s Patio Temporal Mapping

8/12/2019 s Patio Temporal Mapping

1/11

OMMUNITYECOLOGY 1): 69-79,20041585-8553/ 20.00 Akadkmiai Kiadd , Budapest

Spatiotemporal mapping of the dry season vegetation responseof sagebrush steppeR. A. Washington-Allen1'2,R. D. Ramsey' and N. E. West'I Department of Forest, Range, and Wildlg e Sciences, Utah State U niversi@,Old Main Hill 5230, Logan, Utah,84321, USA. E-mail:[email protected],[email protected] [email protected] Ridge, 71v37831-6407, USAFormer address: Environmental Sciences Division, Oak Ridge National Laboratoly, MS 6407, Bldg. 1507,

Keywords: Landsat, Pastoralists, Rangelands, Remote sensing, Retrospective, Time series.Abstract: The vegetation dynamics of semi-arid and ar id landscapes are temporally and spatially heterogeneous and subject to variousdisturbance regimes that act on decadal scales. Traditional field-based monitoring methods have failed to sample adequately in timeand space in order to capture this heterogeneity and thus lack the spatial extent and the long-term continuous time series of datanecessary to detect anomalous dynamics in landscape behavior. Time series of ecological indicators of land degradation that arecollected synoptically fkom local to global spatial scales can be derived from the 33-year and continuing Landsat satellite archive.Consequently, a retrospective study was conducted on a commercially grazed sagebrush steppe dominated Utah landscape using a timeseries of standardized Landsat imagery for the period 1972 to 1997. The study had the objectives to (1) characterize and map thehistorical trends of a remotely-sensed index of vegetation response, a correlate of vegetation cover or phytomass, and (2) toretrospectively infer the cause of this response to historical records of grazing and wet and drought periods. A time series of dry seasonvegetation index maps were statistically clustered t generate a spatio-temporal map of three coarse trends o fvegetation response, Le.,declining, stable, and increasing trends. This study showed that 71% of the landscape's locations had an increasing trend and 29% hada stable trend over the 26-year period. The increasing trend locations were positively correlated with site water balance [the PalmerDrought Severity Index (PDSI)], i.e., vegetation response increased during wet periods and decreased during drought. The increasingtrend was positively and negatively (non-linearly) correlated with grazing in individual paddocks from 1980 to 1997.Nomenclature: Shaw (1989).Abbreviations: AU-Animal unit; AWR- '-Grazing pressure; AUDHa-'-Stocking rate; AUHa-'-Stocking density; DL&L-DeseretLand Livestock C. Ranch; OLS-Ordinary Least Squares regression; PCA-Principal Components Analysis; PDSI-Palmer DroughtSeverity Index; SAVI-Soil Adjusted Vegetation Index; VI-Vegetation Index.

IntroductionLand managers throughout the Western United S tates

are faced with the problem o f managing their natural re-sources in a sustainable manner. Some private land m an-agers such as corporate agribusinesses and ranchers (pas-toralists), have the goal of managing landscapes forcontinued or sustainable profits. This goal implicitly in-corporates sustainable use of soil, vegetation, and othernatural resources. Similarly, federal agencies such a s theUnited States Departments of Defense's A rmed Servicesand the Interior's Bureau of Land Management (BLM)

have the explicit goal of compliance with en vironmentallaw, e.g., the Endangered Species Act, and m anaging theirlandscapes for sustainable use despite differing perspec-tives, e.g., realistic military training and testing environ-ments and available forage for livestock and wild horses,respectively. How ever, both public and private land man-agers are faced with the dilemma of the actua l ecologicalstatus and trend of the natural resources they are chargedto manage.

This problem is compounded by the failure of tradi-tional field monitoring techniques to samp le both the spa-

* The submitted manuscript has been authorized by a contractor of the U.S. Government under contract No. DE-ACOS-960R22464. Accordingly, the U.S. Government retains a non-exclusive, royalty-free license to publish or reproduce the publish-ed form of this contribution, or allow others to do so, for U.S. Government purposes.
mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]


2/11

70 Washington-Allen et al.

tial and temporal heterogeneity of semi-arid and aridlandscap es (Washington-Allen 2003). For exam ple, morefield samples are required as the area of concern becomeslarger and heterogeneous. This increase in sam ple size in-creases tim e, the number of data collectors, and most defi-nitely the expen se of data collection. Additiona lly, thereis the problem of separating declines in natural resourcesdue to land management decisions from those induced bychanges in climatic conditions. The suggested solution tothis matter is to increa se the sam ples in time to at least twotimes the scale of the driving climatic phenomena (Mag-nuson 1990). For example, the fauna and flora of ran-geland ecosystems are constrained by El Nino-SouthernOscillation (ENSO) events that have a 2 to 7 year returninterval, thus requiring continuous monitoring from 4 to15 years (Glantz 2001, Bonan 2002, Holmgren et al.200 1, W ashington-Allen 2003). Very strong ENSOevents have longer repeat intervals. Consider that in thelast 25 years, two major NSO events have occurred in1982-1983 and 1997-1998 (Bonan 2002). However, fewsites at regiona l and national spatial scales or greater aremonitored for this length of time or longer (Magnuson1990).

A technology that has the capability to monitorchanges in natural resources at large spatial and temporalscales is Landsa t satellite imagery (Graetz 1987). Landsa tdata have been collected every 14 to 16 days since 1972.Landsat radiometers collect spectral data from landscapesat a spatial grain of 0.09 and 0.62 ha pixels with nearglobal extent. Ecological indicators can be derived fromthe spectral and textural characteristics of Landsat im-

agery including growth form composition (Washington-Allen 2003), vegetation response (Washington-Allen etal. 1998 and 2003a) including LA1 (Wylie et al. 2001),plant cover (Graetz et al. 1 988, Hostert et al. 2003), pro-ductivity (Reeves et al. 1999), soil erosion (Pickup andNelson 1984), soil quality (Frazier and Cheng 1989,Palacios-Orueta and U stin 1998, Washington-Allen et al.2003b ), and structure and configuration (Spies et al. 1994 ,Washington-Allen 2003). Maps of these indicators canidentify for land managers the locations of degradation,stability, and improvement of natural resources. Conse-quently, the purpose of this study was to de tect the loca-tion of trends of vegetation response and relate the re-sponse to grazing and climate change. The objectives ofthis study were to (1) characterize the historical trend ofthe vegetatio n response of a sagebrush steppe plant com-munity in term s of a vegetation index (VI) time series thatwas d erived from 26-years (1 972 to 1997) of ry seasonLandsat im agery and (2) relate the vegetation respon se towet and dro ught periods and grazing processes.MethodsStudy site

The study area selected for this research was the54,000 ha lower elevation, eastern half of Deseret Landand L ivestock Company Ranch (DL&L ) which is locatedin the northeastern corner of Utah panhandle betw een lati-tudes41 0'and41 30'andlongitudes 1-1l O 'and 111 30'(Fig. 1). DL&L is within the Middle Rocky Mountainphysiograp hic province and is the largest holding of con-

Figure 1. Deseret Land LivestockCompany Ranch is the study site forthis research.


3/11

Spatiotemporal mapping in sag ebrush steppe 71

tiguous private land in the state of Utah at 88,800 ha in-cluding 6,800 ha of imbedded public and state land.DL&L has been comm ercially grazed since 1891 initiallyby sheep until the early 197 0s and is now a commercialmixed grazing operation where primary land-use and in-come generation is through a livestock (cattle and leasedsheep) and a recreational wildlife-hunting program.Vegetation is predominately sagebrush steppe (Knight1996, West and Young 2000). The sagebrush Artemisiatridentata ssp. wyorningensis and other Artemisia spp.) isassociated with introduced and native grasses includingcrested wheatgrass Agropyrum cristaturn L.), westernwheatgrass Pascopyrum srnithii (Rydb.) A. Love) andsandberg bluegrass Poa secunda Vasey).

The vegetation is associated with a rolling terra in withpredominately southern aspect slopes between 0-70%that consists of floodplain alluvium, alluvial fans, andstream terraces that overlie loess and residuum co mpo sedof shale, siltstone, mudstone, and sandstone (Hunt 1974,Chronic 1990, Lageson and Spearing 1991). Soils are pre-dominately gravelly loams in the A ridisol and Mollisolorders. Elevation increases from east to west from 1830m to 2670 m . M ean annual temperatures range between-5C and 15 C. Precipitation increases from east to westacross the ranch [Soil Cons ervation Service (SCS) 19821,with approximately 50% of the precipitation coming assnow (Danvir and Kearl 1996). Mean ann ual rainfall alsovaries with elevation, changing from 24 0 m m in the eastto 449 mm in the western portion of the ranch. Beck(1 994) estimated between 400 - 700 kg ha - r - annualabove-ground net primary productivity (ANPP) forcrested whe atgrass seedings and sagebrush o n control andirrigated sites on DL& L.Data sets

DL&L land management provided their grazing plansfrom 1980 to 1997 that included livestock count data.This data set provided spatial information by paddock,number of livestock, number of herds, timing, length ofstay, and frequency within a paddock, and location ofgrazing from 1980 to 1997.

Cattle count data were standardized to AU using con-version values from Holechek (1988, in Table 5 , p. 12)and calculations for stocking rate (Animal Unit Day perhectare, AU D Ha-') and grazing pressure (AU H a-') fromHeitschmidt and Taylor (1991). Wild herbivore popula-tion data were available, but were no t used in this analysisbecause past studies had shown that livestock were theprimary vertebrate herbivores on D L&L (Ritchie and O lff1999). For the period of this study, 1 972 to 19 97, DL&L

was grazed by a mean of 4059 700 AU within the 71paddocks distributed in Rich County, Utah.

Grazing paddocks were digitized and attributed usingeleven 1991 USGS topographic maps. Climate data, in-cluding information on precipitation, temperature,drought [Palmer Drought Severity Index (PDSI)], andENS0 and La Nifia for Woodruff and Hardware Ranchmeteorological stations, and Utah Region 5 were acquiredfrom the Utah Climate Center, the W estern Region Cli-mate Center, and N ational Oceanic and Atm ospheric Ad-ministration's (NOA A) National Climatic Data Center.Fire data that were derived from BLM field notes andLandsat imagery were available, but because the methodto be described below was developed to detect specificcoarse directional trends, finer temporal behavior wasmasked.

Though no official endorsement should be inferredfrom the use of product names, satellite and GeographicInformation System (GIS) digital data were pro cessed andanalyzed using A RC/INFO 's G RID module (ESRI 1991),ERDAS Imagine (ERDAS 1994), and Idrisi (Eastman1999) software packages. SPSS SPSS Inc 1999) wasused for desc riptive and infere ntial statistical analyses.

Twenty-two (22) dry season anniversary LandsatMSS and TM scenes from 1972 to 1997 were acquiredfrom the US Geological Survey's EROS Data Center(EDC) in Sioux Falls, South Dakota (Table 1). Landsatscenes from 1977, 1978, 198 3, and 1993 were either notavailable or not suitable, due to excess radiometric error(striping) and cloud c over. Local precipitation data, an an-nual MSS dataset from 1985 to 1986 (Washington-Allen2003), AVHRR (Advanced Very High Resolution Radi-ometer) normalized difference vegetation index (ND VI)trends for DL&L (Yorks, Schwartz, and West, personalcommu nication), net primary productivity data (Beck's1994), and image quality criteria provided by EDC, par-ticularly cloud cover estimates of 0 to 30 cloud coverand 0 mm of rain on the imag e acquisition date, were themain criteria used to select the interannual time series.Wet season precipitation events typically run from lateApril to late June on DL&L and are dependent on in-creased temperatures and snowm elt in March and April.The dry season is from late June to September, with peakdrought and highest tem peratures typically in July. How-ever, because grou ndw ater water availability is linked tosnow melt, the AVHRR NDVI trend indicated that thewet season was from June to early July and the dry seasonfrom late August to early Septemb er. Consequently, dryseason imagery was selected for this analysis.


4/11


Table 1. The dry season interannual time series (from 1972 to 1997) of Landsat Multispectral Scanner (MSS) and ThematicMapper (TM) satellite imagery that was used in this study. GCP = Ground Control Points and RMSE = Root Mean SquareError.Acquisition Landsat Scanner &Number Sun Zenith Angle RMSE 30 GCPs)08/07/ 1972 MSS 34 0.1309/07/1973 MSS 42 0.1908/15/1974 MSS 37 0.0409/06/1975 MSS 2 43 0.0409/09/1976 MSS 49 0.0409/03/1979 MSS 42 0.0608/28/1980 MSS 2 41 0.0408/21/1981 MSS 2 40 0.0508/09/1982 MSS 3 3s 0.0208/28/1984 MSS 40 0.1609/16/1985 MSS S 36 0.0409/03/1986 MSS 5 42 0.0409/06/1987 MSS 5 42 0.0409/08/1988 MSS 5 42 0.0808/28/1989 TM S 39 0.0608/29/1990 MSS 5 41 0.0909/17/1991 TM 5 46 0.0309/03/1992 MSS 42 0.3709/09/1994 T M S 44 0.0208/27/1995 TM 5 33 0.0808/29/1996 TM5 41 0.0409/01/1997 T M S 40 0.0

Date

Landsat image scenes were geometrically rectifiedand nearest neighbor resampled to the resolution (60 m)of an August 7,1972 LandsatMSS image, from the NorthAmerican Landscape Characterization (NALC) data set(Lunetta and Sturdevant 1993), between 0.25 and 0.50pixel RMSE (as recommended by Jensen 1996). The im-agery w ere normalized to exo-atmospheric radiance fromdigital numbers and then converted to reflectance valuesusing Landsat MSS and TM post-launch calibration gainsand biases from tables and formulae provided by Mark-ham and B arker (1986).

The interannual data set was then atmosphericallycorrected using a relative atmospheric correction proce-dure developed for multi-temporal imagery (Hall et al.199 1, Jensen 1996, Callahan 2002, W ashington-A llen2003).

The standardized data set was then converted to soil-adjusted vegetation index images (SAVI, Huete 1988)(Fig. 2). SAVI is a measure of vegetation greenness, andby correlation , a surrogate measure of cover, biomass, orleaf area index (Sellers 1985 and 1987, Bastiaanssen1998). SAVI was specifically developed and is recom-mended for arid environments to reduce soil backgroundeffects on the vegetation signal and is calculated as:SAVI = [ NZR- RED)/ NZR+ RED + L)]* (1 + L [11NZR is the near-infra red and RED is red reflectance bands,respectively. T he L is an adjustment factor which variesfrom 0 1 in accordance with soil background conditions(Huete 1988). The recommended L factor of 0.5 was usedfor all images (Huete 1988). The spatial changes within


5/11

Spatiotemporal mapping in sagebrush steppe

1

I73

Figure 2. The dry season time series of soil-adjusted vegctation index (SAVI) images of the sagebrush steppe portion of De-seret Land Livestock Company ranch from 1972 to 1997.

fire and climate could be d etected by clustering a time se-ries of SAVI images.ach SAVI image in the time series were visualized bythresholding the images continuous values by a f 2

standard deviation statistical interval (Jensen 1996). Theyears 1982, 1992, 1994, 199 5, and 1997 have particularlypronounced numbers of high SAV I values (dark red) inDL&Ls western foothills and medium SAVI values(green cells) in the sagebrush-grassland portions of theeastern half of the ranch (Fig. 2). A dry season time seriesof scenes were used because wet season scenes tend tocapture the dynam ics of the more ephem eral compon entsof plant c omm unities such as annu als. Dry season scenesare appropriate for this analysis because they are con-strained to the m ore permanen t ground cover.

Spatiotemporal mapping

Van N iels (1995) procedure for clustering the SAVItime series was followed by first using ERDAS ImaginesISODA TA clustering procedure to initially identify tem-poral s ignatures (trend clusters, ERDAS 1994). Minimumdistance-to-means cluster analysis procedures were thenused to delineate these trend clusters across the landscape(ERDAS 1994). The hypothesis for this multivariate sta-tistical procedure is that four coarse time series trends arepossible: increasing, decreasing, stable, or co mbinationsof these, includ ing threshold dynamics. Cluster analysiswill de tect similar temporal trends across the landsc apethat may be grouped into these three categories andmappe d across the landscape. F or this study, three maintrends (increasing, decreas d stable) were detected.Inference

If a management intervention such as grazing or pre-scribed fire had caused a persistent decline in the v egeta-tion index, a land manager would like to be aware of thelocation of this trend on the landscape. Eastman andMcKendry (1991) sugg ested using principal componentsanalysis (PCA ) for multiple scene comparisons n order toseparate temporal variance from the spectral variance ofthe data. Van Niel (l995 ) found that both the response andrecovery of a landscapes vegetation from the effects of

A significant slope D) s a m easure of the direction oftrend, i.e., stable 0), increasing 0< D +1) a nd decreas-ing 0 > 13 Z -l), and the magnitude of the coefficient ofdetermination Y ) from a linear or polynom ial regressionis a measure of the SAVI trend (Yafee and McGhee2000). The correlations between Palm er Drought Severity

2


6/11


Index (PDSI), different grazing variables, and the tempo-ral SAVI trends from 1972 to 1997 were determined usinga first order difference regression model. Ordinary leastsquares (OLS) regression results between time seriesmust be interpreted with caution. OLS regression analysisassumes that the mean and variance of a time series areconstant over time and the covariance between two timeperiods depends only on the lag or distance between thetime periods, that is, they are stationary. The SSI, PDSI,and grazing variables may contain a stochastic trend andtherefore be nonstationary. Nonstationarity violates theassumption of OLS, which tends to overstate the statisti-cal significance of variables with stochastic trend other-wise termed spurious regressions (Granger and New-bold 1974, Yafee and M cGhee 1998, Zhou e t al. 2000).The Dickey-Fuller test statistic (Dickey and Fuller 1979,Yafee and McGhee 1998) can be used to detect stochastictrend, but is not reliable with short time series (17 to 30observations). One way to reduce the likelihood of a spu-rious regression is to detrend the time series, thus remov-ing the stochastic trend. This entails transforming thetimes series using either order of differencing, runningmeans, lags, or some other smoo thing operation (Yafeeand McGhee 1998). This study followed the lead of Zhou

0 90

0 80

0.mr o*6a4 0 50P 0 40

0.300 .M0 10

l

et al. 2000) in their analysis of an NDV I time series ver-sus temperature and precipitation, by using the followingfirst order difference regression modelAY o + D X + E,in which AY nd AX are the first differences of X and Y ,Do,I3 1 are the regression coefficien ts, and E is a stochasticerror term.Results

PI

Twenty-five temporal signatures were identified fromthe ry season SAV I time series and their trends exam-ined using linear regression (Yafee and McGhee 2000).The 1992 and 1994 images had uncorrectable atmos-pheric and radiometric anomalies and were removed fromthis analysis. The 25 temporal signatures were mergedinto three coarse trajectories representing increasing, de-creasing, and stable trends (Fig. 3A). The three trajecto-ries were then mapped to the landscape (Fig. 3B). Thetrends signify portions of the landscape that are decreas-ing, increasing, and stable in vegetation response. A re-gression line shows the trend and d irection of each o f thetemporal signatures. The stable trend had a non-signifi-

Ubb 7012 sinaeass 34t62 2L I C ~ O P P ~ 68?30 82

Figure 3. The ry season SAVI temporal signatures of the sagebrush steppe portion of Deseret Land Livestock Companyranch fiom 1972 to 1997 (A). The spatiotemporalmap of the three 26-year SAVI trends, increasing, stable, and decreashg-stable, mapped across the landscape (B).


7/11

Spatiotemporal mapping in sagebrush steppe 75

Table 2. Correlations r)and significance (p < 0.1) betweenthe Palmer Drought Severity Index (PDSI) and one-yearlagged soil-adjusted vegetation index (SAVI) trend clustersfrom 1972 to 1996 on Deseret Land Livestock Ranch,Woodruff, Utah.

Cluster trend r PIncreasing 0.38 0.07

Stable 0.20 0.36

cant slope 13 = 0.001, p = 0.62) and the increasing trendhad a s ignificant slope I3 = 0.003, p = 0.07). The decreas-ing trend had a non-significant negative slope I3 =-0.004,p = 0.36). The trends of the decreasing (the black areas,Fig.3B) and increasing (the grey areas, Fig.3B) clusterswere som ewhat negligibly different from the stable trend(the light grey areas, Fig. 3B ). Of course, another year ofdata could also change the s ignificance of the trend. How-ever, though a trend m ay not be statistically significant ornegligible in difference from the stable trend, it may beecologically significant. Ecological significance will beconsidered by retrospective inference or retrodiction(Schumm 1991, Suter 1993), Le., correlation with as-sumed constraints or disturbances. Consequently, forpractical purposes of visualization and ecolog ical signifi-cance, the two trends were mapped to the landscape. Thedecreasing trend encompasse s approximately 6670 ha o fthe landscape and is located predomina ntly in the higherelevation, western portion of the ranch image, particularlyin the riparian and mead ow portions (Fig. 3B). Althoughthe trend is decreasing, it has higher SAVI values than thestable and increasing trends (Fig. 3A). This is becausethese areas are wetlands, wet m eadows, irrigated, and ri-parian areas within DL&L. Irrigated areas usually havehigher vegetation index values relative to other areas inthe surrounding landscape. The stable trend covers some7012 ha and tends to transition spatially from the decrea s-ing trend, not surprising given that there is practically nodifference between their slope s. The stable trend was alsolocated in some of the riparian and meadow portions ofDL&L. The increasing trend covered approximately34,192 ha and was mainly located in the grassland andsagebrush portions of the landscape (Fig. 3B).Relation to climate

PDSI is a regional integrated measure of moistureavailability, site-water balance, or effective p recipitation(Palmer 1965, Alley 1984). PDSI is calculated usingmonthly temperature and precipitation data. PDSI valuesbetw een -2 and + 2 indicate normal moisture conditions.Values less than -2 indicate increasing drought severity

and values greater than +2 indicate increasing moisture(Bonan 2002). Only the increasing trend of one-yearlagged first order difference of S AV I was positively cor-related with the first order difference of PDSI r = 0.38and p = 0.07, Table 2).Relation to grazing

The grazing characteristics of 12 paddocks were spa-tially associated with the increasing and stable mappedtrends (Table 3, Fig. 4). Grazing data were from 1980 to1997. The grazing variables considered were meannumber of days grazed per year, grazing pressure (animalunit per year, A W R- '), stocking density (animal unit perhectare, AUH a-'), an d stocking rate (animal unit per hec-tare per day, AUD Ha-').

Across the first order differences of the graz ing vari-ables, only tw o paddocks lacked sign ificant linear corre-lations with the first order differences of the 2 SAVItrends: Steer South and The Bench. Of the remaining 10paddocks, no ne of them had sign ificant simultaneous cor-relations with the 2 trends. The g razing variables with themost correlations with SAV I trends within paddocks weremean number of days grazed per year and grazing pres-sure, both with 6 paddocks total. Of these 6 paddocks, 3overlapped, including S pring Canyon, Railroad, and FiveSprings Paddocks (Table 4). Within Sp ring Canyon Pad-dock, both grazing pressure (AUYR-') and number ofdays grazed were negatively correlated with the stabletrend, implying that SAVI slightly decreased with in-creased grazing (Table 4). Both va riables were positivelycorrelated with the increas ing trend in Railroad paddock,implying that as grazing increase d SAV I increased (Table4). In Five Springs paddock, grazing pressure was nega-tively correlated with the stable trend, implying that asgrazing increased vegetation response slightly decreased(Table 4).Discussion

In general, herbivory, climate change, and fire areconsidered to be the primary constraints to vegetation re-sponse in sem i-arid and arid landscapes (Noy Meir 1973and 1975, Loehle 1985, Jameson 1988, Lockwood andLockwood 1993) and in sagebrush steppe in particular(Laycock 1991, Knight 1994, West and Young 2000).Pastoralists concerns for the land scapes they manage areprimarily linked to forage availability. However, also ofconcern to land managers is whether the historical trendof vegetation respon se is a persistent decline, increase, orstable. If these trends are detected, then a land managerwould like to know the locations and what causa l factorsmay be responsible.


8/11


Table 3. The grazing characteristics of the 12 paddocks that were spa tially associated with the three mapped spatiotemporaltrends. AUDHa-' is Animal Unit Day per hectare, the standardized number of livestock for a day in a certain area.

Paddock AUDHa-' Mean Days Grazed Area (Ha) Number ofYearsUsed1980- 1997)

The Bench 6 10 929 10Rabbit SouthRailroadBuffalo JumpBrown HollowSpring CanyonRabbit NorthMcKaySteer NorthSuttonSteer SouthFive Springs

81010131314146222360

17 1919 89 1112 613 1080 914 962 835 3281 1016 1472 818 1393 1310 1109 1018 1004 913 1010 1045 2877 9

FemeDeoeasingIncreasingStable

Figure 4. Grazing variables for 12 of Desere t Land Livestock Company Ranch's 71 paddocks, including SC (Spring Can-yon), BH (Browns Hollow), BJ (Buffalo Jum p), SN and SS (north and south Steer), SU (Sutton), RN and RS (north andsouth Rabbit), McKay, Five Springs, BE (T he Bench), and RR (Railroad), were compared to the three 26-year SAVI trendsof the spatio temporal map.

Van Niel 1995) mapped more than 25 temporal sig-natures back to the landscape and detected spatially coin-cident and contemporary fire events by overlaying datedfire boundaries, thus spatially associating fire events to

mapped trends. Fires were not considered in this study.The scale of this study was large enough to exclude fireas a significant factor. The technique used in this studyproduced 3 coarse temporal signatures that masked out


9/11

Spatiotemporal mapping in sagebrush steppe 77

Table 4. Significant correlations r)and p-values of three grazing variables within three grazing paddocks that were spatiallyassociated with the 18-year (1980 to 1997) stable and increasing mapped soil-adjusted vegetation index (SAVI) trend clus-ters:Paddock Grazing Grazing Grazing Mean Mean MeanPressure Pressure Pressure Days Days Days(SAVI r) @) Grazed Grazed Grazed

Trend) (SAVI 4 (PITrend)Spring stable -0.53 0.03 stable -0.64 0.005

~CanyonRailroad increasing 0.63 0.007 increasing 0.65 0.005Five stable -0.48 0.05 NA NA NASprings

Figure 5. The 26-year trend (1972 to 1997) of the dry season soil-adjusted vegetation index (SAVI) for the entire 54,000 hasagebrush steppe dominated landscape on Deseret Land Livestock Company Ranch, Woodruff, Utah.

the acute local effects of fire on SAVI. At the landscapescale, the overall trend for the dry season vegetation re-sponse significantly increased from 1 972 to 1997 r =0.22 and p = 0.0274, Fig. 5). This finding agrees with thespatiotemporal map summary where 71% of the land-scape exhibited an increasing trend (Fig. 3B).

This study is retrospective, thus past changes to thelandscape have already occurred. Where the causes ofthese changes are unknown, inferences in the form of cor-relation with hypothesized constraints can be attempted toretrodict the response (Schumm 1991, Suter 1993). Theoverall trend was sig nificantly linearly correlated to wetand drought periods PDSI) at the landscape scale (Wash-

ington-Allen 2003) and the mapped increasing trend wasalso significantly correlated to the PDSI (Table 2).

The dry season SAVI trend was non-linearly corre-lated with herbivory at the landscape sca le (Washington-Allen 2003). Th e spatiotemporal map partitioned this sin-gle trend into three response trends: stable, increasing,and decreasing, and visually located them within a pad-dock . With the partition of vegetation response, the corre-lations were mixed (negative and positive effects of graz-ing), thus reflecting specifically the non-linearity andcomplexity of plant-herbivore interactions and ecologicalsystems in general (Hogeweg 2002).


10/11


The spatiotemporal map partitions vegetation re-sponse trends and then locates them on the landscape.Management at DL&L has the view that prior to the in-troduction of intense wildlife hunting and grazing man-agement in the early 1980s, the riparian corridors werebeing substantiallydegraded. The decreasingSAVI trend,though not significant,was locatedprimarily in the ripar-ian corridors. If we combine this non-significant trendwith the stable trend, then 29% ofDL&Ls landscapewasstable. These two findings confirm the conclusion of plotstudies by Wolfe et al. (2000) that vegetation cover onDL&L was increasing,but disagrees with their finding ofimprovedriparian cover. The findings in both Table 2 andTable 3 suggest that both localized grazing and regionalclimate have played significant roles in the increasedvegetation response.

ConclusionsIt is very rare to have this comprehensivea time series

of fined grained 0.09 to 0.62 ha) satellite imagery formonitoring changes in vegetation response at this extent54,000 ha) and for this time period 26 years). It is also

rare to be able to retrospectively correlate changes inlandscape vegetation response with spatially associatedmanagement practices, a criticismofboth community andlandscape ecology (Wiens 1995, Wu and Hobbs 2001).Pastoralists equire information on the historical legacyofvegetation change on the landscapes they manage, par-ticularly the location of trends in vegetation response inrelation to their management interventions. The vegeta-tion response to grazing and wet and drought periods of acommercially grazed sagebrush steppe-dominated land-scape in northeastern Utah was measured in time andspace by clustering a time series of 21 ry season SAVIimages. This time serieswas derived from the spectralre-sponse of Landsat satellite imagery from 1972 to 1997.This study indicated to land managers the locationof 7 1of the landscape that had an increasingtrend and 29% thathad a stable trend over the 26-year period. The increasingtrend was positively correlated with site water balancePDSI), i.e., vegetation response increased during wet pe-

riods and decreasedduring drought. The increasing trendwas positively and negatively (non-linearly) correlatedwith grazing in individual paddocks from 1980 to 1997.Acknowledgements. This research was partially funded by theU.S. Environmental Protection Agency (EPA) through a ScienceTo Achieve Results (STAR) grant GADR826112. Thisresearch has not been subjected to EPAs required peer andpolicy review and therefore does not necessarily reflect the viewsof the agency and no official endorsement should be inferred.This research was also sponsored by the Utah AgriculturalExperiment Station, of which this is Journal Paper No.7585, theAllen and Alice Stokes Martin Luther King Jr. MinorityGraduate Fellowship at Utah State University to R.A.

Washington-Allen and the Strategic Environmental Researchand Development program (SERDP) under contractDE-AC05-000R22725 with Oak Ridge National Laboratory,managed by UT-Battelle, LLC. Authors also wish to thank Mr.Gregg Simonds and DL&Ls Mr. William Hopkin for theircooperation and the two anonymous reviewers for their helpfuledits and comments.ReferencesAlley, W. M.1984. The Palmer drought severity index: Limitationsand assumptions.J Clim. App. Meteoro. 23: 1100-1109.Bastiaanssen, W.G.M. 1998. Remote Sensing in Water Resources

Management: The State of the Art. Int. Water Manage. Inst.,Colombo, Sri La&.Beck, E.W. 1994. The effect of resource availability on the activityof white-tailed prairie dogs. MS Thesis, Utah State Univ.,Logan, Utah.Bonan, G. 2002.Ecological Climatology Co ncep ts andApplications.Cambridge Univ. Press, Cambridge.Callahan, K. 2002. Validation of a radiometric normalization proce-dure for satellite derived imagery within a change detectionframework. MS Thesis, Utah State Univ., Logan, Utah.Chronic,H. 1990.Roadside Geology of Utah. Mountain Press Pub-lishing Co., Missoula, Montana.Dickey, D.A. and W.A. Fuller. 1979. Distribution of the estimatorsfor autoregressive ime series with aunit root. J Am. Stat. Assoc.74A27-431.Eastman, J.R. and J.E. McKendry. 1991. Change and Time Series

Analysis. Explorations in Geographic Information SystemsTechnology. Vol. 1 United Nations Inst. for Training and Res.,European Office, Palais desNations CH-1211 Geneva 10, Swit-zerland.

ERDAS. 1994.ERDAS Field Guide. ERDAS Inc., Atlanta, Ga.ESRI (Environmental Systems Research Institute). 1991.Cell-basedGranger, C.W. J. and P. Newbold. 1974. Spurious regressions inFrazier, B.E. and Y .Cheng. 1989. Remote sensing of soils in theeastern Palouse region with Landsat Thematic Mapper.Remote

Sens. Environ. 28:317-25.Glantz, M.H. 2001 Currents ofchange: ElNiiio andLaN iiia Impactson Climate andSociety. Second edition. Cambridge Univ. Press,Cambridge.Graetz, R.D., R.P. Pech and A.W. Davis. 1988. The assessment andmonitoring of sparsely vegetated rangelands using calibratedLandsat data. Int. J Remote Sens.9:1201-1222.Hall, F.G., D.E. Strebel, J.E. Nickeson and J. Goetz. 1991. Radio-metric rectification: Towards a common radiometric responseamong multidate, multisensor images. Remote Sens. Environ.Heitschmidt, R.K. and C.A. Taylor Jr. 1991. Livestock production.In: R.K. Heitschmidt and J.W. Stuth (eds.), Grazing M anage-

ment: An Ecological Perspective. Timberland Press, Portland,Oregon. pp. 161-177.Hogeweg, P. 2002. Computing an organism: on the interface be-tween informatic and dynamic processes. BioSystems 64:97-

109.Holechek, J.L. 1988. An approach for setting the stocking rate.Ran-

Modeling with GRID. ESRI. Redlands, California.econometrics. J . Econometrics 2:lll-120.

35:ll-27.

gelands 1O:lO-14.


11/11

Spatiotemporal mapping in sagebrushsteppe 79

Holmgren, M., M. Scheffer, E. Ezcurra, J.R. Gutitrrez and G.M.J.Mohren. 2001. El Niiio effects on the dynamics of terrestrialecosystems.Trends-Ecol. Evol. 16239-94.Huete, A.R. 1988. A soil-adjusted vegetation index (SAW). Remote

Sens. Environ. 25:295-309.Hostert, P., A. Roder, J. Hill, T. Udlhoven and G. Tsiourlis. 2003.Retrospective studies of grazing-induced land degradation: acase study in central Crete, Greece. Int. J. Remote Sens.Hunt, C.B. 1974. Natural Regions of the United States and Canada.W.H. Freeman and Co., San Francisco, California.Jameson, D.A. 1988. Modeling rangeland ecosystems for monitoringarid adaptive management.In: P. Tueller (ed.), Vegetation Sci-

ence Applications for Rangeland Analysis and Management.Kluwer, Dordrecht, The Netherlands. pp. 189-21Jensen, J.R. 1996. ntroductory Digital Image Processing: A Remote

Sensing Perspec tive. 2nd ed. Prentice Hall, Upper Saddle River,New Jersey.Landcapes. Yale Univ. Press, New Haven, Connecticut.

24~409-4034.

Knight, D. 1994.Mountains and Plains: The Ecology of WyomingLageson, D.R. and D.R. Spearing. 1991. Roadside Geology of Wyo-

ming. Second Edition, Mountain Press Publishing Co., Mis-soula, Montana.

Laycock, W. 1991. Stable states and thresholdsof range conditiononNorth American rangelands: A viewpoint. J.Range Manage.Loehle, C. 1985. Optimal stocking for semi-desert range: a catastro-phe theory model. Ecol. Modeling 27:285-297.Lockwood, J.A. and D.R. Lockwood. 1993. Catastrophe theory: Aunified paradigm for rangeland ecosystem dynamics.J . Range

Manage. 46:282-288.Lunetta, R.S. and Sturdevant, J.A. 1993. The North American land-scape characterization Landsat Pathfinder Project. In: L.R. Pet-tinger (ed.), Land Information from Space-based Systems,Pecora 12 Symposium, Proc. Amer. SOC. hotogrammetry and

Remote Sens., Bethesda, Md. pp. 363-371.Magnuson, J.J. 1990. Long-term ecological research and the invis-ible present. Bioscience 40:495-501.Markham, B.L. and J.L. Barker. 1986. Landsat MSS and TM-postcalibration dynamic ranges, exoatmospheric reflectances and at-satellite emperatures.Earth Observation Satellite Co. (EOSAT)

Landsat Tech. Notes 1.3 - 8 .Noy-Meir, I. 1973. Desert ecosystems: Environment and producers.

Annu. Rev. Ecol. S ys. 4:25-5 1.Noy-Meir, I. 1975. Stability of grazing systems: An application ofpredator-prey graphs.J . Ecol. 63:459-481.Palacios-Orueta, A. and S.L. Ustin. 1998. Remote sensing of soilproperties in the Santa Monica mountains I. Spectral analysis.

Remote Sens. Environ. 65:170-183.

44:427-433.

Palmer, W.C. 1965. Meteorologic Drought. U.S. Weather Bureau,Res. Paper No. 45.Pickup, G. and D.J. Nelson. 1984. Use of Landsat radiance parame-ters to distinguish soil erosion, stability, and deposition in aridcentral Australia.Remote Sens. Environ. 16:195-209.

Reeves, M.C., J.C. Winslow and S.W. Running. 2001. Mappingweekly rangeland vegetation productivity using MODIS algo-rithms. J. Range Manage. 54:A90-A105.Schumm, S . A. 1991. To Interpret the Earth: Ten Ways To Be

Wrong. Cambridge University Press, New York, NY.Sellers, P. J., 1985. Canopy reflectance, photosynthesis and transpi-ration. In?.J. Remote Sens. 6: 1335-1372.Shaw, R.J. 1989. VascularPlants ofNorthem Utah:An Identification

Manual. Utah State Univ. Press, Logan, Utah.Spies, T.A., W.J. Ripple and G.A. Bradshaw. 1994. Dynamics andpattern of a managed coniferous forest landscape in Oregon.

Ecol. Appl. 4555-568.SPSS Inc. 1999.SPSS Trends 10 0 SPSS Inc. Chicago, Illinois.Suter 11, G.W. 1993.Ecological Risk Assessment. Lewis Publishers,Boca Raton, Louisiana.Van Niel, T.G. 1995. Classification of vegetation and analysis of itsrecent trends at Camp Williams, Utah using GIS and remotesensing techniques. MS Thesis, Utah State Univ., Logan, Utah.Washington-Allen, R. A. 2003. Retrospective Ecological Risk As-sessment of Commercially-Grazed Rangelands using Multitem-poral Satellite Imagery. PhD Thesis, Utah State Univ., Logan,Utah.Washington-Allen, R.A., N.E. West, R. D. Ramsey and D.K. Phil-lips. 2003a. Retrospective Assessment of soil stability on a land-scape subject to commercial grazing.African Journal of Range

Forage Science 20:127Washington-Allen, R.A., N.E. West and R. D. Ramsey. 2003b. Re-mote sensing-based dynamical systems analysis of sagebrushsteppe vegetation in rangelands. Aji-ican Journal of Range

Forage Science 20: 100.Washington-Allen, R. A,, R. D. Ramsey, B. E. Norton and N. E.West. 1998. Change detection of the effect of severe droughtonsubsistence agropastoral communities on the Bolivian Alti-plano. Int. J . RemoteSens. 19:1319-1333.West, N.E. and J.A. Young. 2000. Intermountain valleys and lowermountain slopes. In: M.G. Barbour and W.D. Billings (eds),

North American Terrestrial Vegetation. 2nd ed. CambridgeUniv. Press, New York, N.Y. pp. 255-284.Wiens, J.A. 1995. Landscape mosaics and ecological theory. In: L.Hanson, L. Fahrig and G. Merriam (eds), Mosaic Landscapes

and Ecological Processes. Chapman and Hall, London, UK. pp.1-26.Wolfe, M.L., M.E. Ritchie, andR. Danvir.2000.Managing for cattleand wildlife on Deseret Ranch. In: Ecol. SOC.Amer. (ed), Ab-stracts, 85th annual meeting Ecol. SOC.Amer., Snowbird, Utah.Wu, J. and R. Hobbs. 2002. Key issues and research priorities inlandscape ecology: An idiosyncratic synthesis.Landscape Ecol-

ogy 17:355-365.Wylie, B.K., D.J. Meyer, L.L. Tieszen and S.Mannel. 2002. Satellitemapping of surface biophysical parameters at the biome scaleover the North American grasslands: A case study.Remote Sens.

Environ. 79:26&278.Yafee, R.A. and M. McGhee. 2000. Infroduction to Time SeriesAnalysis and Forecasting with Applications of SAS and SPSSAcademic Press, San Diego, Ca.

pp. 37.

Documents

s Patio Temporal Mapping