MULTITEMPORAL REMOTE SENSING OF VEGETATION IN THE CHEYENNE RIVER WATERSHED OF SOUTH DAKOTA AND WYOMING

Monitoring vegetation in a watershed is important for understanding its overall condition and providing information on potential sediment runoff. Most runoff models use general vegetation characterizations or average conditions that do not take into account changes during wet or dry years. The increased availability and lowered cost of remotely sensed data in recent years opens new opportunities for using multi-temporal as well as multispectral techniques. This study attempts to characterize vegetation conditions in the Cheyenne River watershed of Wyoming and South Dakota as a preliminary step towards assessing runoff potential. The immediate goal is to determine which vegetation factors can be discerned from multi-temporal Moderate Resolution Imaging Spectroradiometer (MODIS) data, and to classify land cover based on vegetation changes over the growing season. The study was conducted using Enhanced Vegetation Index (EVI) values from 16-day L3 global 250m MODIS composites for 2008 (MOD13A1.A2008).We constructed an 18-band image containing EVI measurements from January to December 2008, with each band representing a 16-day composite of EVI values. Three composites with extensive cloud cover were excluded from the analysis, and cloud pixels in the remaining composites were masked using a threshold based on the red spectral band. The final 18-band image was classified using an unsupervised isodata classification. The resulting 30 classes were grouped into five superclasses by plotting representative class locations and visually inspecting them in GoogleEarth. The superclasses included forest, grassland, water/barren, cropland, and mixed vegetation. Each superclass exhibits distinctive shapes in plots of EVI versus calendar date. Additional separation within classes and characterization of vegetation condition will be attempted in the future with site visits, and with the addition of ancillary data such as soils and elevation.