Dynamically Characterizing a Variety of Phenological Responses of Semi-Arid Areas to

Hydrological Inputs using Multi-Year AVHRR NDVI Time Series

John F. Hermance, Brown University

Robert W. Jacob, Brown University

Bethany A. Bradley, Princeton University

John F. Mustard, Brown University

e-mail: John_Hermance@Brown.Edu

"NDVI; The Movie"

Location of the Study Area

150 x 150 km: West Central Nevada Note the trend for elevation to increase to SE.

Topography of the Study Area.

Compare elevation (250 m DEM) and vegetation (AVHRR NDVI).

(Large red squares are subareas; small red polygons are farms.)

The NDVI parameter is the upper quartile of all 7 yrs data at each site.

Elevation (200 m contour interval) NDVI (Q75 for 1995 - 2002)

Compare vegetation (AVHRR NDVI) w/ Homer's ('97) landcover classes.

Homer's (1997) Landcover ClassesNDVI (Q75 for 1995 - 2002)

Bradley et als. (2006) landcover classes using this algorithmNDVI (Q75 for 1995 - 2002)

Short-term, multi-year hydrologic indicators for the region: the mean monthly precipitation at 6 stations, and stream runoff in the Humboldt River.

An objective of this study is to more closely identify particular fluctuations in the spatial and temporal patterns of NDVI with other environmental variables. Here, we explore hydrologic inputs.

Landsat: 1998_06-28.Thematic Mapper;Path 42 Row 32

(Post "Wet"Season)

Landsat scene.

Landsat: 2000_10-07.Thematic Mapper;Path 42 Row 32

(Post "Dry"Season)

Landsat scene.

Location map for our study area in west central Nevada, along with the names of adjacent states. Also indicated is the boundary of the Great Basin.

The interannual spline algorithm uses:Model roughness damping;Upper data envelope tracking.

The Algorithm.(Reported previously at this meeting.)

Original NDVI time series from 3 classes of vegetation types. Panel A: Stable agriculture. Panel B: Montane shrubland. Panel C: Invasive grasses (cheatgrass). The vertical gray band at approx. 1999.4 yr (actually from the 1 week data composite centered at 1999.414) denotes an inferred data gap associated with what is observed as a ubiquitous anomalously high NDVI data value for this interval throughout our Basin and Range data base.

Examples of data classes.

Each step in the procedure applied to the Cheatgrass time series. Panel A; Step 1: The simple annual harmonic starting model. Panel B; Step 2: The roughness damped, average annual harmonic model weighted to track the upper data envelope. Panel C; Step 3: A preliminary 8-th order inter-annual spline model with weak upper-envelope weighting. Panel D; Step 4: A 14-th order inter-annual spine model with stronger upper-data envelope weighting.

Stages in "fitting"Cheatgrass data.

Average annual variation models for the three classes of vegetation types. All models were computed using the same control parameters (weighting coefficients, roughness damping, etc.).

Average annual fit to 3 data classes.

Inter-annual (spline) variation models for the three classes of vegetation types. All models were computed using the same control parameters (weighting coefficients, roughness damping, etc.).

Inter-annual fit to 3 data classes.

Comparing the inter-annual spline model in this report (Panel A) with the average annual model (Panel B) of Hermance (2006) for the same time series from a single 1 x 1 km montane shrubland site. The substantial data gap (from 1994.72 to 1995.05) is due to sensor failure on board NOAA-11. Panel A: The inter-annual model uses 14th order annual splines. Panel B: The average annual model uses a 10th order non-classical harmonic series superposed on a 0th order polynomial (mean value). (Note the flags (a), (b) and (c) denote noteworthy tracking attributes.

Fit to "gappy" data.

Topography of the Study Area.

The NDVI animations will begin with the entire study area.

NDVI: Entire Study Area(1995-2002 Interannual variations.)

Note: Agricultural areas.Montane snow cover.Anomalous greenup in valleys.

Run animation.

NDVI from the Dixie Valley / Edwards Creek Valley Subarea.

Run animation.

NDVI: Dixie and Edwards Creek Valleys(1995-2002 Interannual variations.)

Note: Impulsive greenup of Cheatgrass in NE Edwrds Crk Valley in 1998.Behavior of montane snow cover.Concatenation of NDVI w agricultural areas.

NDVI from the Buena Vista Valley Subarea.

Run animation.

NDVI: Buena Vista Valley(1995-2002 Interannual variations.)

Note: Impulsive greenup of Cheatgrass in SW sector of valley in 1998.Behavior of montane snow cover.Concatenation of NDVI w agricultural areas.

NDVI from Desatoya Mountains Subarea.

NDVI: Desatoya Mountains(1995-2002 Interannual variations.)

Note: Asymmetric spatial and temporaldevelopment of montane snow cover and

greenup.

Run animation.

This poster is one of several, related presentations made at this meeting. For more details on the application of this and other methods to classification and monitoring land cover change in the western U.S. (and relations to climate change) see

Jacob, R., B. Bradley, J. Hermance and J. Mustard (2006) Phenology-based Land Cover Classification and Assessment of Land Cover Trends: A Case Study in the Western U.S., Eos Trans. AGU, 87(52), Fall Meet. Suppl., Abstract B33F-05

Bradley, B. A. (2006), Impacts of Climate Variability on Non-native Plant Invasion in the Western U.S., Eos Trans. AGU, 87(52), Fall Meet. Suppl., Abstract B41E-0230.

The Landsat-based landcover classification is from Homer (1997). Citation Information:Collin G. Homer, Dept. of Geography and Earth Resources, Utah State University Logan, Utah 84322 Publication_Date: 1997 Title: Nevada Landcover Classification Online_Linkage: <http://earth.gis.usu.edu/>

For reports on our latest work:Bradley, B. A., and Mustard, J. F., 2005, Identifying land cover variability distinct from land cover change:

Cheatgrass in the Great Basin. Remote Sensing of Environment, 94, pp. 204-213.Bradley, B. A., R.W. Jacob, J.F. Hermance, and J.F. Mustard. 2006. A curve fitting procedure to derive inter-annual

phenologies from time series of noisy satellite NDVI data. submitted to Remote Sensing of Environment, in press.Hermance, J. F., 2006, Stabilizing High-Order, Non-Classical Harmonic Analysis of NDVI Data for Average Annual

Models by Damping Model Roughness, submitted to International Journal of Remote Sensing, in press.Hermance, J. F., R. W. Jacob, B. A. Bradley and J. F. Mustard, 2006, Extracting Phenological Signals from Multi-

Year AVHRR NDVI Time Series: Framework for Applying High-Order Annual Splines with Roughness Damping. Submitted to IEEE Transactions of Geoscience and Remote Sensing, June, 2006.

For copies of reports on the algorithm:http://www.geo.brown.edu/research/Hydrology/NDVI_reports/

Background