1 (8.2) 46.8 (6.3) ‐2.00 0.050 1/x 54.8 (7.2) 45.4 (4.6) 45.2 (6.0) 24.64 <0.001 6.13 <0.001 6.03 <0.001 0.09 0.925 ‐
Sandg (%) 0‐7.5 cm 58 (14) 56 (9) 1932.0* 0.353 asr 61 (9) 58 (10) 52 (14) 2.16 0.122 1.03 0.306 2.01 0.051 1.08 0.285 asr
7.5‐15 cm 57 (16) 56 (10) 1930.5* 0.347 asr 60 (11) 58 (11) 52 (15) 1.33 0.271 0.43 0.669 1.55 0.128 1.16 0.250 asr
Average 57 (15) 56 (9) 2145.5* 0.429 asr 60 (10) 58 (11) 52 (15) 1.36 0.262 0.44 0.659 1.56 0.125 1.17 0.246 asr
Siltg (%) 0‐7.5 cm 29 (10) 29 (6) 1971.5* 0.542 asr 26 (7) 28 (7) 32 (10) 3.36** 0.186 863.0* 0.446 893.0* 0.080 856.5* 0.243 asr
7.5‐15 cm 28 (11) 27 (6) 2055.5* 0.952 asr 25 (7) 26 (7) 31 (10) 3.49** 0.174 939.0* 0.728 881.5* 0.117 886.5* 0.099 asr
Average 28 (10) 28 (6) 2087.5* 0.745 asr 26 (7) 27 (7) 31 (10) 5.84** 0.054 976.5* 0.353 926.5* 0.200 879.0* 0.118 asr
Clayg (%) 0‐7.5 cm 14 (6) 15 (5) 2206.0* 0.201 asr 13 (4) 14 (5) 16 (6) 1.63 0.202 ‐0.74 0.462 ‐1.79 0.079 ‐1.05 0.299 asr
7.5‐15 cm 15 (7) 17 (6) 1814.5* 0.060 asr 15 (6) 16 (6) 18 (7) 0.28 0.761 ‐0.42 0.674 ‐0.73 0.467 ‐0.33 0.741 asr
Average 15 (6) 16 (5) 1926.0* 0.315 asr 14 (5) 15 (6) 17 (7) 3.97** 0.137 930.0* 0.823 901.0* 0.052 873.0* 0.143 asr
an=45; b88 degrees of freedom; cn=30; d89 degrees of freedom; e58 degrees of freedom; fgeometric mean/standard deviation; gas percent of <2mm masstrans. = statistical transformation used to achieve normal distribution; asr = arcsine square root transformation; parameter symbols defined in text*Mann‐Whitney Rank Sum test T‐statistic; **ANOVA on Ranks test H‐Statistic
F or H** p
ANOVAd
trans .
Difference of Meansb Pairwise Differece of Meanse
trans .
LAND USEPARENT MATERIAL
⎯x (σ) ⎯x (σ) t or T* p ⎯x (σ)Forest vs . Lawn Fores t vs . Pasture Lawn vs . Pasture
⎯x (σ) ⎯x (σ)
72
Table 3.6 –Paired locations (adjacent locations with different land uses)
Locationa Parent Material
Forest Lawn Forest Lawn Forest Lawn Forest Lawn
Beasley Creek Alluvium 78 13 1.06 1.31 60.0 50.6 54.2 46.7
Weather Station Alluvium 96 32 0.79 1.11 70.3 58.1 60.2 52.2
Olive Hill Saprolite 66 8 0.98 1.32 63.2 50.2 51.7 44.6
Forest Pasture Forest Pasture Forest Pasture Forest Pasture
Rabbit Creek Alluvium 22 3 0.99 1.49 62.6 43.7 56.2 38.6
Tuckasegee River Alluvium 69 22 0.97 1.44 63.3 45.6 52.2 42.9
Chapel Hill Saprolite 68 4 1.01 1.44 61.8 45.8 53.2 42.5
Gibson Cove Saprolite 41 9 1.19 1.41 55.1 46.7 44.8 40.6
Lawn Pasture Lawn Pasture Lawn Pasture Lawn Pasture
Watauga Hazard Alluvium 8 4 1.38 1.37 48.1 48.2 40.0 41.8
Speedwell Acres Saprolite 14 10 1.31 1.21 50.6 54.5 46.6 47.4
Walnut Creek Saprolite 7 22 1.48 1.36 44.1 48.6 42.1 43.5aThe value for each location represents the mean of three sites; symbols defined in section 3.3Ksat values represent 0‐25 cm depth; Bulk density, porosity, and VMC values represent 0‐15 cm depth
VMCfc (%)ρB (g cm‐3)Ksat (mm h‐1) φT (%)
73
Figure 3.1. Study area and soil sampling locations: Macon Co. and Jackson Co., NC. The white band across the counties represents a geologic zone of interlayered biotite gneiss and amphibolite (Robinson et al., 1992) – site selection was limited to this zone. Vertical exaggeration is 3X. Each mapped sampling location represents three sampling points, and location numbers correspond to the site characteristics in Table 3.2.
74
Figure 3.2. Soil physical characteristics by parent material and land use. The boxes represent the inter‐quartile range, with the black line across each box indicating the median value. The white dotted line represents the arithmetic mean, and the whiskers represent the 5th and 95th percentiles.
75
Figure 3.3. Particle size distributions of forest, lawn, and pasture soils. Textural classes corresponding to particle size distributions observed in these soils are bounded by solid black lines (e.g., Loam, Clay Loam).
Clay Loam
Silt Loam
Loam
Sandy Loam
Loamy Sand
Sand (%)
76
Figure 3.4. Comparison of saturated hydraulic conductivity (Ksat) measurements by field and laboratory methods. Saturated hydraulic conductivity (Ksat) was measured by two methods: 1) in situ at each field sampling point using a compact constant head permeameter at a depth of 0‐25 cm, and 2) by standard laboratory methods on 10 cm‐diameter intact soil cores. The laboratory Ksat values presented here are the average of the two cores from each sampling point (0‐7.5 cm and 7.5‐15 cm). The R2 and p‐value reflect results from polynomial regression analysis performed on the log10‐transformed variables (y = 1.296‐0.1215x+0.2212x2).
77
Figure 3.5. Saturated hydraulic conductivity among sites. The field saturated hydraulic conductivities (Ksat) at the three sampling points at each location are shown, demonstrating the similar variability within a given locations to that seen among all locations within each land use and the similarity in Ksat between the parent materials (saprolite and alluvium).
78
Figure 3.6. Comparison of soil saturated hydraulic conductivities with precipitation intensities occurring in western North Carolina. The solid lines represent the recurrence intervals (RI) of storm events of given precipitation intensity and duration in Franklin, NC (re‐created from Bonnin et al., 2004). The dashed lines represent the mean saturated hydraulic conductivity (Ksat) of soils underlying each land use (n = 30 sites per land use, mean = geometric). As Ksat represents the lower bound of the soil infiltration rate, the figure demonstrates the far greater likelihood of Hortonian overland flow in lawn and pasture soils, especially associated with sustained storm events, during which overland flow is of greatest concern to watershed hydrologic processes.
79
CHAPTER 4
EFFECTS OF LAND USE AND GEOMORPHOLOGY ON STREAM BASEFLOWS IN THE SOUTHERN BLUE RIDGE MOUNTAINS OF GEORGIA AND NORTH CAROLINA, USA1
__________ 1Price, K., Jackson, C. R., Parker, A. J., Reitan, T. and Dowd, J. To be submitted to Hydrological Processes.
80
ABSTRACT
While it has been shown in many settings that both human land use and natural topographic variability
influence stream baseflows, their interactions and relative influences have remained unresolved. Our
objective was to determine the influence of human land use and watershed geomorphic characteristics
in explaining baseflow variability in the southern Blue Ridge Mountains of North Carolina and Georgia.
Continuous discharge data for 35 streams (watershed area 3 to 146 km2) was obtained for a period of
1.5 years, including two late‐summer low flow seasons. Various baseflow metrics were calculated (1‐
day, 7‐day, and 14‐day minimum daily flows, 1 percentile flow, and baseflow index) for three time
subsets of the gage records (low flow season 2007, low flow season 2008, and water year 2008). A
comprehensive suite of watershed characteristics, including factors of watershed topography, channel
network morphometry, soils, land use, and precipitation were used in multiple regression analysis of
baseflow variability among the 35 watersheds. Overall, geomorphic factors of drainage density and
slope variability showed the strongest relationships to baseflow in this region, with watershed forest
cover and colluvium also demonstrating significant influence. Forest cover showed a consistent positive
relationship with baseflow, despite the higher evapotranspiration rates generally associated with forest
compared with other landcovers, which highlights the importance of infiltration and recharge under
undisturbed landcover in sustaining baseflows. The baseflow metrics used in this study demonstrated
inconsistent results in model development, with the 1 percentile flows and BFI indicating a stronger land
use response than the minimum flow metrics, which were most strongly explained by topographic
characteristics.
81
4.1. Introduction and Background
Baseflow refers to the portion of streamflow that enters stream channels via subsurface
pathways and sustains streamflow between precipitation events, supplied by storage reservoirs such as
bedrock, saprolite, colluvium, alluvium, and soil. Natural catchment physical characteristics, such as
geology, geomorphology, and soils, exert important influence on baseflow and serve as the template
upon which other important factors, such as land use and climate, are superimposed [Johnson, 1998].
As many regions are currently experiencing rapid land use change, concurrent with increased demands
on public water supply, a better understanding of watershed function and baseflow is critical to issues of
contaminant dilution, aquatic habitat, and public water use [Barnes and Kalita, 1991; Hornbeck et al.,
1993; Smakhtin, 2001; Konrad and Booth, 2005]. Anthropogenic changes to the landscape alter
baseflow timing and quantity. Aside from direct manipulations, such as impoundments and water
withdrawals from streams and subsurface storage, human activity influences baseflow by indirect
mechanisms associated with changes in land use and land cover. Conversion of native vegetation to
other vegetative covers or artificial surfaces can drastically alter evapotranspiration (ET) [Liu et al.,
2008]. Land use change also may alter surface permeability characteristics, through soil compaction
associated with human land use and addition of impervious surface to watersheds [Rose and Peters,
2001; Price et al., in review]. Of all the important factors relating to recharge and baseflow, ET and
surface permeability are the most susceptible to alteration by human impact. Within a given region,
geology and precipitation may be relatively consistent, but geomorphic characteristics may vary
considerably among watersheds, especially in rugged terrain. It is thus important to evaluate how
anthropogenic and geomorphic characteristics interact, in order to fully understand how land use
change affects baseflows in regions of variable topography. Thus, the goal of research presented here
is to quantitatively resolve the effects of land use change and varied topography on baseflow in small to
mesoscale watersheds of the southern Blue Ridge Mountains.
82
There is a vast body of literature demonstrating reduced streamflow associated with greater
watershed forest cover [e.g., Bosch and Hewlett, 1982; Sahin and Hall, 1996; Jones and Post, 2004], with
many studies specifically demonstrating lower baseflow with higher watershed forest cover [Harr et al.,
1982; Keppeler and Ziemer, 1990; Hicks et al., 1991; Smith, 1991; Black et al., 1995]. This negative
relationship between watershed forest cover and baseflow is attributed to greater interception and
water use by mature trees compared with other land cover types [Bosch and Hewlett, 1982; Calder,
1990; McCullough and Robinson, 1993; Johnson, 1998]. These results have occasionally been
interpreted as a suggestion that watershed management approaches could include deforestation to
increase water yield for public use [Brooks et al., 1991; Chang, 2003]. However, the bulk of this
literature derives from experimental forestry, where soil disturbance is far less pronounced than occurs
when forest is converted to more permanent land uses, such as pasture, residential development, etc.
There is a sound theoretical basis and growing empirical evidence that long‐term forest conversion
reduces baseflows, because the intensive soil compaction and impervious surface increases that
accompany human land uses decrease infiltration rates and subsurface storage recharge [Bruijnzeel,
2004; Price et al., in review]. These changes in surface infiltration may outweigh the reductions in ET
ascribed to removal of mature forest stands. Furthermore, the majority of these studies have focused
on very small systems (generally smaller than 1 km2), and there is evidence that upscaling results from
such studies to larger, more heterogeneous basins is problematic [Blöschl, 2001; Sivapalan, 2003;
Soulsby et al., 2004]. Understanding baseflow response to long‐term land use change at scales relevant
to watershed management and policy is critical, and it is increasingly clear that extrapolation of results
from short‐term vegetation disturbance at the forestry‐plot scale is not appropriate for understanding
the response of larger systems to long‐term forest conversion to other land use types.
A major complication of analyzing and predicting baseflow response to land use change is the
breadth of other factors that influence infiltration rates and transmission rates from subsurface storage
83
to the channel network. Within a context of consistent bedrock geology, topography may exert
substantial influence on baseflow processes, particularly in areas of pronounced relief [Tetzlaff et al.,
2009]. Spatial variability in ET and precipitation may result from differences in topographic
characteristics such as aspect and elevation among watersheds [Kovner, 1956]. Furthermore,
topographic slope and channel network development influence transmission rates of water [Tetzlaff et
al., 2009]. Climatic and topographic variability additionally influence the storage reservoir itself, through
their effects on bedrock weathering and soil development. The effects of land use change on baseflow
timing and quantity may be mitigated or amplified by basin topography, and there may be situations in
which topographic conditions exert such strong control on baseflow that drastic changes in land use are
required to induce detectible changes in baseflow [Konrad and Booth, 2002]. Furthermore, spatial
variability in precipitation may mask baseflow vegetation caused by vegetation or soil disturbance.
Little is known regarding which specific topographic variables are most useful for explaining
baseflow variability among watersheds. Several studies have identified various watershed topographic
and channel network morphometric variables as key influences on baseflow. These studies have
demonstrated that factors such as relief, slope, drainage density, and watershed shape, which all
influence the ability of water to flow to the channel network and out of the watershed, significantly
relate to stream baseflow [Farvolden, 1963; Thomas and Benson, 1970; Vogel and Kroll, 1992; Woods et
al., 1997; Marani et al., 2001; Warner et al., 2003; Cherkauer and Ansari, 2005]. While not directly
addressing variability of baseflow quantity, many additional studies have shown that the residence time
of water in subsurface storage is determined by flowpath length and water transit times, which are
primarily a function of catchment topography [Rodhe et al., 1996; Asano et al., 2002; McGuire et al.,
2005; Tetzlaff et al., 2009]. However, there remains little understanding about which metrics expressing
basin topography and morphometry are most useful for explaining baseflow variability, and it remains
unclear how variables of topography and land use interact to influence baseflow. Thus, the specific
84
objective of this study was to monitor the streamflow of mesoscale watersheds in the southern Blue
Ridge Mountains for a period of 1.5 years, including 2 annual low flow periods, and relate the variability
of baseflow discharge among these systems to land use characteristics and a thorough suite of
topographic metrics.
4.2. Study Area
This study was conducted within the watersheds of the upper sections of the Tuckasegee,
Nantahala, and Little Tennessee Rivers, which together comprise the majority of the Little Tennessee
River system upstream of Lake Fontana, a Tennessee Valley Authority (TVA) reservoir (Figure 4.1). This
study area contains the entirety of Macon County, North Carolina (NC), and portions of Clay and Jackson
Counties, NC and Rabun County, Georgia (GA). These watersheds are located within the Tallulah Falls
thrust sheet of the East Flank Blue Ridge lithotectonic belt [Robinson et al., 1992], a sub‐unit of the Blue
Ridge physiographic province. Regional bedrock consists of intrusive igneous rocks and varied
metasedimentary assemblages [Wooten et al., 2008]. The regional geology is characterized by
crystalline bedrock with minimal fracture flow [Velbel, 1985; Santhi et al., 2008], and the hydraulic
conductivities of all bedrock types are of similar magnitudes [Daniel and Payne, 1990; Mesko et al.,
1999]. A saprolite mantle 1‐30 m thick drapes the ridges and slopes throughout the study area [Hewlett,
1961], and substantial deposits of colluvium are present on benches, coves, and footslopes
[Southworth, 2003; Leigh and Webb, 2006]. The saprolite‐bedrock contact, which largely parallels the
surface topography, is believed to be the predominant subsurface topographic control on hillslope
hydrology [Hatcher, 1988]. The average depth to solid bedrock in the Coweeta Creek basin in the
southern part of the study area is six meters [Swank and Douglass, 1975].
Precipitation in this region typically is quite high but spatially variable, with the general trend of
highest precipitation toward the southern escarpment of the Blue Ridge and lower precipitation toward
85
the northern part of the study area. The 30‐year average precipitation at the Coweeta Experiment
Station’s low elevation rain gage (686 masl) is 1826 mm, whereas 38 km to the northeast in Cullowhee,
NC (elevation = 668 masl), the 30‐year average is only 1313 mm (Figure 4.1). The 30‐year average
annual daily mean temperatures at these stations are quite similar: 19.8 ⁰C at Coweeta and 19.6 ⁰C at
Cullowhee. The study period coincided with a severe drought affecting the southeastern United States.
The average deficits from 30‐year precipitation normals at the Coweeta and Cullowhee stations were
18% for 2007 and 19% for 2008 [NCDC, 2004].
In the absence of human impact, regional land cover would be nearly 100% forest [Yarnell, 1998;
Delcourt and Delcourt, 2004]. Present‐day land cover is predominantly forest, with nonforest land cover
occurring primarily as pasture and low‐density development. The region experienced intensive,
widespread timber harvest and agriculture during the late nineteenth and early twentieth century,
followed by forest regrowth on mountain slopes [Davis, 2000]. In some areas, agricultural land
abandonment and vegetation regrowth have been common since the 1960s, but exurban population
growth and associated expansion of residential and low‐ to medium‐density urban land cover has
affected substantial portions of the region [Wear and Bolstad, 1998; Cho et al., 2004; Gragson and
Bolstad, 2006]. The largest town in the study area is Franklin, with a 2006 population of 3618 [U.S.
Census Bureau, 2007].
4.3. Methods
4.3.1 Site Selection and instrumentation: Over one hundred watersheds were delineated upstream of
public access points and characterized in terms of watershed forest cover and maximum elevation. For
this original characterization, watersheds were delineated manually using USGS digital raster graphics
(DRGs) in ArcView 3.3, and percent forest cover and maximum elevation were calculated using 2001
National Land Cover Database (NLCD) data [USGS, 2003] and 10 m DEM data. A subset of 36 sites was
86
selected for instrumentation with stage recorders. All sites representing land use and topographic end
members within the study area were included. For selection of intermediate sites, K‐means cluster
analysis was used to identify groups of similar watersheds, and study sites were randomly selected from
within each cluster, with number of sites selected proportional to cluster size. No nested watersheds
were included in these 36 sites, which range from 2.68 to 34.10 km2 in watershed area and 44.4 to 99.9
percent watershed forest cover (Table 4.1). Four pairs of sites were identified, in which topographic
characteristics were very similar but watershed forest cover was different. These sites were used for
pairwise comparison of baseflow under different land use conditions with topographic variability
removed.
The majority of these watersheds (35 of 36 sites) were instrumented with Odyssey capacitance
water level recorders. This was achieved by suspending the recorder in 38 mm diameter PVC tubing in
the stream bank, connected to the stream thalweg with a lateral 25 mm diameter PVC pipe, or by
suspending the probe directly into the stream channel by attaching the PVC housing to a wooden bridge
support. Because of shallow bedrock and a lack of a bridge attachment, one site (Fulcher Branch) was
not suitable for Odyssey probe use. At this site, a HOBO pressure transducer was situated under a
bedrock ledge within the stream bed, with an additional transducer installed on the bank for barometric
pressure adjustment. All instruments recorded water level every 10 minutes, with the period of data
recording spanning August 5, 2007 to January 20, 2009. Additionally, three USGS gaged watersheds
(Nantahala River at Rainbow Springs, Cartoogechaye Creek, and Cullasaja River near Highlands) and one
watershed gaged as part of the USFS Coweeta Hydrologic Laboratory network (Shope Fork) were
included. Six of the smaller watersheds are nested within the USGS watersheds. The Coweeta
laboratory maintains a trapezoidal weir on Shope Fork, with a pressure transducer used for continuous
inflection‐point stage monitoring for Shope Fork. This site was additionally instrumented with an
Odyssey capacitance water level recorder for accuracy assessment of the discharge calculation method
87
by comparison with weir data. The USGS gages record instantaneous stage data every 15 minutes.
Efforts were made to limit site selection to watersheds within a large bedrock unit of biotite gneiss and
amphibolite spanning a southwest‐northeast transect across the study area. However, to increase the
number of watersheds included in the study, additional sites with significant portions of different gneiss‐
granite and some fraction of schist‐sandstone assemblages were included from the western portion of
the study area.
4.3.2 Flow measurement and rating curve development ‐ For each of the watersheds instrumented with
Odyssey and HOBO recorders, stage‐discharge rating curves were developed. Several methods were
tested, including ordinary least squares regression and power law curve fitting, but Bayesian power law
curve fitting produced the best results and was applied to all sites (Figure 4.2) [Moyeed and Clarke,
2005; Reitan and Petersen‐Øverlier, 2008; Reitan and Petersen‐Øverlier, 2009]. Discharge was measured
for rating curves using the midsection method [Mosley and McKerchar, 1993], with an acoustic doppler
velocity meter used for 0.6‐depth velocity measurement at no fewer than 13 points per cross‐section.
Discharge was measured at least 10 times for each stream during the study period. For measurement
of high flows on the largest streams, it was necessary to use dye for velocity measurements within five
sections of the channel, with channel area determined by probe stage and laser‐level channel survey.
Additionally, bankfull discharge was modeled using Manning’s equation. Channel dimension parameters
for Manning’s equation were calculated from cross‐sectional and slope data from laser‐level channel
survey, and Manning’s n was calculated from the highest measured discharge value at each site. The
corresponding bankfull stream stage was paired with the Manning’s equation modeled bankfull
discharge and included in rating curve development. Several sites had problems with the stage‐
discharge fit, due to intermittent beaver dams, flood scour, etc. Ultimately, sites included in the analysis
were limited to 35 watersheds, including the USGS and USFS sites.
88
4.3.3 Streamflow analysis ‐ Streamflow records were divided into three time periods for statistical
analysis of topographic and land use influences on baseflow: 1) Low flow season 2007 ‐ August 5 to
November 12, 2007 (LF07), 2) low flow season 2008 ‐ August 1 to November 12, 2008 (LF08), and Water
Year 2008 ‐ October 1, 2007 to September 30, 2008 (WY08). The designation of ‘low flow season’ does
not imply 100% baseflow during the time period, but instead designates the season generally containing
the lowest flows of the year [Smakhtin, 2001]. Low flow season was defined based on regional long
term trends, data availability, and 2007‐2008 conditions (Figure 4.3). Because installation of the stage
recording capacitance probes was not complete until early August, 2007, the LF2007 was assigned a
slightly later start date of August 5. Sites with missing data totaling 5% of a given low flow season or
10% of the water year were not included in the analyses corresponding to that time period. Because of
statistical circularity issues that would arise from filling missing data values with estimations from
neighboring sites, missing data values totaling less than the 5 and 10% thresholds were left as null
values. Ten‐minute interval stage data were available for all Odyssey probe and HOBO transducer sites.
The inflection‐point record of the USFS site at Shope Fork was converted to 10‐minute data by linear
interpolation. Five baseflow metrics were calculated for each stream: 1) the flow exceeded 99% of the
time (Q99), 2) minimum daily mean flow (Qmin1), 3) minimum 7‐day mean flow, (Qmin7) 4) minimum 14‐
day mean flow (Qmin14), and Baseflow Index (BFI). Q99 was calculated from the instantaneous stage
records. Qmin1, Qmin7, and Qmin14 were calculated from daily mean flows. Qmin1 represents the minimum
daily flow value within the specific time period, and Qmin7 and Qmin14 were determined from the 7‐ and
14‐day moving averages of daily mean flow. Daily mean flows were also used for calculation of the
baseflow index (BFI), which is the ratio of baseflow to total streamflow. Baseflow was separated using
the Eckhardt recursive digital filter method [Lim et al., 2005; Eckhardt, 2008]. Additionally, the mean
daily flow from each of the time periods was calculated, to provide context for the baseflow metrics
89
(Qmean). For comparison of flow magnitudes across stream systems of different scales, all metrics except
BFI were standardized by dividing by watershed area. Hereafter, Q99, Qmin1, Qmin7, Qmin14, and Qmean
represent the area‐adjusted flows (m3s‐1km‐2). As a proportion of baseflow to total flow, BFI was not
area‐standardized.
4.3.4 Watershed precipitation summary data ‐ Daily precipitation data for the study period were
obtained for 35 stations throughout the region, from the Coweeta Hydrologic Laboratory, the National
Climatic Data Center (NCDC), the State Climate Office of North Carolina, and the National Weather
Service Integrated Flood Warning System (IFLOWS). Missing data values were filled using double‐mass
curve analysis combining precipitation totals from three neighboring stations [Brutsaert, 2005]. For
spatial interpolations of precipitation data, ordinary kriging, inverse‐distance weighting, and Thiessen
polygons (or Voronoi diagrams) were compared. Interpolations were produced for a short time period
for which nine additional gages were available (total n = 44), and the results of the interpolations with
this more dense station network were compared with results of the same time period using the primary
set of stations (n= 35). Ordinary kriging produced the closest match to the results from the larger set of
stations, and produced the lowest overall RMSE values, confirming the findings of other interpolation
comparisons [Nour et al., 2006]. The mean precipitation depth of all pixels within a watershed was used
to represent the total precipitation during the period of analysis.
4.3.5 Selection of topographic and landscape variables for analysis ‐ For final site characterization,
watersheds were delineated using the Basin 1 extension in Arc View 3.3 [Petras, 2003]. To define
watershed land use, a classification of 2006 Landsat imagery following the NLCD classification scheme
was obtained from the Coweeta Long Term Ecological Research (LTER) program. The NLCD categories
were reduced to 5 more general land use categories: 1) forest and shrub, 2) developed, 3) pasture and
90
agriculture, 4) barren, and 5) open water. Topographic and morphometric variables were selected from
classic and recent literature, based on those which have a strong theoretical relationship to streamflow,
a legacy of inclusion in watershed characterization, and/or have previously been demonstrated to have
statistically significant relationships to streamflow metrics (Table 4.2). Topographic and morphometric
characteristics were calculated from LiDAR data for all sites in NC, for which LiDAR was available (33 of
35 sites) [NCDOT, 2007a,b,c]. For the remaining sites, a 10 m DEM was used. The Basin 1 extension for
ArcView 3.2 was used to delineate the regional stream network, using a threshold of 18,050 m2, which
was shown to best match the ground‐truthed stream network available for a small portion of the study
area [N.C. Department of Environment and Natural Resources, 2009]. Soil parameters were calculated
from STATSGO soils data [USDA‐NRCS, 2005; USDA‐NRCS, 2006, USDA‐NRCS, 2007, USDA‐NRCS, 2008].
Bedrock geology was classified by hydraulic conductivity, following the Blue Ridge regional scheme
presented by Mesko et al. [1999], using digital 1:250,000 maps available for the North Carolina
watersheds [Robinson et al., 1992], and a digital 1:500,000 map available for the state of Georgia
[Georgia Geologic Survey, 1999].
Watershed characterization included calculation of 66 metrics of basin and channel network
geomorphology, soils and bedrock, land use, and precipitation (Table 4.2). Simple correlation analysis
was used to identify strongly correlated variables (operationally defined as |R|> 0.8) among the
watershed characteristics and precipitation totals, of which only one was retained. Preference was
given to variables with previously demonstrated or strong theoretical linkages to streamflow. Principal
Components Analysis (PCA) was used to examine the data structure of the 43 remaining independent
variables. The PCA rotated factor loadings from were used to further reduce the data to 14 candidate
variables for inclusion in multiple regression modeling (Table 4.3).
91
4.3.6 Multiple regression modeling ‐ Separate analyses were performed for each of the five dependent
variables (Q99, Qmin‐1, Qmin‐7, Qmin‐14, and BFI). Each of the three time periods (LF07, LF08, and WY08)
were analyzed separately. From the 14 independent variables, best subsets regression was used to
identify the 3‐5 most promising models, as judged by values of adjusted R2 and Mallow’s Cp.
Determination of the final model was based on overall goodness of fit, significance and direction of
influence of independent variables, parsimony, and logic. Watershed area was only included in the
modeling for BFI, because the flow magnitude metrics were normalized by watershed area (having
demonstrated bivariate correlations 0.87 and greater with watershed area). Standard t‐tests were used
to compare the means baseflows of relatively higher‐ and lower‐forest watersheds, with categories of
higher‐ and lower‐forest determined by K‐means cluster analysis. Significance was defined as p ≤ 0.05
for all analyses, except for within the t‐tests comparing higher‐ and lower‐forest watersheds, in which a
Bonferroni correction established significance at p < 0.0167. Simple correlation analyses were
performed between watershed forest cover and all baseflow metrics.
4.4. Results
4.4.1 Rating Curves – Bayesian power law curve fitting produced a range of goodness of fits for the
rating curves, due to varying amounts of scatter in the stage‐discharge relationships among the sites
(Figure 4.4). Curve fit was evaluated using R2 values along with visual evaluation of credibility. Sites
with intolerable point scatter and poorest fits were not included in further analyses. (Rating curves for
all sites are presented in Appendix A, and stream hydrographs are presented in Appendix B.) Results
from discharge calculations using the velocity‐area method and Bayesian rating curve development at
Shope Fork were compared with measured discharge at the USFS weir, located approximately 50 m
downstream from the capacitance probe gaging site. These results indicated a 7.2% average difference
between 10‐minute interval data from the two methods, and an average 6.5% difference between the
92
daily mean flows. Differences were minor during low flows and most pronounced during high flows, in
which the weir discharge was generally higher than the discharge calculated using the capacitance probe
and natural cross section.
4.4.2 Precipitation interpolation – The 35 regional precipitation stations encompass an area of 6545 km2,
and station elevation ranged from 580 m to 1663 m (Table 4.4). Because of a prevalent south to north
track of large tropical storms from the Atlantic Ocean and Gulf of Mexico, there is much greater rainfall
in the southern part of the study area. Significant spatial autocorrelation of precipitation was present
for all time periods and confirmed kriging as an appropriate approach to spatial interpolation of
precipitation. The insignificant correlation between precipitation and elevation precluded the use of
cokriging to account for elevation in the interpolation. While it is generally assumed that precipitation
increases with elevation [Drogue et al., 2002], the spatial variability of regional rainfall overwhelms local
relationships between elevation and precipitation at this spatial scale and station density. The
correlations between total precipitation and elevation were significant but not strong (LF07: R = 0.381, p
= .024; LF08: R = .421, p = .012; R = 0.357, p =.035), and similarly low correlations were evident for five
individual storms that included nine additional precipitation stations. Ordinary kriging of precipitation
totals over the three time periods demonstrated significant precipitation variability across the study
watersheds (Figure 4.5).
4.4.3 Independent variable reduction – Watershed forest cover emerged as the key land use metric for
inclusion in multiple regression. Two metrics of watershed elevation were included: median elevation
and “hypsometric index 1”, a metric of watershed elevation distribution (Table 4.2). Elongation and
south‐facing slopes (as fraction of the watershed hillslope pixels with aspect between 135 and 215⁰)
were the sole variables included from each of the categories of basin morphometry and aspect. Within
93
the study area, there is a very strong correlation between most watershed hillslope metrics and land use
(as low‐slope land is easier to convert to pasture or developed use than high‐slope land). As a result,
only the fraction of watershed area with slope lower than 2% rise and slope standard deviation (as a
metric of complexity of watershed slope and topography) were included from the suite of slope metrics
originally characterized. Several channel network morphometric variables were included in the multiple
regression modeling: drainage density, percent of stream length that is first order stream, and
bifurcation ratio, all of which express various aspects of channel distribution and potentially relate to
the ability of the watershed to remove water in subsurface storage. The areal percentage of two soil
parent materials (alluvium and colluvium), and clay‐dominant soil texture were selected. Lastly, mean
watershed total precipitation from the three separate time periods was also included as an independent
variable for statistical modeling. More detailed information for these variables is presented in Table 4.2.
4.4.4 Multiple regression modeling of baseflow metrics‐ Area‐adjusted flow magnitudes varied
considerably across sites, spanning an order of magnitude for all flow metrics (Figure 4.6). Models were
created independently for each of the three time periods within each of the five baseflow metrics,
yielding a total of 15 models. While parameters of land use, geomorphology, and/or precipitation were
able to be combined to create statistically significant models for all baseflow metrics, all models left
significant variability unexplained. The weakest model, developed for Qmin1 LF08, only produced an R2 of
0.192 (p=0.007), while the strongest model, developed forQmin1 WY08, produced an R2 of 0.515
(p=0.001).
Regression modeling demonstrated a predominant influence of geomorphic parameters on
stream baseflows in this study area, as opposed to watershed forest cover (Table4.5). Drainage density
and watershed slope standard deviation (an indicator of topographic complexity) were selected in the
greatest number of models. The models indicate that greater drainage density, indicating greater
94
fluvial dissection and connectivity of subsurface storage to the channel network, is associated with
reduced baseflows. Greater slope variability was consistently associated with higher baseflow. Forest
cover, colluvium, and precipitation were also selected in a substantial number of the models and were
positively related to baseflow. Elongation, alluvium, and hypsometric index 1 were not included in any
of the best models.
The three time periods (LF07, LF08, and WY08) showed varied degrees of success with multiple
regression modeling. Among the time periods for a given baseflow metric, the LF07 model was
generally the strongest, often accounting for twice the variance as the models for LF08 or WY08. Qmin1
was the only baseflow metric for which LF07 did not show the strongest results, but for this variable the
fit for both LF07 and WY08 were also relatively strong. Variable selection was inconsistent among the
baseflow metrics. Forest was consistently important in the regression models for Q99 and BFI, but not
for the minimum daily flow metrics. For all of the minimum daily flow metrics, the geomorphic
parameters of drainage density, slope standard deviation, colluvium, and first order streams were
consistently important. BFI demonstrated consistent dependence on watershed area, forest cover,
median elevation, and clay soils.
4.4.5 Difference of means tests and watershed pair comparisons of forest cover influence‐ The results of
the K‐means cluster analysis established two distinct groups of watersheds based on forest cover
(F=60.20, p=0.000). The lower‐forest group included 15 sites and ranged from 44.4 to 86.0% forest
cover, with a center of 75%, and the higher‐forest group included 20 sites and ranged from 90.1 to
99.9% forest cover, with a center of 94%. Standard t‐tests were performed to compare the mean
baseflow of lower‐ and higher‐forest watersheds, with separate analyses performed for each baseflow
metric (Figure 4.7). Because of the replication of tests for each baseflow metric over the three time
periods of analysis, a Bonferroni correction was required. With the a priori significance established at p
95
≤ 0.05, correcting for performing three replicate tests produced a threshold significance level of p =
0.017 for these t‐tests. Results showed that the mean baseflow of higher forest watersheds was greater
than that of the lower forest watersheds, across all baseflow metrics and all time periods. These
differences were statistically significant below the p = 0.017 threshold for the Q99 flows, over all 3 time
periods. The Qmin1, Qmin7, and Qmin14 flows for LF07 were also significant. No time period produced
statistically significant differences in BFI for lower‐ and higher forest streams.
Comparison of the four pairs of topographically similar watersheds also indicated consistently
higher baseflow in the higher‐forested pair member, among all of the flow magnitude metrics (Table
4.6). However, the pairs and time periods demonstrate inconsistent direction of difference for BFI.
Values of Q99, Qmin1, Qmin7, and Qmin14 range from 7 to 132% higher in the higher forest watershed.
Among the four pairs, the difference in forest cover ranged from 6 to 24%, and these results indicate
that greater reductions in forest cover are associated with greater differences in baseflow. Simple
correlation analysis indicated consistently greater baseflow with higher forest cover and reduced
baseflow with greater nonforest land use, especially pasture (Table 4.7).
4.5. Discussion
The expected results for this study were that a mixture of land use and geomorphic variables
would emerge as significant explanatory variables for stream baseflows in the southern Blue Ridge
Mountains. While this generally proved to be the case, the influence of land use was less influential in
the regression modeling than expected, and the amount of baseflow variability left unexplained by
geomorphology, land use, and precipitation (widely understood as the key controls on streamflow) was
high for many of the metrics. This study area is characterized by relatively low human impact, so that
the range of watershed land use among sites included in this study is relatively narrow. Overall, forest
cover ranges from 44.4 ‐99.9% among the sites, but only one watershed has less than 70% forest cover.
96
Despite this relatively small range of conditions and relatively low level of watershed disturbance, forest
cover demonstrated a consistent positive relationship with all baseflow metrics, and showed statistically
significant influence on Q99 and BFI. While the influence of land use is clearly present on stream
basflows, the results of this study indicate the geomorphic influences in this highly variable topographic
setting outweigh the influence of land use. The discussion that follows will emphasize the differences in
models among the time periods, the explanatory models for the five baseflow metrics, the potential
reasons for the unexplained variability, and the insights these results shed on watershed function.
4.5.1 Differences in explanatory models among the time periods of Low Flow Season 2007 (LF07), Low
Flow Season 2008 (LF08), and Water Year 2008 (WY08) – An interesting pattern demonstrated by these
results is that the LF07 period consistently produced stronger multiple regression models , which
incorporated more independent variables, than seen with the later time periods. One possible
explanation for this is the pronounced drought that affected the region during 2007‐2008, along with a
very large tropical storm that coincided with both the LF08 and WY08 periods. It is highly likely that the
LF07 period is more representative of average conditions in this region than the later time periods,
during which the systems were under greater stress from the sustained drought conditions. Under such
conditions, it is probable that the predominant control on baseflow is the availability of long‐term
subsurface storage, which was not directly measured by any of the independent variables included in
this study.
The LF08 period included low flows in mid‐August that were the lowest on all available long‐
term USGS records in the study area, going back as far as the 1940s (Figure 4.3). The Qmin1, Qmin7, and
Qmin14 values for both LF08 and WY08 include values from this period, and the Q99 for both LF08 and
WY08 were both undoubtedly strongly affected by these extremely low flows. Baseflows during these
anomalously intense conditions likely reflect the availability of long residence time storage among the
97
watersheds. Introducing further complexity to these time periods, the remnants of a large Gulf of
Mexico storm system (Tropical Storm Fay), passed directly over the study area on 28‐29 August 2008,
with storm totals in excess of 250 mm in parts of the study area. This storm generated overbank floods
throughout the region. Watershed responses to the intense drought, punctuated by an intense storm
event, were highly variable. For some watersheds, the late‐August 2008 period immediately prior to Fay
were no lower than LF07 minimum flows, whereas others demonstrate reductions from LF07 levels
greater than an order of magnitude. While direct water withdrawals during this period of water scarcity
may have introduced variability, it is also clear that some watersheds have apparently greater drought
resilience, which is likely due to greater long term storage capacity in the watersheds. The watersheds
also accumulated highly variable amounts of recharge from Fay and a smaller storm that occurred
immediately afterward (a small remnant of Hurricane Gustav). Many sites demonstrated much higher
baseflow following the storm series. In contrast, several sites immediately receded to levels even lower
than flows prior to Fay (Figure 4.8). Spatial variability of rainfall during the storm and widely varying
infiltration and retention characteristics among the watersheds explain the wide range of shallow
subsurface recharge as a result of these storms. This pronounced variability in system response to both
the intense drought and the very large storm event introduced baseflow variability that is likely far
greater than in a typical year, and may partially explain the relative weak statistical results for LF08 and
for the minimum flow metrics of WY08.
Similarly, the predominance of the major storm event during the LF08 period likely explains why
precipitation was a significant variable for Q99 in LF08 and WY08, but not for LF07. While the overall
mean precipitation depths among the watersheds do not differ greatly between the two time periods
(206 mm in LF07 vs. 217 mm in LF08), the coefficients of variation (CV) of mean watershed precipitation
depths indicates nearly double variability during LF08 (CV = 0.131 in LF07 vs. 0.209 in LF08). The
minimum flow metrics (Qmin1, Qmin7, and Qmin14) strongly emphasize the period of regionally very low
98
flows prior to the impact of Tropical Storm Fay remnants, and are thus not significantly influenced by the
precipitation totals.
4.5.2 Differences in independent variable selection among the baseflow metrics – Five separate baseflow
metrics were included in these analyses, because there is no consensus within the hydrologic sciences
on any single metric as the best measure of baseflow in the absence of long‐term records. This partially
arises from the dual need to analyze baseflow in terms of a low flow season, and also in terms of
season‐independent recession flows [Smakhtin, 2001]. For the latter, the literature favors the baseflow
index (BFI) [Beran and Gustard, 1977; Nathan et al., 1996; Neff et al., 2005], but there is no such
consensus for characterizing minimum flows. The results of this study do not indicate that any one of
the baseflow metrics (Q99, Qmin1, Qmin7, or Qmin14) is inherently superior for relating watershed
characteristics to baseflow quantity. Instead, these results demonstrate that these metrics are actually
responding to distinct sets of factors and suggest the metrics might represent different aspects of the
hydrologic system.
Overall, Q99 responded most strongly to forest cover, precipitation, and areal percentage of flat
terrain (slope < 2%). Forest cover demonstrated a positive relationship with baseflow, underscoring the
importance of high‐infiltration‐capacity forest soils in recharging shallow subsurface storage [Price et al.,
in review]. Area of slopes < 2% demonstrated a negative relationship with Q99 for LF08, presenting a
relationship contrary to the expectation that flatter terrain would allow for greater infiltration and
recharge [Thomas and Cervione, 1970; Neff et al., 2005; Santhi et al., 2008]. In this study area of
significant relief and historically low population pressures, development has historically concentrated in
flatter areas. Using Cartoogechaye Creek as an example, the area of slope less than 2% is comprised of
11% forest, 19% developed land, and 70% pasture/agriculture. While the correlation between low slope
area and forest cover was below the threshold value of 0.8 used for variable reduction, the relationship
99
is nevertheless quite strong (R = ‐0.787). As various studies have shown that nonforest land use is
associated with reduced infiltration and recharge [e.g., Godsey and Elsenbeer, 2002; Jiménez, 2006; Price
et al., in review], it is highly probable that the negative influence of low slope area on baseflow is in
reality a representation of a negative relationship between nonforest land use and baseflow (which
complies with the positive relationship shown between forest cover and Q99 during the other two time
periods). As discussed above, the low precipitation variability during LF07 likely explains why
precipitation significantly influences Q99 for LF08 and WY08, but not LF07.
All of the minimum daily flow metrics (Qmin1, Qmin7, and Qmin14) relate most strongly to drainage
density, slope standard deviation, and, to a lesser extent, colluvium. Drainage density, or the length of
stream channel per unit watershed area, emerged as the single most selected variable overall. It is quite
logical and theoretically viable that greater fluvial dissection and, thus, greater connectivity between
subsurface storage and the channel network, would have a negative relationship with minimum flows
[Smakhtin, 2001]. The results of several other studies corroborate the negative relationship between
drainage density and baseflow [Gregory and Walling, 1968; Warner et al., 2003]. Greater contact area
between stored water and stream channels facilitates removal of water, thus leaving less water in
subsurface storage when systems are stressed during warmer and drier times of year. Additionally, the
negative relationship between drainage density and baseflow is at least partially due to negative
correlations between subsurface characteristics and drainage density. For example, drainage density is
theoretically greater in watersheds with shallower confining layers, in which channel development
occurs more readily due to a lack of subsurface storage capacity. Thus, it is possible that drainage
density not only relates negatively to baseflow because it facilitates removal of water in subsurface
storage, but also because it is a direct reflection of the subsurface storage conditions.
Topographic complexity (as slope standard deviation) shows a consistently positive relationship
to the minimum flows metrics. Topographic complexity implies a variety of potential storage units, and
100
can be understood through the extreme cases of topographic uniformity. A uniformly steep watershed
would favor rapid transfer of water out of the watershed, while a uniformly flat watershed may not
contain sufficient subsurface storage volume to sustain high baseflows. Furthermore, low slope areas
favor deeper infiltration and the possibility that down‐valley movement of water in alluvial aquifer flows
through the aquifer, below and alongside the stream network itself. High toographic variability reflects
the intermediate condition, in which watersheds contain a range of slopes. This variability results from a
mixture of hillslope and fluvial deposits, such as colluvium and alluvium, favoring greater storage of
subsurface watersheds. This is corroborated by the fact that amount of watershed colluvium also
emerged as an important explanatory variable. The positive relationship between baseflow and
colluvium suggests that colluvium is an important subsurface reservoir in this system dominated by
shallow storage. This corroborates recent findings in Scotland that groundwater storage at lower slopes
in mountainous headwaters (where colluvium accumulates) to be a major source of baseflow [Tetzlaff
and Soulsby, 2008], and to a recent study indicating substantial colluvial water storage in the Cascades
[Schultz et al., 2008].
BFI was consistently influenced by forest, watershed area, clay, and median elevation across the
three time periods, and like most of the other metrics, BFI for LF07 showed a stronger relationship to
watershed characteristics than LF08 or WY08. As the proportion of baseflow to total flow, BFI reflects
sustained streamflow behavior, as opposed to the system response during periods of low precipitation.
As was expected, higher forest cover is associated with higher BFI, indicating a less flashy regime due to
lower overland flow in forested watersheds. The corollary to lower overland flow is increased
subsurface recharge and higher sustained baseflows. While BFI is generally treated as area‐
independent, these results show that there is a scale‐dependence associated with BFI. In high relief
terrain, this relationship arises due to the increased volume of valley bottom storage capacity as systems
increase in size, with alluvial bottomlands often supplying significant water to stream systems [Larkin
101
and Sharp, 1992]. In effect, areal expansion of flat terrain generates an exponential increase in water
storage volume. The size range of systems included in this study, from small headwater watersheds less
than 3 km2 to larger, more complex systems up to 150 km2 likely underscored this effect. This also
explains the negative relationship between BFI and median elevation, as the watersheds representing
the upper part of the system also tend to have higher median elevation. This interpretation that the BFI
of upland, headwater systems is lower than that of watersheds including more alluvial bottomland also
explains the positive relationship between clay and BFI observed in this study. While several previous
studies have shown relationships between soil texture and BFI, in these cases finer soil texture was
associated with lower BFI, due to impeded infiltration [Gustard et al., 1986; Boorman et al., 1995]. In
the study area of the southern Blue Ridge, soil survey data and field observation indicates a greater
presence of clay in alluvial terrace soils than in the saprolitic and colluvial soils of the surrounding
hillslopes [USDA‐NRCS, 2005; USDA‐NRCS, 2006, USDA‐NRCS, 2007, USDA‐NRCS, 2008]. The relationship
between BFI and clay content is again more likely a correlation expressing presence of alluvial
bottomland, rather than a causal relationship between soil texture and BFI.
The lack of consistent independent variable selection among the baseflow metrics implies that
the metrics express separate aspects of watershed function. The distinction of BFI from Q99 and the
minimum flow metrics of Qmin1, Qmin7, and Qmin14 is straightforward. BFI measures baseflow proportion
throughout the entire period of record, rather than isolating extremely low flows, and as such the best
models for BFI emphasize both watershed storage volume (indirectly through watershed area) and
landcover, with a positive relationship with forest cover consistently shown. However, it is less
straightforward why Q99 and the minimum flow metrics selected separate suites of independent
variables. Q99 emphasized forest cover and precipitation, while the minimum flow metrics emphasized
drainage density, slope standard deviation, and amount of colluvium in the watershed. Perhaps Q99,
with no fixed temporal constraints within the time period, is a better reflection of sustained watershed
102
function than the minimum flow metrics, which isolate extreme conditions. Results from the multiple
regression modeling support this interpretation. The Q99 models showed a positive relationship to both
forest cover and precipitation, indicating this variable is a reflection of the amount of water allowed to
enter the subsurface system. The minimum flow metrics, on the other hand, selected variables relating
more directly to subsurface storage capacity and retention, which logically relate to the minimum
amount of water in the stream system during times of low precipitation.
4.5.3 Interpretation of the overall influence of land use, geomorphology, and precipitation on baseflow‐
General understanding of watershed function holds that factors of land use, climate, and
geomorphology/geology are the important controls on baseflow discharge. This study attempted to
quantify the relative importance of the influences of land use and geomorphology (within similar
bedrock) by comparing watersheds of varied land use and topographic characteristics and accounting
for spatial variability of precipitation. However, the best model that resulted from this accounted for
only 51.5% of the variability among the sites, and most models were significantly weaker. These results
are similar to other studies attempting to statistically model baseflow [Thomas and Benson,1970;
Gustard et al., 1989; Kent, 1999; Neff et al., 2005]. While there are studies presenting baseflow
modeling results with very high R2 values using geomorphic, land use, and climate parameters [Vogel
and Kroll, 1992; Nathan et al., 1996; Zhu and Day, 2009], these studies have modeled unstandardized
baseflows, using watershed area as an independent variable. However, area is responsible for such a
large amount of baseflow variability, area‐standardized baseflow was used in order to allow for further
insights into watershed function. For example, baseflow models with area as the sole independent
variable produced R2 values ranging from 0.948 to 0.981 for Pennsylvania streams, leaving very little
variability to explore the influences of other watershed characteristics [Zhu and Day, 2009]. Among the
watersheds in this study, multivariate models including area as an independent variable produced R2
103
values ranging from 0.835 to 0.934 across all metrics of baseflow magnitudes (unstandardized by area,
and not including BFI). Area alone accounted for 74.1 to 86.4% of the variability observed. All of the
independent variables selected by the best models of the area‐adjusted flows retained their significance
with area included, and together accounted for 2‐8% of additional variability among the models.
Overall, the models for LF07 (for both area‐ standardized and unstandardized flows) were most
successful, likely because the anomalously low flows after multiple years of drought introduced
additional variability for LF08 and minimum WY08 flows among the stressed systems.
Precipitation played a lesser role in explaining baseflows than was expected. Total precipitation
emerged as an important explanatory variable in only two of the models (Q99 LF08 and Q99 WY08). As
explained above, this is likely due to the anomalously low precipitation levels during the study period,
punctuated by a major tropical storm event with highly spatially variable precipitation. It was
additionally surprising that no significant relationship emerged between elevation and precipitation in
this study area, despite the general understanding that orographic effects lead to higher precipitation
with increasing elevation [Drogue et al., 2002]. While analysis of major storm precipitation totals in the
Blue Ridge Mountains has shown significant relationships between elevation and precipitation totals
[Sturdevant‐Rees et al., 2001; Nykanen and Harris, 2006; Wooten et al., 2008], low and/or insignificant
correlations between precipitation and elevation have also been demonstrated in mountainous
environments [Chang, 1977; Penk and Schaake, 1990; Konrad, 1995]. The relationship between
elevation and precipitation is clouded in this study area due to the importance of Gulf of Mexico and
western Atlantic storm systems moving into the region from the south. While the highest elevations are
in the northern part of the study area, along the Plott‐Balsam Range and toward the Great Smoky
Mountains National Park, the highest rainfall totals occur in association with the Blue Ridge Escarpment
in the southern part of the study area [Konrad, 1996]. It is likely that orographic effects are evident at
the scale of individual mountains during individual storm events, but that this relationship does not
104
upscale to demonstrate a regional trend, because of the greater importance of the predominant south‐
north track of major storm events. This interpretation is supported by the findings of an analysis of
topographic setting and precipitation patterns among 44 climate stations in the southern Blue Ridge
[Konrad, 1996]. The results of this study showed that elevation was found to correlate significantly with
light precipitation events, but not heavy events. Heavy events were best explained by 1) southern
exposure, and 2) distance from the Gulf of Mexico.
The variability left unexplained by the multiple regression models (Table 4.5) can be accounted
for by three main factors: 1) data uncertainty, 2) direct water withdrawals and surface inputs of
imported water, and 3) variable oversimplification. First, there may be error resulting from the stage‐
discharge method in natural channels, and comparison of this method against discharge calculated from
a trapezoidal weir demonstrated a 6.5% difference for daily flows. All GIS‐derived watershed
characteristics contain some level of uncertainty, with the spatial interpolation of point precipitation
totals being the most notable. Second, the lack of precipitation during the low flow seasons of these
drought years undoubtedly led to some amount of direct water withdrawals from streams for irrigation
or other uses. The extent to which this occurred is not known. The municipal water supplies in this
study area are sourced from large streams outside of all the watersheds included in this study, with the
exception of Cartoogechaye Creek. Non‐municipal water sources are generally wells drilled deep into
bedrock fractures likely disconnected from the surface water system, given what is known about the
regional hydrogeologic characteristics [Velbel 1985, Seaton and Burbey, 2005]. The addition of water
from these sources to the surface water system via septic drainage, outdoor water use, and
infrastructure leakage could be substantial enough to introduce baseflow variability [Lerner, 2002].
Additionally, distribution and size of water retention ponds was not accounted for.
Lastly, the most important reason that these watershed characteristics fail to account for
greater baseflow variability is that the independent variables themselves are only correlates to the
105
actual hydrologic parameters of a basic water budget. Precipitation can be directly estimated, but the
other key hydrologic variables of ET and subsurface storage are only crudely approximated by
watershed characteristics. Factors of land use, aspect, and elevation relate to ET but do not directly
quantify it. Factors of geomorphology relate to subsurface storage capacity, and factors of soil texture
and land use relate to infiltration and recharge – but subsurface storage volumes and aquifer properties
are not directly represented.
Given all these limitations to attempting to represent the watershed system, the ability to
explain 20‐50% of baseflow variability with a small set of surficial watershed characteristics sheds
significant light on watershed function in this study area. While land use was not selected in the best
regression model for every baseflow metric, it was clearly influential to both Q99 and BFI. Its absence in
the minimum flow models can be explained by the fact that these variables seem to be most directly
influenced by subsurface water availability to provide baseflow during the driest times of year. The
importance of land use was confirmed by the pair comparisons, difference of means tests between
watersheds of contrasting forest cover, and simple correlation analyses between various land use
metrics and baseflow variables. The comparisons of baseflow of streams within topographically similar
watersheds that differ in forest cover indicated substantially higher baseflow in the more forested pair
member for all metrics of baseflow magnitude, though not for BFI (Table 4.6). In many cases, the more
highly forested watershed demonstrated baseflow levels more than 50% greater than the lower‐forest
pair member, and there is a clear trend associating greater reductions in watershed forest cover with
greater reductions in stream baseflow. These results were corroborated by the difference of means
tests (Figure 4.7), which demonstrated consistently higher baseflow among the more forested
watersheds, with statistically significant differences in Q99 observed for all time periods. Additionally, all
metrics except BFI showed statistically significant differences between lower‐ and higher‐forest
watersheds for theLF07 time period. Simple correlation analysis showed consistent direction of
106
difference between baseflow metrics and land use classes (Table 4.7), with higher forest cover
associated with higher baseflow in all cases. Again, statistical significance was only demonstrated for
Q99 (all time periods), and for LF07 minimum flow metrics.
4.6. Conclusions
The streamflow from 35 watersheds ranging in size from 3 to 146 km2 in the southern Blue
Ridge of North Carolina was monitored for 1.5 years, encompassing two low flow seasons. The
watershed and channel network morphometry, soil characteristics, land use, and precipitation were
characterized for the 35 watersheds and related to unit‐area baseflow metrics and BFI. The results of
this study indicate that baseflow in the southern Blue Ridge is affected most strongly by factors of
geomorphology, particularly drainage density, topographic variability, and colluvium. While apparently
less influential than watershed geomorphology, watershed forest cover demonstrated a consistent,
positive relationship with baseflow. Five baseflow metrics were considered in this study: the flow
exceeded 99% of the time (Q99), the 1‐day, 7‐day, and 14‐day minimum flows (Qmin1, Qmin7, and Qmin14),
and baseflow index (BFI). All metrics except BFI were standardized by watershed area, which
demonstrated very strong correlations to metrics of baseflow magnitude. Multiple regression modeling
of various landscape factors of topography, land use, and precipitation was used to explain unit‐area
baseflow variability during three time periods: 1) low flow season 2007 (LF07, August 5‐November 12),
2) low flow season 2008 (LF08, August 1‐November 12), and 3) water year 2008 (WY08, October 1, 2007
to September 30, 2008). Q99 showed a strong positive relationship to watershed forest cover and
precipitation, while Qmin1, Qmin7, and Qmin14 were strongly related to drainage density, slope standard
deviation, and colluvium. BFI was most strongly related to forest cover, watershed area, median
elevation, and clay. The lack of a consistent set of explanatory variables suggests these metrics are
sensitive to different aspects of watershed function. Q99 appears to relate to the amount of water that
107
enters subsurface storage, while the minimum flow metrics appear to respond to variables linked to the
storage capacity of the watershed and the connectivity of these reservoirs to the surface water network.
BFI relates both to land use and to the watershed’s position in the overall system, with smaller, higher
elevation headwater watersheds demonstrating lower storage capacity and lower BFI. Regression
models were stronger for the LF07 than the other two time periods. This is attributed to a pronounced
drought that caused severe low flows that were the lowest on regional record in August 2008.
Moreover, Tropical Storm Fay occurred early in the LF08 period and apparently induced considerable
regional variability in hydrologic recharge. Flows during LF07 are more representative of typical
conditions and may better reflect watershed function.
The models for all metrics and time periods left considerable variability unexplained. While the
model strength is not unusually low for this type of study, the results raise questions about the sources
of the remaining variability. It is our contention that the model weakness is caused not by the failure to
consider important aspects of the watershed system, but instead because the variables included only
relate to the gains and losses of water from the system without directly measuring them. The region is
characterized by fairly low variability in land cover, which is the likely reason that forest cover failed to
demonstrate a more consistent role in explaining baseflow variability. The results from t‐tests
comparing means of lower‐ and higher‐forest cover watersheds, paired comparisons of topographically
similar watersheds with varied forest cover, and simple correlations between land use and baseflow
metrics all confirm that higher forest cover is associated with higher baseflow among these watersheds.
The results of this study suggest that as development continues in this region, further land use change
will be associated with reductions in baseflow. This could have negative implications for water
availability for public use, but is also of great concern due to the high number of endangered aquatic
species endemic to the southern Blue Ridge [Mayden, 1987; Warren et al., 2000; Sutherland et al.,
2002].
108
4.7. References
Apaydin, H., F. Ozturk, H. Merdun, and N. M. Aziz (2006), Determination of the drainage basin characteristics using vector GIS, Nordic Hydrol. 37(2), 129‐142. Asano, Y., T. Uchida, and N. Ohte (2002), Residence times and flow paths of water in steep unchannelled catchments, Tankami, Japan, J. Hydrol., 261, 173‐192. Barnes, P. L. and P. K. Kalita (2001), Watershed monitoring to address contamination source issues and remediation of the contamination impairments, Water Sci. and Tech. 44(7), 51‐56. Batelaan, O. and F. De Smedt (2007), GIS‐based recharge estimation by coupling surface‐subsurface water balances, J. Hydrol. 337(3‐4), 337‐355. Beran, M.A. and A. Gustard (1977) A study into the low‐flow characteristics of British rivers, J. Hydrol. 35, 147‐157. Black, A. R., R. C. Johnson, and M. Robinson, (1995), Effects of forestry on low flows in Scotland and Northern Ireland , Report to the Scotland and Ireland Forum for Environmental Research, 52 pp., Institute of Hydrology, Wallingford, UK. Blöschl, G. (2001), Scaling in hydrology, Hydrological Processes 15, 709‐711. Boorman, D. B., J. M. Hollis, and A. Lilly (1995), Hydrology of soil types: a hydrologically based classification of the soils of the United Kingdom, Institute of Hydrology Report, 126. Bosch, J. M. and Hewlett, J. D. (1982), A review of catchment experiments to determine the effect of vegetation changes on water yield and evapotranspiration, J. Hydrol. 55, 3‐23. Brenning, A. and D. Trombotto (2006), Logistic regression modeling of rock glacier and glacier distribution: Topographic and climatic controls in the semi‐arid Andes, Geomorphology 81(1‐2), 141‐154. Brooks, K.N., P. F. Ffolliot, H. M. Gregersen, and J. L. Thames (1991), Hydrology and the Management of Watersheds, 402 pp., Iowa State University Press, Ames, IA. Bruijnzeel, L. A. (2004), Hydrological functions of tropical forests: not seeing the soil for the trees? Agric. Ecosys. Env. 104(1), 185‐228. Brutsaert, W. (2005), Hydrology: an Introduction, 605 pp., University Press, Cambridge. Calder, L. R. (1990), Evaporation in the uplands, 166 pp., Wiley, Chichester, UK. Chang, M. (1977), An evaluation of precipitation gage density in a mountainous terrain, J. Amer. Water Resour. Assoc. 13(1), 39‐46.
109
Chang, M. (2003), Forest Hydrology: an Introduction to Water and Forests, 373 pp., CRC Press, Boca Raton, FL. Cherkauer, D. S. and S. A. Ansari (2005), Estimating ground water recharge from topography, hydrogeology, and land cover, Ground Water 43(1), 102‐112. Cho, S. H., D. H. Newman, and D. N. Wear (2003), Impacts of second home development on housing prices in the southern Appalachian highlands, Rev. Urb. Reg. Dev. Stud.15(3), 208‐225. Daniel, C. C. III and R. A. Payne (1990), Hydrogeologic unit map of the Piedmont and Blue Ridge provinces of North Carolina, U.S. Geol. Surv. Water Res. Inv. Rep., 90‐4035. Davis, D. E. (2000), Where There are Mountains: an Environmental History of Appalachia, 320 pp., University of Georgia Press, Athens. Delcourt, P. A. and H. R. Delcourt (2004), Prehistoric Native Americans and Ecological Change: Human Ecosystems in Eastern North America since the Pleistocene, 205 pp., University Press, Cambridge. Drogue, G., J. Humbert, J. Deraisme, N. Mahr, and N. Frelson (2002), Statistical‐topographic model for using an omnidirectional parameterization of the relief for mapping orographic rainfall, Int. J. Climatol. 22, 599‐613. Eckhardt, K. (2008), A comparison of baseflow indices, which were calculated with seven different baseflow separation methods, J. Hydrol., 352(1‐2), 168‐173. Farvolden, R. N. (1963), Geologic controls on ground‐water storage and base flow. J. Hydrol. 1, 219‐249. Fitzpatrick, F. A., I. R. Waite, P. J. D’Arconte, M. R. Meador, M. A. Maupin, and M. E. Gurtz (1998), Revised methods for characterizing stream habitat in the National Water‐Quality Assessment Program, U.S. Geol. Surv. Water Res. Inv. Rep., 98‐4052. Georgia Geologic Survey (1999), Digital Geology Map of Georgia, 2nd edition, Georgia Geol. Surv. Doc. Report, 99‐20. Godsey, S., and H. Elsenbeer (2002), The soil hydrologic response to forest regrowth: a case study from southwestern Amazonia, Hydrological Processes 16, 1519‐1522. Gragson, T. L. and P. Bolstad (2006), Land use legacies and the future of southern Appalachia, Soc. and Nat. Res. 19(2), 175‐190. Gregory, K. J. and D. E. Walling (1968), The variation of drainage density within a catchment, Int. Assoc. of Scientif. Hydrologists Bull. 13(2), 61‐68. Gustard, A., D. C. W. Marshall, and F. Sutcliffe (1986), Low flow study of Northern Ireland, 10 pp, Institute of Hydrology, Wallingford, UK. Gustard, A., L. A. Roald, S. Denuth, H. S. Lumadjeng, and R. Gross (1989), Flow Regimes from Experimental and Network Data (FREND). Inst. of Hydrol., Hydrological Studies Volume 1. 344 pp.
110
Harr, R. D., A. Levno, and R. Mersereau (1982), Streamflow changes after logging 130‐year‐old douglas fir in two small watersheds, Water Resour. Res. 18(3), 644‐647. Hatcher, R. D. (1988), Bedrock geology and regional geologic setting of Coweeta Hydrologic Laboratory in the eastern Blue Ridge, in Forest Hydrology and Ecology at Coweeta, edited by W. T. Swank and D. A. Crossley, Jr., pp. 81‐92, Springer‐Verlag, New York. Hewlett, J. D. (1961), Soil moisture as a source of base flow from steep mountain watersheds, U.S. Dept. of Agric. – Forest Service, Southern Research Station Paper, 132. Hicks, B. J. , R. L. Beschta, and R. D. Harr (1991), Long‐term changes in streamflow following logging in western Oregon and associated fisheries implications, Water Res. Bull 27(2), 217‐226. Hornbeck, J. W., M. B. Adams, E. S. Corbett, E. S. Verry, and J. A. Lynch (1993), Long‐term impacts of forest treatment on water yield: a summary for northeastern USA, J. Hydrol. 150, 323‐344. Jiménez, C. C., M. Tejedor, G. Morillas, and J. Neris (2006), Infiltration rate in andisols: Effect of changes in vegetation cover (Tenerife, Spain), Journal of Soil and Water Conservation 61(3), 153‐158. Johnson, R., (1998), The forest cycle and low river flows: a review of UK and international studies, Forest Ecology and Management 109, 1‐7. Jones, J. A. and D. A. Post (2004), Seasonal and successional streamflow response to forest cutting and regrowth in the northwestern and eastern United States, Water Resour. Res .40, 1‐19. Kent, C.A. (1999) The influence of changes in land cover and agricultural land management practice on baseflow in southwest Wisconsin, 1968‐1998. Ph.D. Dissertation, University of Wisconsin, Madison, WI.
Keppeler, E. T. and R. R. Ziemer (1990), Logging effects and streamflow: water yield and summer low flows at Caspar Creek in northwestern California, Water Resour. Res. 26(7), 1669‐1679. Konrad, C. E. II (1995), Maximum precipitation rates in the Blue Ridge Mountains of the Southeastern United States, Climate Res.5(2), 159‐166.
Konrad, C. E. II (1996), Relationships between precipitation event types and topography in the southern Blue Ridge Mountains of the southeastern USA, Int. J. of Climatology 16(1), 49‐62.
Konrad, C. P. and D. B. Booth (2002), Hydrologic trends associated with urban development for selected streams in the Puget Sound Basin, Western Washington. U.S. Geological Survey, Water‐Resources Investigation Report 02‐4040.
Konrad, C. P. and D. B. Booth (2005), Hydrologic changes in urban streams and their ecological significance, Am. Fisher. Soc. Symp. 47, 157‐177. Larkin, R. G. and J. M. Sharp, (1992), On the relationship between river‐basin geomorphology, aquifer hydraulics, and ground‐water flow direction in alluvial aquifers, Geol. Soc. Am. Bull. 104, 1608‐1620.
111
Leigh, D.S. and P. A. Webb (2006), Holocene erosion, sedimentation, and stratigraphy at Raven Fork, Southern Blue Ridge Mountains, Geomorphology 78, 161‐177. Lerner, D. N. (2002), Identifying and quantifying urban recharge: a review, Hydrogeol. J. 10, 143‐152. Lim, K. J., B. A. Engel, T. Zhenxu, J. Choi, K.‐S. Kim, S. Muthukrishnan, and D. Tripathy (2005), Automated web GIS based hydrograph analysis tool, WHAT, J. Am. Water Res. Assoc., 41(6), 1407‐1416. Liu, M. L., H. Q. Tian, G. S. Chen, W. Ren, C. Zhang, and J. Y. Liu (2008), Effects of land‐use and land‐cover change on evapotranspiration and water yield in China during 1900‐2000, J. Am. Water. Resour. Assoc. 44(5), 1193‐1207. Marani, M., E. Eltahir, and A. Rinaldo (2001), Geomorphic controls on regional base flow, Water Resour. Res.37(10), 2619‐2630. Mayden, R. L. (1987), Historical ecology and North American highland fishes: a research program in community ecology, in Community and Evolutionary Ecology of North American Stream Fishes, edited by W. J. Matthews and D. C. Heins, pp**, University of Oklahoma Press, Norman, OK. McCullough, J. S. G. and M. Robinson (1993), History of forest hydrology, J. Hydrol. 150, 189‐216. McGuire, K. J., J. J. McDonnell, M. Weiler, C. Kendall, B. L. McGlynn, J. M. Welker, and J. Siebert (2005), The role of topography on catchment‐scale water residence time, Water Resour. Res. 41(W05002), 1‐14. Melton, M. A. (1957), An analysis of the relations among elements of climate, surface properties, and geomorphology, Colombia University Department of Geology Technical Report, 11. Mesko, T.O., L. A. Swain, and E. F. Hollyday (1999), Hydrogeology and hydrogeologic terranes of the Blue Ridge and Piedmont physiographic provinces in the eastern United States, U.S. Geol. Surv. Hydrologic Investigations Atlas, HA‐732‐B. Mosley, M. P. and A. I. McKerchar (1993), Streamflow, in Handbook of Hydrology, edited by D. R. Maidment, pp.** to **, McGraw Hill, **city**. Moyeed, R. A. and R. T. Clarke (2005), The use of Bayesian methods for fitting rating curves, with case studies, Adv.Water Res., 28(8), 807‐818. NCDC – National Climatic Data Center (2004), Monthly Station Climate Summaries, 1971‐2000, Climatography of the U.S., 20. N.C. Department of Environment and Natural Resources (2009), Lower Little Tennessee /Upper Little Tennessee / Tugaloo Hydrography, Accessed March 9, 2009, <http://www.ncstreams.org/DataAccess/tabid/257/Default.aspx> NCDOT, 2007a. Clay County ‐ Elevation Grid. Accessed September 15, 2007. <http://www.ncdot.org/it/gis>
112
NCDOT, 2007b. Jackson County ‐ Elevation Grid. Accessed September 15, 2007. <http://www.ncdot.org/it/gis> NCDOT, 2007c. Macon County ‐ Elevation Grid. Accessed September 15, 2007. <http://www.ncdot.org/it/gis> Nathan, R. J., K. Austin, D. Crawford, and N. Jayasuriya (1996), The estimation of monthly yield in ungauged catchments using a lumped conceptual model, Australian J. Water Resour., 1(2), 65‐75. Neff, B. P., S. M. Day, A. R. Piggott, and L. M. Fuller (2005), Base Flow in the Great Lakes Basin, U.S. Geological Survey Scientific Investigations Report, 2005‐2517. Nour, M. H., D. W. Smit, and M. Gamal El‐Din (2006), Geostatistical mapping of precipitation: implications for rain gauge network design, Water Sci. and Tech. 53(10), 101‐110. Nykanen, D. K. and D. Harris (2003), Orographic influences on the multiscale statistical properties of precipitation, J. Geophys. Res. 108, 8381 Peck, E. L. and Schaake, J. C. (1990), Network design for water‐supply forecasting in the West Petras, I. (2003), Arcview Avenue Script: Basin1, Accessed December 12, 2008, <http://arcscripts.esri.com/details.asp?dbid=10668> Price, K., C. R. Jackson, and A. J. Parker, Variation of surficial soil hydraulic properties across land uses in the southern Blue Ridge Mountains, USA, submitted to J. Hydrol. Reitan, T. and A. Petersen‐Øverlier (2008), Bayesian power‐law regression with a location parameter, with applications for construction of discharge rating curves, Stoch. Environ. Res. Risk Assess., 22(3), 351‐365. Reitan, T. and A. Petersen‐Øverlier (2009), Bayesian methods for estimating multi‐segment discharge rating curves, Stoch. Environ. Res. Risk Assess., 23(5), 627‐642. Robinson, G. R., Jr., F. G. Lesure, J. I. Marlowe II, N. K. Foley, and S. H. Clark (1992), Bedrock geology and mineral resources of the Knoxville 1⁰ x 2⁰ quadrangle, Tennessee, North Carolina, and South Carolina, U.S. Geol. Surv. Bulletin, 1979. Rodhe, A., L. Nyberg, and K. Bishop (1996), Transit times for water in a small till catchment from a step shift in the oxygen 18 content of the water input, Water Resour. Res., 32(12), 3497‐3511. Rose, S. and N. E. Peters (2001), Effects of urbanization on streamflow in the Atlanta area (Georgia, USA): a comparative hydrological approach, Hydrol. Processes 15, 1441‐1457. Sahin, V. and M. J. Hall (1996), The effects of afforestation and deforestation on water yields, J. Hydrol. 178, 293‐309. Santhi, C., P. M., R. S. Muttiah, J. G. Arnold, and P. Tuppad (2008), Regional estimation of base flow for the conterminous United States by hydrologic landscape regions, J. Hydrol., 351(1‐2), 139‐153.
113
Schultz, W. H., D. J. Lidke, and J. W. Godt (2008), Modeling the spatial distribution of landslide‐prone colluvium and shallow groundwater on hillslopes of Seattle, WA, Earth Surf. Proc. and Landforms 33, 123‐141. Seaton, W. J. and T. B. Burbey (2005), Influence of ancient thrust faults on the hydrogeology of the Blue Ridge Province, Ground Water 43(3), 301‐313. Shehata, M. A. and F. M. Al‐Ruwaih (2005), Quantitative geomorphological analysis of some watersheds on the Eastern Bank of the River Nile with relation to basin hydrogeology, Egypt, Kuwait J. of Sci. and Eng. 32(1), 195‐212. Sivapalan, M. (2003), Process complexity at the hillslope scale, process simplicity at the watershed scale: is there a connection? Hydrol. Proc. 17, 1037‐1041. Smakhtin, V. U. (2001), Low flow hydrology: a review, J. Hydrol. 240, 147‐186. Smith, R. E. (1991), Effect of clearfelling pines on water yield in a small eastern Transvaal catchment, South Africa, Water S.A. 17(3), 217‐224. Soulsby, C., P. J. Rodgers, J. Petry, D. M. Hannah, I. A. Malcolm, and S. M. Dunn (2004), Using tracers to upscale flow path understanding in mesoscale mountainous catchments: two examples from Scotland, J. Hydrol. 291, 174‐196. Southworth, S., A. Schultz, D. Denenney, and J. Triplett (2003), Surficial geologic map of the Great Smoky Mountains National Park, U.S. Geol. Surv. Open File Report, 03‐081. Sturdevant‐Rees, P., J. A. Smith, J. Morrison, and M. L. Baeck (2001), Tropical storms and the flood hydrology of the central Appalachians, Water Resour. Res. 37(8), 2143‐2168. Sutherland, A. S., J. L. Meyer, and E. P. Gardiner (2002), Effects of land cover on sediment regime and fish assemblage structure in four southern Appalachian streams, Freshwater Bio. 47, 1791‐1805. Swank, W. T. and J. E. Douglass (1975), Nutrient flux in undisturbed and manipulated forest ecosystems in the southern Appalachian Mountains, in Proceedings of the Tokyo Symposium on the Hydrologic Characteristics of River Basins, Tokyo, Japan, pp. 445‐456. Tetzlaff, D. and C. Soulsby (2008), Sources of baseflow in larger catchments – Using tracers to develop a holistic understanding of runoff generation, J. Hydrol. 359, 287‐302. Tetzlaff, D., J. Siebert, K. J. McGuire, H. Laudon, D. A. Burns, S. M. Dunn, and C. Soulsby (2009), How does landscape structure influence catchment transit times across different geomorphic provinces? Hydrol. Process. 23, 945‐953. Thomas, D. M. and M. A. Benson (1970), Generalization of streamflow characteristics from drainage‐basin characteristics, U.S. Geol. Surv. Water‐Supply Paper 1975.
114
Thomas, M. P. and M. A. Cervione (1970), A proposed streamflow data program for Connecticut, Conn. Water Resour. Bull. 23, ** U.S. Census Bureau (2007), 2006 Population Estimates. Accessed Ocober 28, 2007. <http://factfinder.census.gov> USDA‐NRCS (2005), Soil Survey Geographic (SSURGO) Database for Macon County, North Carolina. Accessed September 10, 2007. <http://soildatamart.nrcs.usda.gov> USDA‐NRCS (2006), Soil Survey Geographic (SSURGO) Database for Rabun and Towns Counties, Georgia, Accessed January 20, 2009 <http://soildatamart.nrcs.usda.gov> USDA‐NRCS (2007), Soil Survey Geographic (SSURGO) Database for Jackson County, North Carolina. Accessed September 10, 2007. <http://soildatamart.nrcs.usda.gov> USDA‐NRCS (2008), Soil Survey Geographic (SSURGO) Database for Clay County, North Carolina, Accessed January 20, 2009 <http://soildatamart.nrcs.usda.gov> USGS (2003), North Carolina Land Cover Database Zone 57 Land Cover Layer, Accessed September 20, 2007. <www.seamless.usgs.gov> Velbel, M. A. (1985), Geochemical mass balances and weathering rates in forested watersheds of the southern Blue Ridge, Am. J. Sci. 285, 904‐930. Vogel, R. M. and C. N. Kroll (1992), Regional geohydrologic‐geomorphic relationships for the estimation of low‐flow statistics, Water Resour. Res. 28(9), 2451‐2458. Warner, G. S., A. R. García‐Martinó, F. N. Scatena, and D. L. Civco (2003), Watershed characterization by GIS for Low Flow Prediction, in GIS for Water Resources and Watershed Management, edited by J. G. Lyon, pp. 101‐107, Taylor and Francis, London. Warren, M. L., B. M. Burr, S. J. Walsh, H. L. Bart, R. C. Cashner, D. A. Etnier, B. J. Freeman, B. R. Kuhajda, R. L. Mayden, H. W. Robison, S. T. Ross, and W. C. Starnes (2002), Diversity, distribution, and conservation status of the native freshwater fishes of the southern United States, Fisheries 25, 7‐31. Wear, D. N. and P. Bolstad (1998), Land use changes in the Southern Appalachian landscapes: spatial analysis and forecast evolution, Ecosystems 1(6), 575‐594. White, B. A. and Burbey, T. J. (2007), Evidence for structurally controlled recharge in the Blue Ridge province, Virginia, USA, Hydrogeology Journal 15(5), 929‐943. Woods, R. A., M. Sivapalan, J. S. Robinson (1997), Modeling the spatial variability of subsurface runoff using a topographic index, Water Resour. Res. 33(5), 1061‐1073. Wooten, R. M., K. A. Gillon, A. C. Witt, R. S. Latham, T. J. Douglas, J. B. Bauer, S. J. Fuemmeler, and L. G. Lee (2008), Geologic, geomorphic, and meteorological aspects of debris flows triggered by Hurricanes Frances and Ivan during September 2004 in the Southern Appalachian Mountains of Macon County, North Carolina (southeastern USA), Landslides 5, 31‐44.
115
Yarnell, S. L. (1998), The southern Appalachians: a history of the landscape. U.S. Dept. of Agric. – Forest Service, Southern Research Station General Technical Report, SRS‐18. Zhu, Y. and R. L. Day (2009), Regression modeling of streamflow, baseflow, and runoff using geographic information systems, J. Env .Mgmt. 90, 946‐953.
116
Table 4.1 – General topographic and land use characteristics of study watersheds. Site numbers correspond to Figure 4.1
# Stream Name Area Developed Forest/Shrub Pasture/Ag Max.Elev. Relief (km2) (%) (%) (%) (m) (m)
1 Buck Cr. 33.8 3.3 96.5 0.2 1535 555
2 Roaring Fork 4.7 0.2 99.9 0.0 1590 634
3 Nantahala River 134.9 2.1 97.5 0.2 1676 739
4 Wayah Cr. 30.6 2.9 96.7 0.3 1650 964
5 Poplar Cove Cr. 9.6 8.5 90.1 1.4 1416 743
6 Allison Cr. 15.2 6.1 90.4 3.4 1514 828
7 Jones Cr. 15.3 3.3 94.3 2.4 1533 840
8 Shope Fork 7.8 1.6 98.4 0.0 1593 893
9 Cartoogechaye Cr. 145.5 8.0 85.9 5.9 1661 1041
10 Iotla Cr. 23.5 8.8 77.5 13.3 1157 550
11 Crawford Br. 6.0 48.9 44.4 6.3 886 276
12 Blaine Br. 3.3 8.4 82.3 9.1 968 341
13 McDowell Br. 3.5 17.0 70.9 11.9 886 269
14 North Fork Skeenah Cr. 6.4 6.5 84.7 8.5 1081 447
15 South Fork Skeenah Cr. 6.0 5.7 90.5 3.8 1113 475
16 Bates Br. 6.3 10.7 76.8 12.1 996 379
17 Fulcher Br. 2.7 12.4 76.3 11.1 1170 551
18 Cowee Cr. 24.3 2.6 95.4 1.9 1510 871
19 Caler Fork 17.4 4.4 93.4 1.9 1361 742
20 Watauga Cr. 16.7 13.0 82.4 4.3 1232 614
21 Rabbit Cr. 22.9 8.5 77.9 13.5 1345 730
22 Nickajack Cr. 6.1 3.6 95.0 0.5 1281 651
23 Savannah Cr. 34.1 5.7 93.7 0.6 1422 731
24 Tathams Cr. 5.9 1.1 98.6 0.3 1311 600
25 Little Ellijay Cr. 10.8 3.0 96.4 0.4 1464 785
26 Little Savannah Cr. 10.1 9.1 83.5 7.0 1048 432
27 Cullowhee Cr. 27.6 3.5 95.6 0.8 1459 797
28 Buff Cr. 9.3 2.7 96.1 1.0 1840 1152
29 Blanton Br. 5.3 11.0 86.0 2.7 1147 488
30 Cope Cr. 8.5 15.4 80.3 4.0 1084 460
31 Cane Cr. 7.7 4.0 95.0 1.0 1238 613
32 Wayehutta Cr. 16.3 3.4 95.8 0.7 1469 840
33 Darnell Cr. 13.7 0.5 99.2 0.0 1405 742
34 Mud Cr. 13.1 23.4 75.1 0.9 1432 778
35 Cullasaja River 48.2 26.7 71.6 1.0 1525 544
117
Metric Abbreviation Unit Calculation Method Reference Excl. Trans.
Basin topographyBas in elevation ‐ max ElevMax m Elevation at highest point in bas in Petras , 2003 SC ‐Bas in elevation ‐ mean ElevMean m Mean elevation of watershed DEM pixels Petras , 2003 SC ‐Bas in elevation ‐ median ElevMed m Median elevation of watershed DEM pixels ‐ * ‐Bas in elevation ‐ min ElevMin m Elevation at bas in outlet Petras , 2003 SC xBas in elevation ‐ st. dev. ElevSD m Standard deviation of watershed DEM pixels Petras , 2003 SC ‐Bas in relative relei f RelRel ief ‐ Bas in Rel ief / Bas in perimeter Fi tzpatrick et al ., 1998 SC ‐Bas in rel ief TotRel ief m Maximum elevation ‐ minimum elevation Fitzpatrick et al ., 1998 PCA ‐Hypsometric index 1 Hyp1 ‐ % change between 25‐75%i le of curve Warner et a l ., 2003 * log10Hypsometric index 2 Hyp2 km‐2 Hypsometric Index 1 / Area Warner et a l ., 2003 PCA ‐Hypsometric index 3 Hyp3 ‐ Hypsometric Index 1 / % change between 50‐75%i le of curve Warner et a l ., 2003 PCA XHypsometric integra l HypInt ‐ Integra l of hypsometric curve McGuire et al ., 2005 PCA ‐Hypsometric kurtos is HypKurt ‐ Kurtos is of hypsometric curve ‐ PCA ‐Hypsometric skewness HypSkew ‐ Skewness of hypsometric curve ‐ PCA ‐Topographic Index ‐ mean TImean ‐ Mean TI of pixels (ln(tan(s lope))/dim. accumul . area) McGuire et al ., 2005 PCA XTopographic Index ‐ st. dev. TISD ‐ Standard deviation of TI of pixels ‐ PCA X
Basin morphometryBas in area Area km2 Area enclosed by dra inage divide Fitzpatrick et al ., 1998 * log10Bas in circulari ty ratio Circ ‐ 4*∏*Bas in area / (Bas in perimeter)2 Apaydin et al ., 2006 SC ‐
Bas in compactness ratio Comp ‐ Bas in perimeter / 2*(∏*Bas in Area)0.5 Apaydin et al ., 2006 PCA ‐
Bas in elongation Elong ‐ 2*(Bas in area/∏)0.5 / Bas in length Apaydin et al ., 2006 * ‐Bas in length Length km Length from watershed outlet to dra inage divide Fitzpatrick et al ., 1998 SC ‐Bas in length ‐ equiva lent L_Equiv km (Bas in perimeter + (Bas in perimeter2 ‐ 16*Bas in area)0.5) / 4 Petras , 2003 SC log10Bas in length ‐ relative L_Rel ‐ Bas in length / (Bas in area)0.5 Petras , 2003 SC log10Bas in perimeter Perim km Length of bas in boundary Warner et a l ., 2003 SC log10Bas in rel ief ratio RelRat ‐ Bas in rel ief / Bas in length*1000 Fitzpatrick et al ., 1998 SC ‐
Bas in ruggedness 1 Rugg1 ‐ Bas in Rel ief / (Bas in area)0.5 Apaydin et al ., 2006 SC ‐Bas in ruggedness 2 Rugg2 ‐ Bas in Rel ief * Drainage dens ity Melton, 1957 SC ‐Bas in shape Form ‐ Bas in area / (Bas in length)2 Fi tzpatrick et al ., 1998 SC ‐
Bas in thickness V/A m (Pixel area * sum of al l pixel elevations ) / Bas in area ‐ SC ‐AspectAspect ‐ east‐facing EF ‐ Fraction of pixels facing 45‐135 degrees Warner et a l ., 2003 PCA asrAspect ‐ north‐facing NF ‐ Fraction of pixels facing 315‐45 degrees Warner et a l ., 2003 PCA asrAspect ‐ south‐facing SF ‐ Fraction of pixels facing 135‐225 degrees Warner et a l ., 2003 * asrAspect ‐ west‐facing WF ‐ Fraction of pixels facing 225‐315 degrees Warner et a l ., 2003 SC asr, expCos(aspect) cos(asp) ‐ Mean cos(aspect) of watershed pixels Brenning and Tromboto, 2006 PCA ‐Sin(aspect) s in(asp) ‐ Mean s in(aspect) of watershed pixels Brenning and Tromboto, 2006 PCA ‐
Table 4.2 (page 1 of 2) –Explanation of watershed characteristics considered for use in multiple regression modeling. The "Excl." column conveys the analysis that removed the variable (SC = simple correlation, PCA = principal components analysis, and * = not removed). The "Trans." column presents the transform that was used to achieve a normal distribution ("‐" = variable was normally distributed raw, "X" = no standard transform normalized the variable, and "asr" indicates an arcsin square root transform was used (in cases of proportions).
118
Metric Unit Calculation Method Reference Excl. Trans.
Basin slopeSlope ‐ 95th percenti le Slope95th ‐ 95th percenti le of pixel s lope distribution ‐ SC ‐Slope ‐ bas in area < 2% Slope<2 ‐ Fraction of pixels less than 2% s lope Warner et a l ., 2003 * log10Slope ‐ bas in area < 5% Slope<5 ‐ Fraction of pixels less than 5% s lope Warner et a l ., 2003 SC log10Slope ‐ bas in area < 10% Slope<10 ‐ Fraction of pixels less than 10% s lope Warner et a l ., 2003 SC log10Slope ‐ bas in area < 20% Slope<20 ‐ Fraction of pixels less than 20% s lope Warner et a l ., 2003 SC log10Slope ‐ kurtos i s SlopeKurt ‐ Kurtos is of pixel s lope dis tribution ‐ PCA XSlope ‐ maximum SlopeMax ‐ Maximum pixel s lope Petras , 2003 * recipSlope ‐ mean SlopeMean ‐ Mean s lope of watershed pixels McGuire et al ., 2005 SC ‐Slope ‐ median SlopeMed ‐ Median s lope of watershed pixels ‐ SC ‐Slope ‐ skewness SlopeSkew ‐ Skewness of pixel s lope distribution ‐ PCA ‐Slope ‐ standard dev. SlopeSD ‐ Standard deviation of watershed s lope pixels ‐ * X
Channel network morphometryBifurcation ratio (count) BR_C ‐ Average (# stream segments orderx) / (# stream segments Shehata and Al ‐Ruwaih, 2005 * ‐Bifurcation ratio (length) BR_L ‐ Average (sum length orderx) / (sum length orderx+1) ≠ orderx_max ‐ PCA XTributary/trunk ratio ChaTri ‐ Tota l tributary length / Trunk stream length Warner et a l ., 2003 SC ‐Drainage dens ity DD km‐1 Tota l s tream length / Bas in Area Fitzpatrick et al ., 1998 * ‐Entire s tream gradient Bas inSl ‐ (Elev. at 85% length ‐ elev. at 10% length / (85% ‐ 10% length) Fi tzpatrick et al ., 1998 PCA log10Firs t order s tream fraction %1st ‐ Fi rs t order stream length / Tota l s tream length ‐ * ‐Tota l stream length TotLength km Sum of segment lengths (us ing accum. threshold of 18,050m2) Fi tzpatrick et al ., 1998 SC log10
Soil and bedrockSoi l ‐ al luvium Al luv ‐ Fraction of bas in area mapped as al luvium parent matera l Tetzlaff and Soulsby, 2008 * asrSoi l ‐ col luvium Col luv ‐ Fraction of bas in area mapped as col luvium parent matera l Tetzlaff and Soulsby, 2008 * asrSoi l ‐ res iduum Res id ‐ Fraction of bas in area mapped as res iduum parent matera l White and Burbey, 2007 PCA asr, XSoi l ‐ sand fraction Sand ‐ Area ‐weighted mean sand fraction of watershed soi l s Batelaan and De Smedt, 2007 PCA asr, XSoi l ‐ s i l t fraction Si l t ‐ Area ‐weighted mean s i l t fraction of watershed soi l s Batelaan and De Smedt, 2007 PCA asr, XSoi l ‐ clay fraction Clay ‐ Area ‐weighted mean clay fraction of watershed soi l s Batelaan and De Smedt, 2007 * asr, XBedrock ‐ class 2 conuctivi ty Geol2 ‐ Fraction of watershed bedrock geology mapped as category 2 ‐ PCA X
Land useImpervious surface area Imperv ‐ As fraction of watershed area, ca lculated from NLCD class SCDeveloped Dev ‐ Fraction of watershed area in NLCD classes 21, 22, 23, and 24 SCForest and shrub For ‐ Fraction of watershed area in NLCD classes 41, 42, 43, and 52 *Pasture and agricul ture Pas ‐ Fraction of watershed area in NLCD classes 71, 81, and 82 SCBarren Barren ‐ Fraction of watershed area in NLCD class 31 PCAWetland Wetland ‐ Fraction of watershed area in NLCD class 90 PCAOpen Water Water ‐ Fraction of watershed area in NLCD class 11 PCA
Table 4.2 (page 2 of 2) ‐ Explanation of watershed characteristics considered for use in multiple regression modeling. The "Excl." column conveys the analysis that removed the variable (SC = simple correlation, PCA = principal components analysis, and * = not removed). The "Trans." column presents the transform that was used to achieve a normal distribution ("‐" = variable was normally distributed raw, "X" = no standard transform normalized the variable, and "asr" indicates an arcsin square root transform was used (in cases of proportions).
119
For
ElevM
edHy
p1
Elong
SF Slope
SD
Slope
<2BR
_C
DD %1st
Alluv
Collu
v
Clay
PPTL
F07
PPTL
F08
PPTW
Y08
Area
For R 1 0.667 ‐0.270 0.243 ‐0.083 0.165 ‐0.787 0.297 ‐0.184 ‐0.077 ‐0.752 0.507 ‐0.151 ‐0.126 0.180 0.121 0.178p . 0.000 0.117 0.159 0.634 0.344 0.000 0.083 0.291 0.661 0.000 0.002 0.387 0.472 0.301 0.487 0.307
ElevMed R 0.667 1 ‐0.036 ‐0.029 0.129 ‐0.062 ‐0.583 0.290 ‐0.317 ‐0.208 ‐0.590 0.230 ‐0.112 ‐0.381 0.421 0.357 0.399p 0.000 . 0.839 0.868 0.459 0.723 0.000 0.091 0.064 0.230 0.000 0.184 0.522 0.024 0.012 0.035 0.018
Hyp1 R ‐0.270 ‐0.036 1 0.055 ‐0.029 ‐0.082 0.425 ‐0.182 ‐0.096 ‐0.088 0.218 0.015 ‐0.205 ‐0.141 0.184 0.191 0.469p 0.117 0.839 . 0.755 0.870 0.641 0.011 0.297 0.582 0.616 0.209 0.930 0.236 0.419 0.291 0.271 0.004
Elong R 0.243 ‐0.029 0.055 1 ‐0.012 0.248 ‐0.181 ‐0.139 ‐0.004 ‐0.194 ‐0.277 0.322 ‐0.295 0.242 ‐0.189 ‐0.139 ‐0.022p 0.159 0.868 0.755 . 0.947 0.151 0.297 0.427 0.982 0.264 0.107 0.059 0.086 0.161 0.278 0.427 0.900
SF R ‐0.083 0.129 ‐0.029 ‐0.012 1 0.033 0.124 0.219 ‐0.218 0.058 0.117 ‐0.171 0.123 0.198 ‐0.093 ‐0.073 0.150p 0.634 0.459 0.870 0.947 . 0.850 0.478 0.206 0.209 0.741 0.505 0.326 0.481 0.255 0.597 0.678 0.391
SlopeSD R 0.165 ‐0.062 ‐0.082 0.248 0.033 1 0.081 0.091 0.164 0.118 ‐0.092 0.470 ‐0.237 0.166 ‐0.016 ‐0.009 0.156p 0.344 0.723 0.641 0.151 0.850 . 0.644 0.603 0.348 0.501 0.600 0.004 0.170 0.340 0.926 0.960 0.371
Slope<2 R ‐0.787 ‐0.583 0.425 ‐0.181 0.124 0.081 1 ‐0.182 0.210 0.178 0.687 ‐0.186 0.029 0.048 0.024 0.082 0.155p 0.000 0.000 0.011 0.297 0.478 0.644 . 0.295 0.226 0.307 0.000 0.284 0.867 0.784 0.891 0.640 0.374
BR_C R 0.297 0.290 ‐0.182 ‐0.139 0.219 0.091 ‐0.182 1 ‐0.094 0.039 ‐0.279 0.080 0.044 ‐0.069 0.015 ‐0.024 0.174p 0.083 0.091 0.297 0.427 0.206 0.603 0.295 . 0.591 0.823 0.105 0.649 0.804 0.694 0.933 0.893 0.319
DD R ‐0.184 ‐0.317 ‐0.096 ‐0.004 ‐0.218 0.164 0.210 ‐0.094 1 0.004 0.313 0.073 0.014 0.115 ‐0.215 ‐0.191 ‐0.093p 0.291 0.064 0.582 0.982 0.209 0.348 0.226 0.591 . 0.983 0.067 0.676 0.938 0.512 0.214 0.271 0.595
%1st R ‐0.077 ‐0.208 ‐0.088 ‐0.194 0.058 0.118 0.178 0.039 0.004 1 0.014 ‐0.133 0.103 0.108 ‐0.135 ‐0.174 ‐0.020p 0.661 0.230 0.616 0.264 0.741 0.501 0.307 0.823 0.983 . 0.935 0.448 0.554 0.537 0.439 0.319 0.908
Alluv R ‐0.752 ‐0.590 0.218 ‐0.277 0.117 ‐0.092 0.687 ‐0.279 0.313 0.014 1 ‐0.511 0.525 0.116 ‐0.058 ‐0.007 ‐0.071p 0.000 0.000 0.209 0.107 0.505 0.600 0.000 0.105 0.067 0.935 . 0.002 0.001 0.507 0.740 0.969 0.684
Colluv R 0.507 0.230 0.015 0.322 ‐0.171 0.470 ‐0.186 0.080 0.073 ‐0.133 ‐0.511 1 ‐0.459 ‐0.148 0.138 0.155 0.090p 0.002 0.184 0.930 0.059 0.326 0.004 0.284 0.649 0.676 0.448 0.002 . 0.006 0.395 0.430 0.372 0.605
Clay R ‐0.151 ‐0.112 ‐0.205 ‐0.295 0.123 ‐0.237 0.029 0.044 0.014 0.103 0.525 ‐0.459 1 ‐0.038 ‐0.192 ‐0.208 ‐0.128p 0.387 0.522 0.236 0.086 0.481 0.170 0.867 0.804 0.938 0.554 0.001 0.006 . 0.830 0.270 0.229 0.462
PPTLF07 R ‐0.126 ‐0.381 ‐0.141 0.242 0.198 0.166 0.048 ‐0.069 0.115 0.108 0.116 ‐0.148 ‐0.038 1 ‐0.556 ‐0.573 0.052p 0.472 0.024 0.419 0.161 0.255 0.340 0.784 0.694 0.512 0.537 0.507 0.395 0.830 . 0.001 0.000 0.766
PPTLF08 R 0.180 0.421 0.184 ‐0.189 ‐0.093 ‐0.016 0.024 0.015 ‐0.215 ‐0.135 ‐0.058 0.138 ‐0.192 ‐0.556 1 0.985 0.190p 0.301 0.012 0.291 0.278 0.597 0.926 0.891 0.933 0.214 0.439 0.740 0.430 0.270 0.001 . 0.000 0.275
PPTWY08 R 0.121 0.357 0.191 ‐0.139 ‐0.073 ‐0.009 0.082 ‐0.024 ‐0.191 ‐0.174 ‐0.007 0.155 ‐0.208 ‐0.573 0.985 1 0.139p 0.487 0.035 0.271 0.427 0.678 0.960 0.640 0.893 0.271 0.319 0.969 0.372 0.229 0.000 0.000 . 0.425
Area R 0.178 0.399 0.469 ‐0.022 0.150 0.156 0.155 0.174 ‐0.093 ‐0.020 ‐0.071 0.090 ‐0.128 0.052 0.190 0.139 1p 0.307 0.018 0.004 0.900 0.391 0.371 0.374 0.319 0.595 0.908 0.684 0.605 0.462 0.766 0.275 0.425 .
Table 4.3 – Correlations among independent variables included in regression analysis. N = 35 for all variables. Variable abbreviations are defined in Table 4.2, except for PPTLF07, PPTLF08, and PPTWY08, which represent the precipitation totals for Low Flow Season 2007, Low Flow Season 2008, and Water Year 2008 respectively. Only one precipitation total was included in any given analysis (corresponding to the appropriate time period).
120
Table 4.4. Precipitation stations. Coordinates are UTM‐NAD 83. Data sources are the State Climate Office of North Carolina (SCO), National Climatic Data Center (NCDC), Integrated Flood Warning System (IFLOWS), and the USFS‐Coweeta Hydrologic Laboratory (CHL). Station numbers correspond to Figure 4.5.
Station #
Station Name Elevation
(m) Easting (m)
Northing (m)
Source
1 Brasstown Bald 991 241770 3854462 SCO 2 Hayesville 1 SW 580 241879 3882247 NCDC 3 Robbinsville Ag 5 678 243547 3909913 NCDC 4 Cheoah 640 243754 3913182 SCO 5 Stecoah Gap 1122 255334 3913631 IFLOWS 6 Robbinsville 1 S 608 255852 3917445 NCDC 7 Wayah Bald Mountain 1663 264956 3895294 SCO 8 Harrison Gap 825 272275 3887741 IFLOWS 9 Potato Knob 1004 272949 3921735 IFLOWS 10 Moody Gap 1364 274786 3879556 CHL 11 Coweeta Exp Station 686 278354 3882470 NCDC 12 Mountain City 2 SSW 1056 278877 3864872 NCDC 13 Bryson City 2 618 279127 3924823 NCDC 14 Macon County Airport 611 279717 3900135 SCO 15 Clayton 1 SSW 584 280633 3860932 NCDC 16 Mountain City 2 N 657 280685 3868728 NCDC 17 Franklin 648 281471 3895566 NCDC 18 Otto 623 282421 3882066 IFLOWS 19 Macon Middle 626 285105 3893706 CHL 20 Wesser 1291 286219 3914016 IFLOWS 21 Oconaluftee 614 291413 3932039 NCDC 22 W Frk Dicks Crk 966 295698 3923137 IFLOWS 23 Pumpkintown 1417 295821 3903772 IFLOWS 24 Sylva Kings Mtn 915 298100 3915523 IFLOWS 25 Highlands 1170 300164 3880792 NCDC 26 Cullowhee 668 301577 3912076 NCDC 27 Soco Gap 1609 305594 3929924 IFLOWS 28 Tuckasegee 1036 306297 3900192 IFLOWS 29 Cedar Cliff 671 308648 3903162 IFLOWS 30 Balsam Gap 1027 311364 3923043 IFLOWS 31 Jocassee 8 WNW 762 311368 3873391 NCDC 32 Argura WS 1012 314258 3904752 IFLOWS 33 Charley Ridge 1256 319027 3906113 IFLOWS 34 Waynesville 1 E 810 321681 3928771 NCDC 35 Lake Toxaway 838 326117 3888363 IFLOWS
121
Dependent Variable
Time Period R2 Adj. R2 F (p)
Total DF
Independent Variables
Std. Coeff
t (p)
Q99 LF07 0.441 0.377 6.841 (0.002) 29 Forest 0.432 2.848 (0.008)1st order ‐0.318 ‐2.133 (0.043)Slope Std. Dev. 0.314 2.056 (0.050)
Q99 LF08 0.363 0.319 8.269 (0.001) 31 Precipitation 0.496 3.349 (0.002)Slope < 2% ‐0.348 ‐2.357 (0.026)
Q99 WY08 0.265 0.215 5.241 (0.011) 31 Precipitation 0.364 2.256 (0.032)Forest 0.310 2.131 (0.043)
Qmin1 LF07 0.537 0.463 7.238 (0.001) 29 Slope Std. Dev. 0.445 2.79 (0.010)Drainage Density ‐0.395 ‐2.706 (0.012)1st order ‐0.312 ‐2.248 (0.034)Colluvium 0.364 2.311 (0.029)
Qmin1 LF08 0.218 0.192 8.382 (0.007) 31 Drainage Density ‐0.467 ‐2.895 (0.007)
Qmin1 WY08 0.578 0.515 9.232 (0.001) 31 Slope Std. Dev. 0.488 3.789 (0.001)Drainage Density ‐0.332 ‐2.519 (0.018)Bifurcation Ratio 0.274 2.086 (0.047)South‐facing slopes 0.261 2.078 (0.048)
Qmin7 LF07 0.472 0.411 7.742 (0.001) 29 Drainage Density ‐0.409 ‐2.688 (0.012)Colluvium 0.411 2.526 (0.018)Slope Std. Dev. 0.418 2.523 (0.018)
Qmin7 LF08 0.292 0.243 5.982 (0.001) 31 Drainage Density ‐0.444 ‐2.839 (0.008)Colluvium 0.319 2.042 (0.032)
Qmin7 WY08 0.298 0.222 3.955 (0.018) 31 Drainage Density ‐0.459 ‐2.871 (0.022)Slope Std. Dev. 0.361 2.256 (0.023)
Qmin14 LF07 0.542 0.468 1.382 (0.000) 29 Slope Std. Dev. 0.470 2.952 (0.007)Colluvium 0.374 2.402 (0.024)Drainage Density ‐0.360 ‐2.480 (0.027)1st order ‐0.282 ‐2.098 (0.048)
Qmin14 LF08 0.191 0.164 7.102 (0.012) 31 Drainage Density ‐0.438 ‐2.665 (0.012)
Qmin14 WY08 0.281 0.232 5.677 (0.008) 31 Drainage Density ‐0.446 ‐2.787 (0.009)Slope Std. Dev. 0.337 2.355 (0.026)
BFI LF07 0.482 0.399 5.811 (0.002) 29 Forest 0.727 3.551 (0.002)Area 0.524 3.303 (0.003)Clay 0.483 3.027 (0.006)Median Elev. ‐0.519 ‐2.500 (0.019)
BFI LF08 0.314 0.241 4.275 (0.013) 31 Area 0.483 2.832 (0.008)Forest 0.529 2.481 (0.019)Median Elev. ‐0.547 ‐2.409 (0.023)
BFI WY08 0.302 0.227 4.031 (0.017) 31 Area 0.412 2.496 (0.019)Clay 0.414 2.310 (0.028)Forest 0.370 2.124 (0.043)
Table 4.5 ‐ Best models for each dependent variable, as determined by best subsets regression and Mallow’s Cp. Dependent variable and time period abbreviations are explained in section 3.3 of the text.
122
Table 4.6 ‐ Comparison of paired watershed flows. Each pair is comprised of topographically similar, proximal waterhseds, whose forest cover differs. These pairs demonstrate that greater reductions in watershed forest cover are associated with greater reductions in stream baseflows. All flows are expressed as watershed area‐standardized discharge (m3s‐1km‐2). “ % +/‐ “ indicates differences in flow, calculated as percent of the flow of the lower forest watershed; watershed forest cover differences (in parentheses) are simple magnitude differences. Site numbers correspond to Table 4.1 and Figure 4.1. Flow metric abbreviations are explained in section 4.3.3 of the text.
Flow Metric
Time Period
Site 12 Site 17 % +/‐ Site 24 Site 29 % +/‐ Site 19 Site 21 % +/‐ Site 33 Site 34 % +/‐
Mean LF07 ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ 9.22E‐03 7.74E‐03 +19 LF08 1.17E‐02 1.21E‐02 ‐3 4.58E‐03 5.26E‐03 ‐13 6.44E‐03 5.15E‐03 +25 1.17E‐02 7.38E‐03 +59 WY08 1.43E‐02 8.91E‐03 +60 9.22E‐03 6.98E‐03 +32 ‐ ‐ ‐ 1.93E‐02 1.61E‐02 +20 Q99 LF07 ‐ ‐ ‐ ‐ ‐ . ‐ ‐ ‐ 3.91E‐03 3.31E‐03 +18 LF08 2.79E‐03 2.06E‐03 +36 1.79E‐03 1.22E‐03 +47 3.28E‐03 1.41E‐03 +132 4.17E‐03 2.38E‐03 +78 WY08 2.54E‐03 2.31E‐03 +10 2.06E‐03 1.38E‐03 +49 ‐ ‐ ‐ 4.81E‐03 2.62E‐03 +83 Qmin1 LF07 ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ 4.05E‐03 3.57E‐03 +13 LF08 3.07E‐03 2.46E‐03 +25 1.82E‐03 1.45E‐03 +26 3.46E‐03 1.94E‐03 +78 4.59E‐03 2.60E‐03 +76 WY08 2.64E‐03 2.46E‐03 +7 1.84E‐03 1.45E‐03 +27 ‐ ‐ ‐ 4.60E‐03 2.60E‐03 +77 Qmin7 LF07 ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ 4.38E‐03 3.92E‐03 +12 LF08 3.70E‐03 2.61E‐03 +41 0.00213 0.001548 +38 3.63E‐03 2.08E‐03 +74 5.08E‐03 2.96E‐03 +72 WY08 3.43E‐03 2.61E‐03 +31 2.13E‐03 1.55E‐03 +38 ‐ ‐ ‐ 5.08E‐03 2.96E‐03 +72 Qmin14 LF07 ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ 4.70E‐03 4.12E‐03 +14 LF08 3.85E‐03 2.83E‐03 +36 2.18E‐03 1.64E‐03 +33 3.71E‐03 2.34E‐03 +58 5.62E‐03 3.12E‐03 +80 WY08 3.49E‐03 2.87E‐03 +22 0.002215 1.64E‐03 +35 ‐ ‐ ‐ 5.62E‐03 3.12E‐03 +80 BFI LF07 ‐ ‐ 7.25E‐01 5.88E‐01 +23 LF08 5.72E‐01 6.66E‐01 ‐14 0.657 7.55E‐01 ‐13 7.58E‐01 7.26E‐01 +4 7.56E‐01 5.39E‐01 +40 WY08 6.85E‐01 7.39E‐01 ‐7 0.72 6.54E‐01 +10 7.62E‐01 6.85E‐01 +11 Watershed Forest (%) 82.3 76.3 (6.0) 98.6 86.0 (12.6) 93.4 77.9 (15.5) 99.2 75.1 (24.1)
123
Table 4.7 ‐ Correlations between land use and baseflow metrics. All baseflow metrics except the dimensionless BFI are expressed as watershed area‐standardized discharge (m3s‐1km‐2). Baseflow metric abbreviations are explained in section 4.3.3 of the text.
Baseflow Metric
Time Period
Forest Developed Pasture Impervious n
R (p) R (p) R (p) R (p)
Q99 LF07 0.529 (0.003) ‐0.455 (0.011) ‐0.496 (0.005) ‐0.448 (0.013) 30
Q99 LF08 0.426 (0.015) ‐0.355 (0.046) ‐0.472 (0.006) ‐0.413 (0.019) 32
Q99 WY08 0.370 (0.037) ‐0.307 (0.088) ‐0.390 (0.027) ‐0.343 (0.054) 32
Qmin1 LF07 0.379 (0.039) ‐0.320 (0.085) ‐0.358 (0.052) ‐0.308 (0.098) 30
Qmin1 LF08 0.237 (0.192) ‐0.159 (0.385) ‐0.370 (0.037) ‐0.254 (0.160) 32
Qmin1 WY08 0.249 (0.170) ‐0.197(0.279) ‐0.284 (0.115) ‐0.231 (0.202) 32
Qmin7 LF07 0.381 (0.038) ‐0.323 (0.082) ‐0.339 (0.067) ‐0.294 (0.115) 30
Qmin7 LF08 0.339 (0.058) ‐0.269 (0.137) ‐0.413 (0.019) ‐0.335 (0.061) 32
Qmin7 WY08 0.341 (0.056) ‐0.270 (0.135) ‐0.405 (0.021) ‐0.303 (0.092) 32
Qmin14 LF07 0.371 (0.043) ‐0.309 (0.097) ‐0.359 (0.052) ‐0.290 (0.120) 30
Qmin14 LF08 0.348 (0.051) ‐0.282 (0.118) ‐0.414 (0.019) ‐0.354 (0.047) 32
Qmin14 WY08 0.287 (0.111) ‐0.228 (0.209) ‐0.329 (0.066) ‐0.267 (0.139) 32
BFI LF07 0.276 (0.140) ‐0.303 (0.104) ‐0.059 (0.756) ‐0.183 (0.333) 30
BFI LF08 0.251 (0.167) ‐0.272 (0.131) ‐0.122 (0.508) ‐0.200 (0.273) 32
BFI WY08 0.264 (0.144) ‐0.261 (0.149) ‐0.157 (0.391) ‐0.256 (0.158) 32
124
Figure 4.1 ‐ Study area and monitored watersheds. Watershed numbers correspond to site numbers in Table 4.1. Points A and B are the Coweeta and Cullowhee climate stations, respectively, for which long‐term normals are presented in the text.
125
Figure 4.2 ‐ Representative example of rating curves developed using different curve fitting methods.
126
Regional hydrologic conditions during study periodLittle Tennessee River at Prentiss ‐ USGS 03500000
1/1/07
2/1/07
3/1/07
4/1/07
5/1/07
6/1/07
7/1/07
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
11/1/08
12/1/08
1/1/09
Discharge
(m3 s
‐1)
1
10
August 2008: Lowest flows
on record (69 years)
Low Flow Season 2007
(LF07)
Low Flow Season 2008
(LF08)
Tropical Storm Fay
Water Year 2008 (WY08)
Figure 4.3 – Hydrologic conditions during the study period
127
Figure 4.4 ‐ Examples of Bayesian power law rating curves used in this study, representing the range of scatter in the stage‐discharge relationships. Rating curves with R2 values for all sites are presented in Appendix A.
Excellent QualityWatauga Creek
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7
0.1
1
10
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7 0.8
Q (m
3 s‐1)
0.01
0.1
1
Relatively Low QualityCope Creek
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.01
0.1
1
10
Representative QualityMcDowell Branch
R2 = 0.997 R2 = 0.950 R2 = 0.829
128
Figure 4.5 ‐ Interpolations of precipitation for the three time periods used in this study. Precipitation station numbers correspond to Table 4.4.
129
Figure 4.6 ‐ Ranges of values for baseflow metrics across all study watersheds. The upper and lower limits of the boxes represent the interquartile range, with the black line representing the median value and the white line representing the mean. Whiskers represent one standard deviation, and dots represent all outliers. Baseflow metric abbreviations are explained in section 4.3.3 of the text.
Variability of area‐standardized flows and Baselow Index (BFI) across sitesWatershed
area‐standardized
discharge (m
3 s‐1km
‐2)
0.001
0.01
Qmean
Q99
Qmin1
Qmin7
Qmin14
LF07
LF07LF07
LF07 LF07
LF08
LF08
LF08
LF08
LF08
WY08
WY08WY08WY08 WY08
BFI (baseflo
w / to
tal streamflo
w)
0.3
0.4
0.5
0.6
0.7
0.8
0.9
LF07
LF08
WY08
BFI
130
Figure 4.7 ‐ Difference of mean baseflows between lower‐ and higher‐forest cover watersheds. T and p values presented reflect the results of standard t‐tests. Lower‐ and higher‐forest cover groups were determined using K‐means cluster analysis. Baseflow metric and time period abbreviations are explained in section 4.3.3 of the text.
Comparison of mean baseflows of lower‐ and higher‐forest cover watersheds
Watershed
area‐standardized
discharge
(m3 s
‐1km
‐2)
0.000
0.002
0.004
0.006
0.008
0.010Q99 Qmin1 Qmin7 Qmin14
LF 07t = ‐3.794p = 0.000
LF 08t = ‐2.929p = 0.006
WY 08t = ‐3.001p = 0.005
WY 08t = ‐2.363p = 0.025
LF 08t = ‐1.880p = 0.070
LF 07t = ‐2.719p = 0.011
LF 07t = ‐2.664p = 0.013
LF 08t = ‐2.163p = 0.039
WY 08t = ‐2.267p = 0.031
LF 07t = ‐2.665p = 0.013
WY 08t = ‐2.299p = 0.029
LF 08t = ‐2.176p = 0.036
BFI (baseflow / to
tal flow)
0.0
0.2
0.4
0.6
0.8
1.0BFI
LF 07t = ‐‐1.063p = 0.296
WY 08t = ‐0.745p = 0.462
LF 08t = ‐1.304p = 0.202
Lower forest cover (44‐86%) Higher forest cover (90‐100%)
131
Figure 4.8 – Examples of varied recharge responses to Tropical Storm Fay.
Examples of varied rechargeresponse to a large tropical storm
8/18/0
8
8/25/0
8
9/1/08
9/8/08
9/15/0
8
Q (m
3 s-1)
0.1
1
10
Iotla CreekCullowhee Creek
132
CHAPTER 5
A TEST OF A DISTRIBUTED, GIS‐BASED HYDROLOGIC MODEL FOR EVALUATING BASEFLOW RESPONSE TO LAND‐USE CHANGE IN THE SOUTHERN BLUE RIDGE MOUNTAINS OF NORTH CAROLINA1
________ 1Price, K., Jackson, C.R., and Parker, A.J. To be submitted to Journal of the American Water Resources Association
133
ABSTRACT
Watershed land‐use change has been shown to influence baseflows in many settings. Empirical studies
have shown that soil compaction and impervious surface reduce infiltration of water to subsurface
storage, thereby reducing baseflows. A major hindrance to a complete understanding of the influence
of watershed characteristics on streamflow is the unavailability of empirical streamflow data for a wide
range of watershed conditions within a given study area. The objectives of this study were: 1) to test
the distributed model WetSpa in predicting streamflow for four watersheds in the Blue Ridge province
of North Carolina, using 16 months of measured streamflow as the basis of comparison, 2) to evaluate
baseflow over a 30‐year simulation period under five development scenarios, ranging from undisturbed
to extreme additions of impervious surface, and 3) to evaluate baseflow response over a 30‐year
simulation period to three spatial arrangements of land use. WetSpa produced moderately successful
simulations of the 16‐months of measured streamflow, but failed to reproduce important trends
demonstrated by empirical data from the region. Empirical data shows higher watershed forest cover
associated with higher baseflows, compared with more‐developed watersheds. However, the simulated
streamflow showed universal increases in baseflow magnitudes as watershed forest cover decreased
and developed land use increased. This contradiction is attributed to problematic theoretical
assumptions within the model that the evapotranspiration rates of forests over‐ride gains to subsurface
storage and baseflow due to higher soil infiltration rates. While simulated streamflows showed
increases in baseflow magnitude with greater development, the ratio of baseflow to total streamflow
(baseflow index, or BFI) was higher in less‐developed land use scenarios, corroborating empirical data.
Varying the spatial arrangement of land use did alter the simulated streamflows, but there were no
trends that were consistent across all four watersheds. The contrast between simulated and measured
streamflows regarding baseflow response to development, resulting from embedded theory within the
134
model, should serve as a strong caution regarding the application of distributed hydrologic modeling
without an adequate understanding of the assumptions built into the structure of such models.
5.1. Introduction
Baseflow is the portion of streamflow that is sustained between precipitation events, fed to
stream channels by subsurface and other slow pathways. Watershed land use has been shown to
strongly influence baseflows in many settings (refer to Chapter 2 for a thorough review). Human
landscape impacts such as soil compaction and impervious surface additions often drastically reduce the
amount the water entering subsurface storage, often resulting in reduced baseflows (Rose and Peters,
2001; Chang, 2007). Understanding baseflow is of great importance, as these flows are critical to issues
of water quality, water supply, and aquatic habitat (Johnson, 1998).
A major limitation to understanding the influence of watershed characteristics on streamflow is
the unavailability of streamflow records for a wide range of watershed conditions within a given region.
Toward this end, many regulatory agencies, watershed protection groups, and water resource scientists
are turning to the use of hydrologic models for prediction of streamflow response to watershed
development, for estimation of streamflow extremes for the needs of civil engineering and related
purposes. As the user‐friendliness of geographic information systems (GIS) continues to improve, GIS‐
interfaced hydrologic models are becoming increasingly emphasized as cost‐effective approaches to
watershed management. In such models, spatial data on watershed characteristics such as landcover,
soils, and topography are used as a basis for routing precipitation through a watershed and predicting
streamflow at a designated outlet. Ever‐increasing computing capabilities have allowed for widespread
use of highly‐paramterized, fully‐distributed models, in which water budgets and flow routing are
calculated on the scale of an individual pixel. The underlying assumption behind the use of these more
135
complex models is that the added complexity improves model accuracy, but there is little or no evidence
to support this (Jakeman and Hornberger, 1993; Beven, 2001; Abu El‐Nasr et al., 2005).
It is certainly the case that the graphical user interface (GUI) provided by the GIS software
packages affords more readily usable modeling than options in the past, which required extensive
programming knowledge. However, as increasing numbers of user‐friendly models become available,
along with increasing publicly‐available spatial data, the potential ramifications of exaggerated trust in
such models becomes of paramount concern. Hydrologic models are sophisticated tools, but much of
the theory upon which these models are built remains far from truly established. Furthermore, many of
the physically‐based processes upon which the models are built are scale‐dependent but not treated as
such within the modeling framework (Beven, 2001).
Further complicating matters for predicting baseflow response to land‐use change, most fully
distributed model development has emphasized flood prediction, and few distributed models have been
evaluated to predict extreme low flows. Nyenje and Batelaan (2009) used WetSpa to predict baseflow
response to climate change in Uganda. Cao et al. (2006) successfully calibrated SWAT for baseflow
prediction in highly topographically variable mountainous terrain. Mulungu et al., (2005) independently
developed a physically‐based, distributed model with an emphasis on subsurface‐surface interactions
and baseflow prediction. Coarse‐resolution, grid‐based distributed models have been used in Germany
for analysis of recharge volumes associated with varied bedrock types (Bogena et al., 2005). These
examples highlight many important applications of baseflow simulation using distributed modeling, but
also underscore the absence of recent applications in which distributed models are used to analyze
baseflow levels under varied land use scenarios.
The objective of this study was to test the capabilities of the model WetSpa (Water and Energy
Transfer among Soil, Plants, and Atmosphere ) to 1) reproduce observed discharges during extreme
drought conditions, and 2) predict streamflow differences under varied landcover scenarios, as observed
136
in a related empirical study. Four watersheds in Macon and Jackson counties, NC, were selected for
study, based on the availability of continuous streamflow data for model calibration and evaluation, and
proximity to climate stations with at least 30 years of record. Shope Fork, Iotla Creek, and
Cartoogechaye Creek are tributaries of the Little Tennessee River, and Cane Creek is a tributary to the
Tuckasegee River (Figure 1). These watersheds range in size from 7.8 to 145.5 km2. The sites are
located in the Blue Ridge physiographic province, which is characterized by relatively high topographic
relief and dense, crystalline bedrock.
5.2. Model Description
There are many publicly available hydrological models. Most notably, SWAT (Soil and Water
Assessment Tool), BASINS‐HSPF (Hydrologic Simulation Program – Fortran), DHSVM (Distributed
Hydrologic Soil and Vegetation Model), and WetSpa have been widely used. Of these models, SWAT and
HSPF are semi‐distributed, meaning that water budgets and flow routing are calculated at the
subwatershed scale, and WetSpa and DHSVM are fully distributed, meaning water is routed through
each individual pixel. Exploratory comparisons of model accuracy among SWAT, BASINS‐HSPF, using
Black Earth Creek in Wisconsin and Shope Fork in North Carolina as test watersheds, showed that flows
simulated by WetSpa best matched observed flows. For this reason, WetSpa was chosen for use in this
study.
WetSpa is a physically based, distributed hydrological model for flood prediction and watershed
management. The original model was developed in the Vrije Universiteit Brussel, Belgium in 1996 and
has since been upgraded to incorporate additional spatially distributed parameters. This study
specifically uses the WetSpa Extension, a GIS‐based variant of the original model that has been modified
to include flow routing components, snowmelt modeling, lateral interflow, groundwater flow, and
depression storage (Liu and DeSmedt, 2004). This model captures the physical characteristics of
137
topography, soil, and land use on runoff production, as well as the water and energy balance, through
distributed model parameters over grid cells of a basin. WetSpa has been successfully used for
assessment of land use impacts on flooding in various countries (Liu and DeSmedt, 2005; Liu et al., 2005;
Bahremand et al., 2006), for spatial analysis of runoff distribution (Liu et al., 2006), and for prediction of
streamflow response to varied restoration scenarios (Liu et al., 2004).
The structure of WetSpa’s simplified hydrologic system consists of four control volumes: the
plant canopy, the soil surface, the root zone, and the saturated groundwater aquifer. Based on these
volumes, the water balance for each grid cell is maintained by accounting for the processes of
precipitation, interception, snowmelt, depression storage, infiltration, evapotranspiration, percolation,
surface runoff, interflow, and groundwater flow. The model’s parameterization has two main
components: 1) a hillslope pixel component in which a water budget is calculated to determine the
quantity of water available for routing as overland flow, interflow (shallow subsurface flow), and
percolation to groundwater storage; and 2) channel network characteristics (Figure 2). Both
components are used for flow routing to determine rates of water transmission through the watershed.
WetSpa operation begins with GIS pre‐processing in ESRI ArcView 3.x software, using the
WetSpa extension, to develop spatial attribute files, for subsequent incorporation into a FORTRAN
program that performs flow routing through the watershed. Required spatial data inputs for GIS pre‐
processing include 1) topographic data as a digital elevation model (DEM) in grid format, 2) land use
data in grid format, 3) soil textural data in grid format, and 4) locations of precipitation stations and
stream gages in point shapefile format. Additional database files must be developed to accompany
these spatial layers to serve as look‐up tables during GIS pre‐processing.
GIS pre‐processing results in assignment of hydrologic attributes to each cell, which summarize
the likelihood of evapotranspiration loss, surface runoff vs. interflow, or retention of water based on
topographic, land use, and soils characteristics as defined by the look‐up tables. Model operation
138
incorporates precipitation input over the watershed (via inverse‐distance‐weighted interpolation) and
calculates the outflow from each cell based on velocity and dissipation characteristics for overland flow,
and Darcy’s Law and kinematic approximations for interflow. GIS pre‐processing additionally includes
the development of a flow path network and assignment of channel characteristics. The water that exits
each cell ultimately intersects the stream network, at which point the quantification of flow response at
the flow path level is followed. More detailed information on mathematical formulation for water
budgets and flow routing are presented in Appendix C, and a complete discussion of the underlying
theory is presented in the WetSpa user manual (Liu and DeSmedt, 2004).
There are 10 “global parameters” whose main function is model calibration. These are applied
to the entire watershed, and cannot be adjusted to individual land use or soil classes:
1. Correction factor for estimating AET from PET
2. Interflow scaling factor
3. Groundwater recession constant scaling factor
4. Initial soil moisture
5. Initial groundwater storage
6. Base temperature for snowmelt
7. Temperature degree‐day coefficient
8. Rainfall degree‐day coefficient
9. Exponent that determines the proportion of surface runoff associated with very low rainfall
intensity
10. Rainfall intensity at which surface runoff becomes a linear function of soil moisture content
In addition to the spatial watershed characteristics, these global parameters influence the proportion of
streamflow that contributes to flood peaks during storm events vs. entering the stream system as
baseflow, and are thus very important for calibrating the model to correspond to observed streamflow.
139
However, because these parameters are uniform for the entire watershed, they cannot be adjusted to
account for spatial variation among land‐use or soil classes within watersheds.
5.3. Methods
5.3.1 Data preparation: All preparation of spatial data was performed using ArcGIS 9.2. A regional DEM
(10 m pixels) was obtained from the National Map Seamless Server (USGS, 2008). SSURGO digital vector
soil data for Macon and Jackson counties were obtained from the National Resources Conservation
Service (USDA‐NRCS, 2005; USDA‐NRCS, 2007) and converted to raster datasets based on soil texture
class. Landcover data from 2006, with a classification scheme identical to the 2001 National Land Cover
Database (USGS, 2003), was obtained from the Coweeta Long Term Ecological Research Program (Fievet
and Collins, 2009). To preserve the information contained in the 10 m DEM, the coarser‐resolution soil
and land use raster layers were resampled to 10 m pixel size for modeling.
For the purposes of this study, classes were limited to evergreen forest, deciduous forest,
pasture, and developed land (Table 1). The NLCD classes were grouped to create these four classes.
NLCD classes of deciduous forest, mixed forest, and shrub were combined to create the deciduous forest
land‐use class used herein, and evergreen forest was retained as a separate class. Forest areas in these
study watersheds are characterized by predominantly deciduous forests, with occasional patches of
evergreen shrubs and trees occurring in steep stream valleys and at the highest elevations. The pasture
class was a composite of the grassland, pasture, and agriculture NLCD classes. Only very small areas of
row‐crop agriculture occur in the study area, and there are very few, small natural grasslands on
mountain peaks. The developed land use was created as a composite of all intensities of developed land
in the NLCD classification. Among these study watersheds, there are no substantial areas of high‐
intensity development. All developed land was estimated to have 30% impervious surface.
140
For model simulations of hydrologic response to varied land use scenarios, eight raster
landcover datasets were created for each watershed (Figure 3). These datasets were part of two
frameworks for analysis of streamflow differences among the land‐use scenarios. In the first
framework, watershed land use becomes progressively more developed, following development trends
in the region. In the second framework, all landcover scenarios contain 50% forest, 25% pasture, and
25% developed land, but in varied spatial arrangements.
Progressive Development Framework: In the first stage (or P1, for Progressive Development
Stage 1), the landcover is 100% forest, with evergreen forest limited to the highest elevations, mimicking
the present‐day actual land use. Stage 2 (P2) contains forest in all areas of the watershed with slope >
15%, and pasture in all areas with slope <15%. This 15% threshold was based on observed land‐use
trends in the region, and closely approximates the “classic” development patterns the region
experienced throughout the mid‐twentieth century. Stage P3 follows this 15% threshold as well, but in
this development stage the land use in areas lower than 15% slope is split evenly between developed
and pasture land, with developed land concentrated in the areas of lowest slope. Stage P4 was
intended to replicate the watershed land use that is expected with increasing development pressures.
In this scenario, watershed land use is evenly split between forest, pasture, and developed land use,
with forest cover in the 33% of watershed area with highest slopes, developed land use in the 33% of
watershed area with the lowest slopes, and pasture occupying the intermediate slope region. In stage
P5, representing extreme development, the 33% of the watershed area on the higher slopes is forest, as
in P4, but the entire remaining 67% percent of the watershed is in developed land use.
Spatial Distribution Framework: The objective of using this framework was to determine
whether streamflow differences are evident among different spatial arrangements of land use classes.
In each of these three scenarios, S1‐S3, watershed land use is 50% forest, 25% pasture, and 25%
developed, but the spatial distribution of the land use is varied. In Spatial Scenario 1 (S1), the
141
distribution is random, with landcover reclassified from the slope raster and randomly distributed
among 25 slope quantiles. In Scenario S2, the distribution is modeled after the classic regional
development pattern, in which the steeper slopes are under forest cover, and the lowest slopes are in
developed land use, with pasture occupying the areas of intermediate slopes. In Scenario S3, the classic
development pattern is inverted, with forest occupying the lower slope valley bottomlands, developed
land use on the intermediate slopes, and pasture on the higher slopes.
Daily climate data from four stations in Macon and Jackson counties were obtained for the
period of August 1, 1977 to December 31, 2008 (Figure 1). Daily mean temperature measurements and
precipitation totals were directly used as inputs to WetSpa, while daily maximum and minimum values
of temperature and relative humidity were used to estimate potential evapotranspiration (PET) using
the program ETo (Raes, 2009). This program uses available climate information, along with station
latitude, longitude, and elevation, to estimate PET values in the absence of comprehensive climatic
records that allow its direct calculation. Streamflow data for model calibration and evaluation were
obtained from multiple field‐based sources, for the period of August 1, 2007 to November 30, 2008. For
Iotla Creek and Cane Creek, 10‐minute instantaneous stage data were obtained using capacitance
probes installed in the stream edge. Stage data were converted to discharge values using rating curves
developed from ten stage/discharge points from discharge measurements using the velocity‐area
method with an acoustic Doppler velocity meter (Mosley and McKerchar, 1993). Instantaneous
discharge data for Cartoogechaye Creek were obtained from the U.S. Geological Survey (USGS gage
03500240), and for Shope Fork from the U.S. Forest Service (USFS) Coweeta Hydrologic Laboratory.
Daily mean discharge values were calculated for all streams and used for model calibration and
evaluation. Simulations were run spanning the period of August 1, 1978 to November 30, 2008, and the
first year of the simulations was excluded from analyses to allow for model spin‐up.
142
5.3.2 Model calibration and evaluation – Model calibration and validation were based on the 16‐month
period spanning August 1, 2007, to November 30, 2008, for most sites, but limited to 15 months for
Shope Fork. The first eight months of the period, August 1, 2007 to March 31, 2008, was used for
calibration, while the remainder of the record was used for model validation. For the purposes of model
calibration, the time series records of climate and streamflow data were duplicated to simulate a year of
model spin‐up for all sites except Cartoogechaye Creek, but only the second year was used for
calibration.
The ten global parameters, which affect all watershed pixels, irrespective of land use or soil
texture class, can be adjusted for calibration of WetSpa. Additionally, the look‐up tables associated with
secondary raster creation, may be modified to represent conditions specific to the study area. For
calibration of the four watersheds in this study, two look‐up tables were modified, 1) the “soil_remap”
table, which links the soil hydrologic characteristics to soil texture class, and 2) the “runoff_coefficient”
table, which combines information on slope, land‐use class, and soil texture class to determine the
portion of water that can run off as overland flow during storm events (as opposed to infiltrate to
become interflow or percolate to groundwater). Empirical data on the average soil porosity and
hydraulic conductivity for each texture class were available from the study area, and were directly used
in the “soil_remap” look‐up table. Empirical data from the study area have shown that within a given
soil texture class, there are pronounced differences in the hydraulic conductivities of forest and
nonforest soils (Price et al., in review). The observed magnitudes of difference between the hydraulic
conductivities of forest and nonforest soils were used to scale the runoff coefficients, assigned by
combinations of land‐use and soil texture classes, to reflect regional conditions. Several of the global
parameters (evapotranspiration correction factor, interflow scaling factor, recession constant scaling
factor, and surface runoff exponent) were also manually varied for model calibration. The Nash‐Sutcliffe
efficiency coefficient (NSE), expressing model predictive power, was calculated using the WHAT program
143
(Nash and Sutcliffe, 1970; Lim et al., 2005). Along with visual comparison of simulated and observed
stream hydrographs for the calibration period, these scores were used to select the optimal set of global
parameters and for model validation.
5.3.3 Calculation of streamflow metrics – Three baseflow metrics were calculated for each scenario, the
one percentile flow (Q99), seven day low‐flow (Qmin‐7), and the ratio of baseflow to total streamflow
(baseflow index, or BFI). Q99 was calculated as the one percentile flow occurring during the entire
period of record. Qmin‐7 was calculated from the moving seven‐day average throughout the 29‐year
period of record. The minimum 7‐day average from the entire period of record was used to represent
each land‐use scenario, and the minimum 7‐day average for each year was used for graphical analysis of
low flow frequency. BFI was calculated as total baseflow volume divided by total streamflow volume for
the 29‐year period, with baseflow volume determined by baseflow separation using the Eckhardt
recursive digital filter method (Eckhardt, 2008). Baseflow separation was performed using the online
WHAT hydrograph analysis tool (Lim et al., 2005).
While the emphasis of this study was on baseflow, additional streamflow variables were
calculated to provide context for interpretation of baseflow response to varied land‐use scenarios. The
mean daily streamflow for the 29‐year record (Qmean) was calculated for each scenario. The peak daily
flow was calculated for each year of record, with the year designated as August 1 to July 31, based on
the study start date. The median value of the 29 peak annual daily flows (Qpeak) was used to represent
each land‐use scenario. For graphical analysis of streamflow response to the varied land‐use scenarios,
flow duration curves were calculated from the continuous daily record. Additionally, recurrence plots
were calculated for the 7‐day annual low flows, using the Gringorten plot position (Stedinger et al.,
1993).
144
5.4. Results
WetSpa produced fair estimates of streamflow for the study watersheds (Figure 4). Nash‐
Sutcliffe coefficients ranged from 0.29 to 0.46 (perfect model prediction = 1.0). Development scenarios
of increasing intensity (P1‐P5) resulted in overall higher flows, including median annual peak flow, mean
flow, and extreme low flows. The proportion of baseflow to total streamflow (baseflow index, or BFI)
decreased with intensifying development. Among the three spatial scenarios evaluated for streamflow
differences under varied landcover distributions, there was not consistent response among the study
watersheds.
5.4.1 Progressive development framework – The flow duration curves (Figure 5) of the varied
development stages show two clear trends among all four watersheds: 1) Stages P1‐P3 produce almost
indistinguishable flow regimes, and 2) The substantial impervious surface that is introduced in stages P4
and P5 results in increased flood peaks, but not reduced baseflows. These counterintuitive trends are
also shown by the recurrence intervals of 7‐day annual low flows that increase moving from P1‐P5
(Figure 6), although these plots do show separation of stages P1‐P3 during very high recurrence interval
(infrequent) low‐flow events. During extreme dry periods represented by recurrence intervals greater
than 10 years, baseflows are lowest in watersheds where forest covers are highest, due to the
assumptions of evapotranspirative loss in forest land use built into the model structure. These results
contradict empirical data from the region, in which watersheds with higher forest cover have been
shown to have greater baseflow volume, during drought and non‐drought periods (Price and Jackson,
2007; Chapter 4 herein). All modeled flow magnitudes (Qpeak, Qmean, Q99, and Qmin‐7) increased as
watershed development increased and forest cover decreased (Table 2). This ubiquitous increase is
explained by the pronounced ET ascribed to forest landcover. Differences in flow among the
development stages were substantial. Values of Qpeak associated with the most pronounced
145
development (P5) were 1.7 to 3.6 times greater than values simulated for the 100% forest watersheds
(P1), Qmean values were 1.3 to 1.9 times greater, and Qmin‐7 values were 1.6‐3.1 times greater (Figure 7).
While baseflow magnitudes were shown to increase with development, this increase was not as great as
increases in peak flows, resulting in decreased proportions of baseflow to total flow observed in the
more developed watersheds. Therefore, BFI ratios were 1.2 to 1.4 times greater in the 100% forest
watersheds (P1) than the maximum development scenario (P5).
5.4.2 Spatial distribution framework – No trends emerged between the spatially varied land‐use
scenarios and simulated streamflow characteristics that were consistent among all four watersheds.
Land‐use scenario S2 has the highest peak flows and lowest BFI ratios in three of the four watersheds
(Figure 8). This scenario, in which the flatter portions of the watershed are in pasture and developed
land use, and the steeper portions are forested, represents the prevalent present‐day patterns and likely
continued development trends. The flow duration curves (Figure 5), and especially the 7‐day annual low
flow recurrence intervals (Figure 6), show that low flows under scenario S3 are greatly reduced from all
other land‐use scenarios in Iotla Creek and Cartoogechaye Creek. In Iotla Creek, the average Qmin‐7 of S1
and S2 is nearly twice as high as that of S3. In Cartoogechaye Creek, the average of S1 and S2 is more
than three times greater than the Qmin‐7 of S3. Scenario S3 represents the inverse of prevalent
development trends, with the flat bottomlands are under forest and steeper uplands in developed land
use.
5.4. Discussion
WetSpa was only moderately successful in simulating streamflow among these four watersheds.
It proved exceedingly difficult to optimize the model parameters to fit both low and high flows. Other
studies using WetSpa have had much greater success, demonstrating NSE values twice as high as those
146
observed in this study. Studies published by the model developers present results with NSE values
ranging from 0.72 to 0.84 among watersheds in Vietnam, Luxembourg, and Slovakia (Liu et al., 2005; Liu
et al., 2006; Bahremand et al., 2007). It should be noted that in this study, the calibration and validation
period of August 2007 to November 2008 coincided with an extreme regional drought, with the lowest
streamflows on record for all USGS gaging stations within the Little Tennessee River system. The
anomalous hydrologic conditions likely confounded the model calibration. These conditions may have
reduced the overall predictive capabilities of the model. However, identical trends emerged among the
land‐use scenarios using the 30‐year or 16‐month simulated records. Adding additional complication to
model calibration, this region has high relief, with pronounced variability of rainfall and subsurface
storage distribution. The lowest NSE values (0.29 for Cane Creek and 0.31 for Iotla Creek) were derived
from streamflow data with relatively high uncertainty, given the method of discharge calculation. The
NSE values from the long‐term USFS and USGS gaging stations were higher (0.41 for Cartoogechaye
Creek and 0.46 for Shope Fork). While the WetSpa NSE values from this study were substantially lower
than those observed by the developers, they are comparable to other hydrologic modeling studies,
which have demonstrated similar NSE values (0.3‐0.4) for other commonly used models such as RHESSys
(Tague, 2009), SWAT (Abu El‐Nasr et al., 2005; Dietrich and Funke, 2009), and BASINS‐HSPF (Abdulla,
2009).
5.5.1 Progressive development framework ‐ A surprising outcome of these simulated flows was that
decreasing watershed forest cover universally resulted in increase baseflow magnitude. Empirical data
from 35 watersheds in this study area, including the four study watersheds simulated herein, has shown
significantly greater baseflow among watersheds with more forest cover than those with less forest
cover (Chapter 4, this dissertation). The empirically documented positive relationship between
watershed forest cover and baseflow was attributed to greater soil infiltration and water holding
147
capacity associated with forest land use. These soil traits allow a greater portion of water to infiltrate to
subsurface storage and sustain baseflows, compared with the more compacted and less porous soils
associated with nonforest land use. This interpretation was reinforced by soil physical properties
measured in the study area, which showed saturated hydraulic conductivities approximately seven
times greater in forest soils, compared with lawn and pasture soils (Price et al., In Review). Wide,
statistically significant contrasts in conductivities, bulk densities, and water‐holding capacities were
demonstrated among soils of the same textural class, under forest vs. nonforest land use. The results of
the WetSpa simulations indicating higher BFI among more forested watersheds agrees with the findings
of these empirical studies, but the reduced baseflow magnitudes (Q99, Qmin‐7) contradict ample empirical
data, and are thus highly suspect.
It appears that the limitations on user input of empirical data to parameterize WetSpa, along
with highly problematic assumptions embedded into the model structure, preclude the model from
linking higher baseflow with lower watershed forest cover. The user can only modify soil characteristics
by soil texture class, so any known spatial variability among soils associated with varied land use or
other factors cannot be incorporated into the model function. It is our contention that the results of the
progressive development framework part of this study, and possibly those of the spatial distribution
framework as well, would have been completely different if we had the capability to incorporate a
greater amount of empirical data related to soil hydrology. Furthermore, the model appears to assume
that the evapotranspirative losses from forest cover are sufficient to over‐ride any gains to subsurface
storage that occur due to greater soil infiltration capacity. It is not necessarily a valid assumption that ET
rates always are greater in forests. While interception and transpiration in forests are certainly
important factors within a water budget, developed areas are associated with surface ponding of water
and pronounced direct evaporation from impervious surfaces and unshaded soils. Recent research has
shown developed areas to demonstrate ET rates greater than forests of all but the most water‐intensive
148
species (Batelaan and DeSmedt, 2007). The model’s theoretical basis appears to be the vast body of
twentieth century forestry experimentation literature demonstrating streamflow increases with forest
removal. Recent literature and research into scaling issues in hydrology have shown that upscaling from
small forestry plots to large, heterogeneous watersheds is extremely problematic (Blöschl, 2001).
Furthermore, it has been demonstrated that the impacts of temporary forest harvest are not
comparable to the hydrologic effects of long‐term land‐use conversion (Bruijnzeel, 2004).
The model’s embedded assumptions about forest ET explains the trends seen in the flow
duration curves (Figure 5), in which stages P1‐P3 are virtually indistinguishable. In the progressive
development framework, it is not until P4 that substantial amounts of watershed area are covered in
impervious surface (33%). Even though attempts were made, by modifying the runoff coefficient, to
inform the model that pasture in this study area demonstrates far lower infiltration rates than forest,
the high ET rates assigned to forest cover preclude any substantial differences in the water budgets of
forest vs. pasture land cover. Several reality‐based manipulations of the land use/soil based runoff
coefficient were tried, and under no tested scenario did baseflows increase with watershed forest cover,
as was observed in the empirical studies. Admittedly, WetSpa was originally developed for flood
prediction, and there are many studies demonstrating success with predicting high flows and flood
response to land‐use change using WetSpa (e.g., Liu et al., 2005; Liu et al., 2006; Bahremand et al.,
2007). However, the results of this study should serve as a resounding caution regarding the ever‐
increasing appeal of using distributed watershed modeling for purposes of watershed planning and
management.
5.5.2 Spatial distribution framework – In this framework, all scenarios consisted of watershed land use
with 50% forest, 25% pasture, and 25% developed land, but the spatial arrangement of the land use
differed. No relationship between spatial land‐use arrangement and streamflow patterns emerged that
149
was consistent among all four watersheds. However, there were some intriguing patterns that may
suggest fruitful areas of future investigation. The spatial land‐use scenario that replicates current land
use and development trends, S2, demonstrated the highest peak flows and lowest BFI in three of the
four watersheds (Figure 8). This finding may have important implications for issues of flooding and
drought tolerance among watersheds in the region. It is possible that planning efforts to vary the
developed‐bottomland, protected‐upland pattern may be advisable for reducing floods and protecting
baseflow, as development continues. However, these results also suggest that the reverse spatial
distribution, S3, in which bottomlands are forested and uplands developed, may also be problematic. In
Iotla Creek and Cartoogechaye Creek, the S3 scenario showed pronounced reductions in the 7‐day
annual low‐flow (Figures 6 and 8). For these watersheds, the Qmin‐7 was 2‐3 times higher under S1 and
S2. These watersheds have much more alluvial bottomland than Cane Creek or Shope Fork, introducing
the likely possibility that watershed geomorphology is affecting the streamflow response to land‐use
change. While the results from this analysis are insufficient to draw any conclusions, there is a
suggestion from these simulations that the spatial arrangement of land use, especially combined with
watershed geomorphic characteristics, may be an important factor in explaining extreme low and high
streamflows.
The suspicious and counterintuitive results of the progressive development framework of this
simulation study, which contrast a robust body of empirical data, underscore the caution that needs to
be taken when using distributed hydrologic models. The temptation is strong to put increasing faith in
models that have ever‐higher grid resolution, ever‐easier user operation, and ever‐faster performance
as computing capabilities increase. However the underlying, fundamental issues of parameter
designation and flow routing processes do not change, even with a more advanced interface (Beven,
2001). Most recently‐developed distributed models, including WetSpa, rely on decades‐old theory and
150
frighteningly small amounts of empirical data, and do not “properly reflect the collective intelligence of
the hydrological community” (Beven, 2001, p. 10). There are many well‐established problems with
distributed modeling, of which users outside of the theoretical modeling community may never be
informed. A major problem with complex distributed models is overparameterization (Jakeman and
Hornberger, 1993). It has been argued that simpler, less data‐intensive models provide equal or
superior results to more physically‐based models (Loague and Freeze, 1985), and that physically‐based
models should contain no more than 3‐5 parameters (Beven, 1989). WetSpa has ten global parameters,
and if all of the modifiable values in the look‐up tables are considered individual parameters, the
number is closer to 100. Even setting the issue of overparameterization aside, there is clear difficulty
with assigning realistic values to the parameters, as seen in this study. While admitting the necessity of
hydrologic models, Beven (2001) identifies nonlinearity in the hydrologic system, theoretically
unresolved process scaling issues, equifinality due to overparameterization, and uncertainty as
remaining key problems with distributed hydrologic models. The results of this study, in which model
structures consistently generated outcomes diametrically opposed to the well established empirical
behavior of these watersheds, underscore the critical need for caution among users of distributed
models. This is particularly germane, as increasing numbers of user‐friendly models are readily available
online, requiring little to no prior knowledge or education about hydrology.
5.6. Conclusions
Daily streamflow from four watersheds (7.8 to 145.5 km2) was simulated using the WetSpa fully‐
distributed hydrologic model over a period of 30 years. Validation of WetSpa indicated moderate
success in reproducing streamflows for these watersheds. However, streamflow conditions were
anomalously low during the study period, due to extreme regional drought, which likely confounded the
151
calibration. Streamflow was simulated for each stream under eight land‐use scenarios. Five of these
scenarios represented a progression from fully forested conditions to extreme development. The other
three contained constant areal coverages of forest, pasture, and developed land, but the spatial
arrangement of the land‐use classes was varied. Varied spatial arrangements of landcover did not
demonstrate any consistent relationships to streamflow among all four watersheds. Results from the
progressive development simulations indicated increases in peak, mean, and low flows as forest cover
decreased and developed landcover increased, while development was associated with a lower
proportion of baseflow to total streamflow (BFI). The simulated increases in baseflow magnitude with
watershed development contradict regional empirical data showing higher baseflows in watersheds
with greater forest cover, whereas the simulated results that BFI decreases with development
corroborate empirical results. Theoretical assumptions built into the model framework apparently
preclude a more realistic model outcome of decreased baseflow with increased development, because
evapotranspirative losses from forest land use are assumed to be sufficiently high to override any
subsurface storage gains from higher surface infiltration. This contrast with empirical data, due to
embedded theory within the model, should serve as a strong caution regarding the application of
distributed hydrologic modeling without an adequate understanding of the assumptions built into the
structure of such models.
5.7 References Abdulla, F., Eshtawi, T., and Assaf, H. 2009. Assessment of the impact of potential climate change on the water balance of a semi‐arid watershed. Water Resources Management 23(10): 2051‐2068. Abu El‐Nasr, A., Arnold, J. G., Feyen, J., and Berlamot, J. 2005. Modelling the hydrology of a catchment using a distributed and a semi‐distributed model. Hydrological Processes 19(3): 573‐587. Bahremand, A., De Smedt, F., Corluy, J., Liu, Y.B., Poorova, J., Velcicka, L., Kunikova, E. 2006. Application of WetSpa model for assessing land use impacts on floods in the Margecany‐Hornad watershed, Slovakia. Water Science and Technology 53(10): 37‐45.
152
Bahremand, A., De Smedt, F., Corluy, J., and Liu, Y. B. 2007. WetSpa Model application for assessing reforestation impacts on floods in Margecany‐Horland Watershed, Slovakia. Water Resources Management 21: 1373‐1391. Batelaan, O. and De Smedt, F. 2007. GIS‐based recharge estimation by coupling surface‐subsurface water balances. Journal of Hydrology 337: 337‐355. Beven, K. 1989. Changing ideas in hydrology: The case of physically‐based models. Journal of Hydrology 105: 157‐172. Beven, K. 2001. How far can we go with distributed hydrologic modeling? Hydrology and Earth System Sciences 5(1): 1‐12. Blöschl, G. 2001. Scaling in hydrology. Hydrological Processes 15: 709‐711. Bogena, H., Kunkel, R., Scobel, T., Schrey, H. P., Wendland F. 2005. Distributed modeling of groundwater recharge at the macroscale. Ecological Modeling 187(1): 15‐26. Bruijnzeel, L. A. 2004. Hydrological functions of tropical forests: not seeing the soil for the trees? Agriculture, Ecosystems, and Environment 104(1): 185‐228. Cao, W., Bowden, W. B., Davie, T., and Fenemor, A. 2006. Multi‐variable and multi‐site calibration and validation of SWAT in a large mountainous catchment with high spatial variability. Hydrological Processes 20: 1057‐1073. Chang, H., 2007. Comparative streamflow characteristics in urbanizing basins in the Portland Metropolitan Area, Oregon, USA. Hydrological Processes, 21: 211‐222. Dietrich, J. and Funke, M. 2009. Integrated catchment modeling within a strategic planning and decision making process: Werra case study. Physics and Chemistry of the Earth 34: 580‐588. Eckhardt, K. 2008. A comparison of baseflow indices, which were calculated with seven different baseflow separation methods. Journal of Hydrology 352: 168‐173. Fievet, C. and Collins, B. 2009. Land Cover Across Space and Time: The Southern Appalachians Since 1986. Accessed February 10, 2009. <http://coweeta.ecology.uga.edu/ecology/gis/landcover.html> Jakeman, A. J. and Hornberger, G. M. 1993. How much complexity is warranted in a rainfall‐runoff model? Water Resources Research 29(8): 2637‐2649. Johnson, R. 1998. The forest cycle and low river flows: a review of UK and international studies. Forest Ecology and Management 109: 1‐7. Lim, K. J., Engel, B .A., Zhenxu, T., Choi, J., Kim, K.‐S., Muthukrishnan, S. and Tripathy, D. 2005. Automated web GIS based hydrograph analysis tool, WHAT, Journal of the American Water Resources 41(6), 1407‐1416.
153
Liu, Y. B. and De Smedt, F. 2004. WetSpa extension: Documentation and user manual. Department of Hydrology and Hydraulic Engineering, Vrije Universiteit Brussel, Belgium. Liu, Y.B. and De Smedt, F. 2005. Flood modeling for complex terrain using GIS and remote sensed information. Water Resources Management 19(5): 605‐624. Liu, Y.B., Gebremeskel, S., De Smedt, F., Hoffmann, L., and Pfister, L. 2004. Simulation of flood reduction by natural river rehabilitation using a distributed hydrologic model. Hydrology and Earth System Sciences 8(6): 1129‐1140. Liu, Y. B., Batelaan, O., De Smedt, F., Huong, N. T., and Tam, V. T. 2005. Test of a distributed modeling approach to predict flood flows in the karst Suoimuoi catchment in Vietnam. Environmental Geology 48: 931‐940. Liu, Y. B., Gebremeskel, S., De Smedt, F., Hoffman, L., and Pfister, L. 2006. Predicting storm runoff from different land‐use classes using a geographical information system‐based distributed model. Hydrological Processes 20: 533‐548. Loague , K. M. and Freeze, R. A. 1985. A comparison of rainfall‐runoff modelling techniques on small upland catchments. Water Resources Research 21: 229‐248. Mosley, M. P. and McKerchar, A. I. 1993. Streamflow. In Maidment, D. R. (ed.), Handbook of Hydrology. McGraw Hill, New York, p. 8.1‐8.39. Mulungu, D. M. M., Ichikawa, Y., and Shiiba, M. 2005. A physically‐based distributed subsurface‐surface flow dynamics model for forested mountainous catchments. Hydrological Processes 19: 3999‐4022. Nash, J. E. and Sutcliffe, J. V. 1970. River flow forecasting through conceptual models, Part 1: A discussion of principles. Journal of Hydrology 10: 282‐290. Nyenje, P. M. and Batelaan, O. 2009. Estimating the effects of climate change on groundwater recharge and baseflow in the upper Ssezibwa catchment, Uganda. Hydrological Sciences Journal 54(4): 713‐726. Price, K. and Jackson, C.R., 2007. Effects of forest conversion on baseflows in the southern Appalachians: A cross‐landscape comparison of synoptic measurements. Proceedings of the 2007 Georgia Water Resources Conference. <http://cms.ce.gatech.edu/gwri/uploads/proceedings/2007/2.3.4.pdf> Raes, D. 2009. The ETo Calculator Reference Manual version 3.1. Food and Agriculture Organization of the United Nations, Land and Water Division. Rome, Italy. Accessed June 15, 2009. < http://www.biw.kuleuven.be/lbh/lsw/iupware/> Rose, S. and Peters, N. E. 2001. Effects of urbanization on streamflow in the Atlanta area (Georgia, USA): A comparative hydrological approach. Hydrologic Processes 15: 1441‐1457. Stedinger, J. R., Vogel, R. M., and Foufoula‐Georgiou, 1993. Frequency analysis of extreme events. In Maidment, D. R. (ed.), Handbook of Hydrology. McGraw Hill, New York, p. 18.1‐18.66.
154
Tague, C. 2009. Modeling hydrologic controls on denitrification: sensitivity to parameter uncertainty and landscape representation. Biogeochemistry 93: 79‐90. USDA‐NRCS, 2005. Soil Survey Geographic (SSURGO) Database for Macon County, North Carolina. Accessed September 10, 2007. <http://soildatamart.nrcs.usda.gov> USDA‐NRCS, 2007. Soil Survey Geographic (SSURGO) Database for Jackson County, North Carolina. Accessed September 10, 2007. <http://soildatamart.nrcs.usda.gov> USGS, 2003. North Carolina Land Cover Database Zone 57 Land Cover Layer, Accessed September 20, 2007. <www.seamless.usgs.gov> USGS, 2008. National Elevation Dataset (NED) 1/3 Arc Second, Accessed June 15, 2009. <www.seamless.usgs.gov>
155
WetSpa Class
Original NLCD Class
Evergreen Forest 5.8% 4.5% 1.8% 0.3%42 ‐ Evergreen Forest 5.8% 4.5% 1.8% 0.3%
Deciduous Forest 92.6% 73.1% 84.2% 94.7%41 ‐ Deciduous Forest 91.1% 67.2% 81.3% 93.2%
43 ‐ Mixed Forest 0.7% 3.4% 1.4% 0.5%52 ‐ Shrub/Scrub 0.8% 2.3% 1.5% 1.0%
91 ‐ Woody Wetland 0.0% 0.2% 0.1% 0.0%
Pasture 0.0% 13.3% 5.9% 1.0%71 ‐ Grassland/Herbaceous 0.0% 1.8% 1.0% 0.5%
81 ‐ Pasture/Hay 0.0% 10.5% 4.6% 0.5%82 ‐ Cultivated Crops 0.0% 1.0% 0.3% 0.0%
Developed 1.6% 9.0% 8.1% 4.0%21 ‐ Developed Open 1.6% 8.4% 7.2% 4.0%
22 ‐ Developed Low‐Intensity 0.0% 0.3% 0.6% 0.0%23 ‐ Developed Medium‐Intensity 0.0% 0.1% 0.2% 0.0%
24 ‐ Developed High‐Intensity 0.0% 0.1% 0.0% 0.0%31 ‐ Barren Land 0.0% 0.1% 0.1% 0.1%
Open Water 0.0% 0.1% 0.1% 0.0%11 ‐ Open Water 0.0% 0.1% 0.1% 0.0%
Shope Fork Iotla CreekCartoogechaye
CreekCane Creek
Table 5.1. 2006 land use of study watersheds and explanation of reclassification of NLCD land use scheme. NLCD class numbers correspond to 2001 classification (USGS, 2003). (Source for 2006 land use data: Fievet and Collins, 2009)
156
Stream Land Use Qpeak Q99 Qmean Qmin‐7 BFIm3s ‐1 m3s ‐1 m3s ‐1 m3s‐1 ‐
Shope Fork S1 3.097 0.050 0.274 0.033 0.703S2 3.799 0.060 0.301 0.043 0.661S3 2.870 0.051 0.273 0.033 0.716P1 1.953 0.037 0.246 0.019 0.759P2 2.050 0.037 0.246 0.020 0.754P3 2.118 0.038 0.248 0.021 0.750P4 3.610 0.053 0.287 0.035 0.665P5 4.001 0.066 0.316 0.048 0.639
Iotla Cr. S1 4.699 0.065 0.445 0.036 0.709S2 5.237 0.061 0.444 0.033 0.697S3 4.959 0.047 0.457 0.018 0.626P1 2.173 0.050 0.376 0.020 0.878P2 3.335 0.050 0.382 0.022 0.802P3 4.162 0.058 0.414 0.029 0.742P4 6.127 0.068 0.476 0.042 0.659P5 7.900 0.081 0.564 0.048 0.574
Cartoogechaye Cr. S1 24.747 0.355 3.258 0.214 0.730S2 23.711 0.342 3.248 0.210 0.658S3 28.034 0.241 3.298 0.065 0.720P1 17.631 0.229 2.892 0.111 0.772P2 18.440 0.234 2.923 0.120 0.765P3 18.503 0.240 2.939 0.125 0.763P4 26.644 0.388 3.433 0.233 0.702P5 32.772 0.543 4.107 0.349 0.648
Cane Cr. S1 1.819 0.023 0.119 0.013 0.674S2 1.920 0.023 0.121 0.013 0.658S3 1.729 0.023 0.117 0.013 0.764P1 0.689 0.012 0.088 0.008 0.751P2 0.719 0.012 0.088 0.008 0.748P3 0.756 0.013 0.090 0.009 0.743P4 1.725 0.020 0.132 0.012 0.606P5 2.306 0.020 0.167 0.013 0.536
Table 5.2. Flow summary values for varied land use scenarios. Land use scenario symbols are explained in Section 5.3.1.
157
Figure 5.1. Study area and watersheds used for streamflow simulation. Stream names correspond to the numbered watersheds on the map, with watershed area in parentheses.
158
Figure 5.2 Summary of WetSpa spatial parameterization. Bold parameters indicate parameters that were modified based on empirical data from this study area. For all others parameters, model default values were used.
159
Figure 5.3 Simulated land use scenarios. The Iotla Creek watershed is used as an example. Progressive development stages 1‐5 (P1‐P5) represent intensifying development in a spatial pattern characteristic of the region, while the spatial distributions of S1‐S3 have equivalent land cover by areal extent, but vary in the arrangement of the land use pixels. Stage P1 is 100% forest. A complete explanation is provided in section 5.3.1.
P2
P3
P4
P5
S1
S2
S3
0 2.5 51.25 km
$Legend
Forest
Pasture
Developed
160
Figure 5.4a Comparison of simulated vs. observed streamflow. NRE is Nash‐Sutcliffe efficiency coefficient (Nash and Sutcliffe, 1970). The linear scale on the y‐axis (5.4a) accentuates the low absolute error in low flows, while the log scale on the y‐axis emphasizes the high relative error.
Iotla Creek
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
11/1/08
Discharge (m
3 s‐1)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Precipitation (mm)
0
50
100
150
200
250
Modeled streamflowObserved streamflowPrecipitation
Shope Fork
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
Discharge (m
3 s‐1)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Precipitation (mm)
0
50
100
150
200
250
Cartoogechaye Creek
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
11/1/08
Discharge (m
3 s‐1)
0
5
10
15
20
Precipitation (mm)
0
50
100
150
200
250
Cane Creek
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
11/1/08
Discharge (m
3 s‐1)
0.0
0.2
0.4
0.6
0.8
Precipitation (mm)
0
50
100
150
200
250
Calibration
Validation
Calibration
Validation
Calibration
Validation
Calibration
Validation
NSE = 0.46 NSE = 0.31
NSE = 0.41 NSE = 0.29
161
Figure 5.4b Comparison of simulated vs. observed streamflow. NRE is Nash‐Sutcliffe efficiency coefficient (Nash and Sutcliffe, 1970). The linear scale on the y‐axis (5.4a) accentuates the low absolute error in low flows, while the log scale on the y‐axis emphasizes the high relative error.
Iotla Creek
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
11/1/08
Discharge (m
3 s‐1)
0.1
1
10
Precipitation (mm)
0
50
100
150
200
250
Modeled streamflowObserved streamflowPrecipitation
Shope Fork
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
Discharge (m
3 s‐1)
0.1
1
10
Precipitation (mm)
0
50
100
150
200
250
Cartoogechaye Creek
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
11/1/08
Discharge (m
3 s‐1)
1
10
100
Precipitation (mm)
0
50
100
150
200
250
Cane Creek
8/1/07
9/1/07
10/1/07
11/1/07
12/1/07
1/1/08
2/1/08
3/1/08
4/1/08
5/1/08
6/1/08
7/1/08
8/1/08
9/1/08
10/1/08
11/1/08
Discharge (m
3 s‐1)
0.1
1
Precipitation (mm)
0
50
100
150
200
250
Calibration
Validation
Calibration
Validation
Calibration
Validation
Calibration
Validation
NSE = 0.46 NSE = 0.31
NSE = 0.41 NSE = 0.29
162
Figure 5.5 (page 1 of 2). Flow duration curves for the varied land use scenarios. Symbols correspond to Progressive (P) and Spatial (S) scenarios described in section 5.3.1. These graphs convey the percent of time a given flow level is exceeded. For example, very high flows are exceed for only a small percentage of the time, while extremely low flows are exceeded nearly 100% of the time. The overlain curves represent the varied land use simulations, for which the distributions of varied flow levels can be compared.
Fraction of Time Exceeded
0.01
0.05 0.1 0.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 99
.899.999.95
99.99
Discharge
(m3 s
‐1)
0.01
0.1
1
10
P1P2P3P4P5 S1S2S3
Fraction of Time Exceeded
0.01
0.05 0.1 0.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 99
.899.999.95
99.99
Discharge
(m3 s
‐1)
0.01
0.1
1
10
Shope Fork
Iotla Creek
P1P2P3P4P5 S1S2S3
163
Figure 5.5 (page 2 of 2). Flow duration curves for the varied land use scenarios. Symbols correspond to Progressive (P) and Spatial (S) scenarios described in section 5.3.1. These graphs convey the percent of time a given flow level is exceeded. For example, very high flows are exceed for only a small percentage of the time, while extremely low flows are exceeded nearly 100% of the time. The overlain curves represent the varied land use simulations, for which the distributions of varied flow levels can be compared.
Fraction of Time Exceeded
0.01
0.05 0.1 0.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 99
.899.999.95
99.99
Discharge
(m3 s
‐1)
0.01
0.1
1
10
Fraction of Time Exceeded
0.01
0.05 0.1 0.2 0.5 1 2 5 10 20 30 50 70 80 90 95 98 99 99
.899.999.95
99.99
Discharge
(m3 s
‐1)
0.1
1
10
100
P1P2P3P4P5S1S2 S3
Cartoogechaye Creek
Cane Creek
P1
P2
P3
P4
P5
S1
S2
S3
164
Figure 5.6 (page 1 of 2). Recurrence intervals of 7‐day annual low flows , 1979‐2008. Recurrence intervals were calculated using the Gringorten plot position (ref*). Symbols correspond to Progressive (P) and Spatial (S) scenarios described in section 5.3.1.
Recurrence Interval (years)1 10 100
7‐day annu
al low flow
(m3 s
‐1)
0.1
P1P2P3 P4P5S1S2S3
Recurrence Interval (years)1 10 100
7‐day annu
al low flow
(m3 s
‐1)
0.01
0.1
Shope Fork
Iotla Creek
P1P2P3 P4P5S1S2S3
165
Figure 5.6 (page 2 of 2). Recurrence intervals of 7‐day annual low flows , 1979‐2008. Recurrence intervals were calculated using the Gringorten plot position (ref*). Symbols correspond to Progressive (P) and Spatial (S) scenarios described in section 5.3.1.
Recurrence Interval (years)1 10 100
7‐day annu
al low flow
(m3 s
‐1)
0.1
1
P1P2P3P4P5P6P7 P8
Recurrence Interval (years)1 10 100
7‐day annu
al low flow
(m3 s
‐1)
0.01
0.1
Cartoogechaye Creek
Cane Creek
P1P2P3P4P5P6P7 P8
166
Figure 5.7 Streamflow summary metrics for each of the progressive development land use scenarios. Symbols P1‐P5 are described in section 5.3.1.
Mean Flow
0.01
0.1
Median Peak Annual FlowW
ater
shed
are
a-st
anda
rdiz
ed d
isch
arge
(m3 s-1
km-2
)
0.01
0.1
1
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
7-day Low Flow
0.0001
0.001
0.01
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
Baseflow Index
0.0
0.2
0.4
0.6
0.8
1.0
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
Wat
ersh
ed a
rea-
stan
dard
ized
dis
char
ge (m
3 s-1km
-2)
P1P2P3 P4 P5
167
Figure 5.8 Streamflow summary metrics for each spatial land use scenario. Symbols S1‐S3 are described in section 5.3.1.
Mean Flow
0.01
0.1
Median Peak Annual FlowW
ater
shed
are
a-st
anda
rdiz
ed d
isch
arge
(m3 s-1
km-2
)
0.1
1
S1S2S3
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
7-day Low Flow
0.0001
0.001
0.01
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
Baseflow Index
0.0
0.2
0.4
0.6
0.8
1.0
Shope Fork
IotlaCreek
CartoogechayeCreek
CaneCreek
Wat
ersh
ed a
rea-
stan
dard
ized
dis
char
ge (m
3 s-1km
-2)
168
CHAPTER 6
SUMMARY AND CONCLUSIONS
Human conversion of forests to agricultural or residential land alters hydrologic processes
affecting baseflows in competing ways – by reducing evapotranspiration which should increase
baseflows, and by compacting soils which should reduce groundwater recharge and thus reduce
baseflows. This research attempted to determine which hydrologic alteration is dominant in southern
Appalachian watersheds and how such alterations compare to the importance of the natural watershed
geomorphic template. There is a large body of literature investigating the influence of land use on
baseflows, but the majority of these studies have emphasized two specific types of impact: 1) temporary
forest harvest, and2) urbanization. Very little is known about the effects of conversion of natural forest
to many land uses that are prevalent on the landscape, such as pasture, low‐, and medium‐intensity
development. Urbanization studies demonstrate no clear trend in baseflow response, because of
complications to water budgets due to issues such as importation of water, infrastructure leakage, etc.
Forestry studies have shown that forest removal increases baseflows, because the interception and
transpiration associated with forests reduce the amount of water available for streamflow. However,
there is little basis for extrapolating the results from small‐plot, forestry experimentation studies to
larger watersheds undergoing long‐term land‐use change. Such experiments are generally conducted in
very small watersheds, and the process of forest removal, in which little long‐term soil disturbance
occurs, do not show much similarity to the impacts of permanent land‐use conversion. In the southern
Blue Ridge Mountains, there are currently development pressures driven primarily by exurban and
vacation home markets, and the land‐use change occurring in this region primarily takes the form of
residential development and the medium‐density commercial growth to support increased populations.
The current understanding of watershed hydrology does not provide insight into what changes in low‐
169
flow hydrology to expect from continued land‐use change in developing regions like the southern Blue
Ridge.
To address this problem, three separate but complementary studies were conducted. In the
first study, the hydraulic characteristics of soils underlying forest, pasture and turfgrass (lawn) land use
were compared. Soils designated as the same series, and located in similar physical settings, were
compared under the varied land uses. The saturated hydraulic conductivity, bulk density, and water
holding capacity were measured for 30 sites in each land‐use class, giving a total of 90 samples. The
average saturated hydraulic conductivity of forest soils was seven times higher than pasture and lawn
soils. The water holding capacities of forest soils were significantly higher than lawn and pasture soils,
and the bulk densities of the forest soils was significantly lower than the nonforest land uses. For no
parameter did pasture and lawn soils significantly differ. These results carry important implications for
watershed hydrology, as the nonforest soils are much more highly susceptible to overland flow during
storm events. Greater overland flow is associated with reduced infiltration and baseflows, as well as
impaired water quality. These results also show that the flashier hydrologic regimes generally
associated with highly developed watersheds result not only from impervious surface increases, but also
from soil compaction and altered soil hydrology in non‐natural land uses.
In the second study, 35 streams of varied land use and geomorphology were instrumented with
continuous stage recorders for discharge measurement. Watersheds ranging from 45 to 100% forest
cover were monitored for 16 months, encompassing two low‐flow seasons (late summer and fall) during
pronounced drought conditions. Baseflow metrics of one percentile flow, 1‐day, 7‐day, and 14‐day
minimum flows, and baseflow index were calculated for three subsets of the study period: low‐flow
season 2007, low‐flow season 2008, and water year 2008. Multiple regression analysis indicated that
watershed geomorphic characteristics, such as drainage density, slope variability, and colluvium were
the most important variables in predicting the various baseflow metrics. Grouped watersheds with
170
higher forest cover demonstrated significantly greater baseflow than watersheds with lower forest
cover. These results contradict the relationship seen in experimental forestry studies, and show that the
reduced infiltration associated with the conversion of forest to nonforest land use over‐rides any water
gains from forest removal due to reduced interception and transpiration. Human impact in the
southern Blue Ridge is still relatively minor, with most watersheds containing well over 80% forest
cover, but baseflow differences among more‐ and less‐impacted watersheds are already apparent. The
results of this study indicate that continued development in this region (and its associated reductions in
forest cover and increase in impervious surface and compacted soils) will reduce baseflows and
potentially lead to impaired water quality, degraded aquatic habitat, and increased vulnerability to
future drought.
In the third study, the distributed, GIS‐based hydrologic model WetSpa was used to simulate 30
years of streamflow for four watersheds, under 8 land use scenarios. The primary objective was to use
the model to evaluate baseflow response to development levels more intense than were available for
empirical study. However, the model contained embedded theoretical assumptions about the
importance of evapotranspirative losses from forest land use, without adequately accounting for the far
greater hydraulic conductivities of forest soils in contributing to subsurface storage recharge and
baseflow. Consequently , model results for developed watersheds were associated with greater
baseflow magnitude, despite a great deal of empirical data that shows the opposite relationship. These
counterintuitive and empirically‐contradicted results should serve as a strong caution regarding
hydrologic modeling. WetSpa model structures are likely drawn from hydrologic theory related to the
forestry experimentation literature, and do not accommodate any new empirical information on the
importance of soil hydraulic characteristics on stream baseflows. As such distributed models become
more readily‐available and user‐friendly, increasingly watershed management and protection agencies
are coming to rely on such tools. It is of great importance that model results be treated with great
171
caution, and that users remain alert to the fact that these models often contain equations based on
highly generalized theory and minimal empirical information.
Overall, four principal findings emerged from this dissertation research:
1. Changes in soil physical properties associated with permanent forest removal are
substantial, approaching impervious surfaces in importance for understanding hydrologic
behavior of altered watersheds in the southern Blue Ridge.
2. Empirical findings for watersheds ranging from 3 to 146 km2 demonstrate a positive
relationship between watershed forest cover and stream discharge. This is counter to plot‐
level forest harvest studies, and underscores the danger of upscaling from localized studies
to predict regional hydrologic behavior.
3. The loss of forest cover and modification of soil physical properties with land clearance and
development is associated with reduced baseflows, especially during dry periods. Hence,
these landscape pressures magnify drought vulnerability and declines in water quality and
aquatic habitat.
4. Models that uncritically accept the plot‐level linkage of tree removal with reduced
evapotranspiration baseflow can generate results that deviate dramatically from empirically
observed results. This should serve as a powerful caution to potential users of models
regarding the importance (and possible inaccuracies) of underlying assumptions within
model frameworks.
172
APPENDIX A
DISCHARGE RATING CURVES
The discharge rating curves presented in the following appendix correspond to discharge values presented in Chapter 4. Data for sites unrepresented in this appendix (Nantahala River, Cartoogechaye Creek, and Cullasaja River) were obtained directly as discharge values from the U.S. Geological Survey and did not require rating curve development.
These rating curves were primarily developed from discharge measurements using the velocity‐area method in natural stream channels, with stream stage determined from continuous water level recorders instrumented on the stream cross‐section. The black points represent all stage‐discharge points using this method. Green points represent high stage flows modeled using Manning’s Equation. Manning’s n was calculated from the highest measured discharge values, with cross‐sectional dimensions and channel slope determined by laser‐level survey. In most cases, the green high flow point represents bankfull discharge, as determined by channel morphology. Exceptions include entrenched or modified channels where bankfull stage could not be determined from channel morphology. For these sites, an arbitrary high stage point was used. Blue stage‐discharge points indicate discharge measurements using dye tracers during high flows on large streams, when conditions were unsafe for wading.
The curve fit lines represent the equation used to convert stream stage values to discharge values. These curves were developed using Bayesian power‐law curve fitting. The R2 values for each graph correspond to the Bayesian fit. In some cases, a multi‐segment rating curve was more appropriate than a simple curve. Curve development was achieved using a program developed by Trond Reitan at the University of Oslo, Norway.
The rating curve for Shope Fork was used solely to test the accuracy of the velocity‐area discharge method paired with the rating curve development, by comparison with the U.S. Forest Service weir immediately downstream from the stage recording and discharge measurement site. The discharge values from the weir were used for the analyses presented in Chapter 4 and for model calibration in Chapter 5.
173
APPENDIX A: Discharge Rating Curves page 2 of 7
1. Buck Creek
Stage (m)0.4 0.6 0.8 1.0 1.2
Q (m
3 /s)
0.1
1
10
2. Roaring Fork
Stage (m)0.3 0.4 0.5 0.6 0.7 0.8
0.01
0.1
1
4. Wayah Creek
Stage (m)0.4 0.6 0.8 1.0
Q (m
3 /s)
0.1
1
10
5. Poplar Cove
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.1
1
10
6. Allison Creek
Stage (m)0.2 0.4 0.6 0.8 1.0
0.1
1
10
Q (m
3 /s)
7. Jones Creek
Stage (m)0.4 0.6 0.8 1.0
0.1
1
10
R2 = 0.955 R2 = 0.953
R2 = 0.964 R2 = 0.838
R2 = 0.953 R2 = 0.990
174
APPENDIX A: Discharge Rating Curves page 3 of 7
8. Shope Fork
Stage (m)0.3 0.4 0.5 0.6 0.7 0.8
0.1
1
10
14. North Fork Skeenah Creek
Stage (m)0.4 0.6 0.8 1.0 1.2 1.4
Q (m
3/s)
0.01
0.1
1
12. Blaine Branch
Stage (m)0.2 0.3 0.4 0.5 0.6
0.01
0.1
1
13. McDowell Branch
Stage (m)0.2 0.4 0.6 0.8
0.01
0.1
1
11. Crawford Branch
Stage (m)0.2 0.4 0.6 0.8 1.0 1.2
0.01
0.1
1
10. Iotla Creek
Stage (m)0.4 0.6 0.8 1.0 1.2
0.1
1
10
Q (m
3 /s)
Q (m
3/s)
R2 = 0.968 R2 = 0.981
R2 = 0.966 R2 = 0.889
R2 = 0.950 R2 = 0.994
175
APPENDIX A: Discharge Rating Curves page 4 of 7
17. Fulcher Branch
Stage (m)0.2 0.3 0.4 0.5
0.01
0.1
1
18.Cowee Creek
Stage (m)0.3 0.4 0.5 0.6 0.7 0.8
0.1
1
10
19. Caler Fork
Stage (m)0.2 0.4 0.6 0.8
0.01
0.1
1
16. Bates Branch
Stage (m)0.4 0.6 0.8 1.0
0.01
0.1
1
15. South Fork Skeenah Creek
Stage (m)0.3 0.4 0.5 0.6 0.7
0.1
1
10
20. Watauga. Creek
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7
0.1
1
10
Q (m
3/s)
Q (m
3/s)
Q (m
3/s)
R2 = 0.995 R2 = 0.973
R2 = 0.974 R2 = 0.950
R2 = 0.807 R2 = 0.997
176
APPENDIX A: Discharge Rating Curves page 5 of 7
21. Rabbit Creek
Stage (m)0.4 0.5 0.6 0.7 0.8 0.9 1.0
0.1
1
25. Little Ellijay Creek
Stage (m)0.3 0.4 0.5 0.6 0.7 0.8 0.9
0.1
1
24. Tathams Branch
Stage (m)0.30 0.35 0.40 0.45 0.50 0.55 0.60
0.01
0.1
1
22. Nickajack Creek
Stage (m)0.25 0.30 0.35 0.40 0.45 0.50 0.55
0.01
0.1
1
10
23. Savannah Creek
Stage (m)0.5 0.6 0.7 0.8 0.9
0.1
1
26. Little Savannah Creek
Stage (m)0.4 0.5 0.6 0.7 0.8 0.9 1.0
0.1
1
Q (m
3/s)
Q (m
3/s)
Q (m
3/s)
R2 = 0.972 R2 = 0.985
R2 = 0.992 R2 = 0.949
R2 = 0.945 R2 = 0.823
177
APPENDIX A: Discharge Rating Curves page 6 of 7
30. Cope Creek
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.01
0.1
1
10
31. Cane Creek
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7 0.8
0.1
1
29. Blanton Branch
Stage (m)0.3 0.4 0.5 0.6 0.7 0.8 0.9
0.01
0.1
1
27b. Cullowhee Creek (after 1.10.08)
Stage (m)0.4 0.5 0.6 0.7 0.8
0.1
1
27a. Cullowhee Creek (to 1.10.08)
Stage (m)0.4 0.5 0.6 0.7 0.8
0.1
1
28. Buff Creek
Stage (m)0.1 0.2 0.3 0.4 0.5 0.6 0.7
0.01
0.1
1
Q (m
3/s)
Q (m
3 /s)
Q (m
3/s)
R2 = 0.989 R2 = 0.968
R2 = 0.855 R2 = 0.911
R2 = 0.829 R2 = 0.999
178
APPENDIX A: Discharge Rating Curves page 7 of 7
34. Mud Creek
Stage (m)0.3 0.4 0.5 0.6 0.7 0.8
0.1
1
33. Darnell Creek
Stage (m)0.4 0.6 0.8 1.0
0.1
1
10
32. Wayehutta Creek
Stage (m)0.2 0.3 0.4 0.5 0.6 0.7
0.01
0.1
1
10
Q (m
3 /s)
Q (m
3/s)
R2 = 0.999 R2 = 0.963
R2 = 0.999
179
APPENDIX B
HYDROGRAPHS
This appendix contains streamflow hydrographs from all sites included in analyses presented in Chapter 4. For all sites except Nantahala River, Shope Fork, Cartoogechaye Creek, and Cullasaja River, the hydrographs were created by converting 10‐minute interval stage data to flow values using the rating curves presented in Appendix A. The hydrographs for Nantahala River, Cartoogechaye Creek, and Cullasaja River represent 15‐minute interval flow data obtained from the U.S. Geological Survey. The hydrograph for Shope Fork was developed from data obtained from the U.S. Forest Service Coweeta Hydrologic Laboratory.
180
2. Roaring Fork
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
1. Buck Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
Q (m
3 s-1)
0.01
0.1
1
10
Q (m
3 s-1)
APPENDIX B: Hydrographs page 2 of 19
181
3. Nantahala River
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.1
1
10
100
4. Wayah Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 3 of 19
182
5. Poplar Cove Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
6. Allison Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 4 of 19
183
7. Jones Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
8. Shope Fork
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 5 of 19
184
9. Cartoogechaye Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.1
1
10
100
10. Iotla Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 6 of 19
185
11. Crawford Branch
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
0.01
0.1
1
12. Blaine Branch
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 7 of 19
186
13. McDowell Branch
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
14. North Fork Skeenah Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.0001
0.001
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 8 of 19
187
15.South Fork Skeenah Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
16. Bates Branch
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
0.01
0.1
1
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 9 of 19
188
17. Fulcher Branch
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
18. Cowee Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 10 of 19
189
19. Caler Fork
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
20. Watauga Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 11 of 19
190
21. Rabbit Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
22. Nickajack Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 12 of 19
191
23. Savannah Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
24. Tathams Branch
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
0.01
0.1
1
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 13 of 19
192
25. Little Ellijay Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
26. Little Savannah Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 14 of 19
193
27. Cullowhee Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
0.1
1
10
28. Buff Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 15 of 19
194
29. Blanton Branch
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
30. Cope Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 16 of 19
195
31. Cane Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
32. Wayehutta Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 17 of 19
196
33. Darnell Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.01
0.1
1
10
34. Mud Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09 0.001
0.01
0.1
1
10
Q (m
3 s-1)
Q (m
3 s-1)
APPENDIX B: Hydrographs page 18 of 19
197
35. Cullasaja Creek
7/1/07 8/1/07 9/1/07 10/1/07 11/1/07 12/1/07 1/1/08 2/1/08 3/1/08 4/1/08 5/1/08 6/1/08 7/1/08 8/1/08 9/1/08 10/1/08 11/1/08 12/1/08 1/1/09 2/1/09
0.1
1
10
100Q
(m3 s-1
)
APPENDIX B: Hydrographs page 19 of 19
196
APPENDIX C
WETSPA PARAMETERIZATION AND FLOW ROUTING
This appendix presents the mathematical framework behind WetSpa parameterization and flow routing. These equations are taken from the WetSpa user manual (Liu and De Smedt, 2004), which provides a more thorough explanation of the theory behind these equations.
Section 1 presents the parameterization of the catchment water balance, which is used to partition precipitation and snowmelt into surface runoff, evapotranspiration, changes in soil storage, and changes in groundwater storage. Section 2 presents the equations that are used to route water from pixel to pixel, both as overland flow among hillslope pixels and channel flow. Section 3 contains the equations for determining flow response.
199
200
201
202
203
204