Irrigation and Drainage
393590 - Estimation of Vegetation Coverage based on Seasonal Variabilities in MODIS-based Vegetation Indices
Tuesday, June 5
4:00 PM - 5:30 PM
Location: Skyway Room
Determination of the Vegetation Coverage (VC) plays an important role in different aspects of watershed management. Remote sensing products can provide estimates of VC; however, the uncertainty of the data acquisition and processing has always challenged the usability of spaceborne imagery data. In this study, Moderate Resolution Imaging Spectroradiometer (MODIS) products used to model VC of three basins: Lower Fox (LF) in Wisconsin, Lower East Fork White (LEFW) in Indiana, and Lower San Joaquin River (LSJR) in California. MODIS MOD13Q1 product is used to obtain VC based on Enhanced Vegetation Index (EVI) and Normalized Difference Vegetation Index (NDVI). However, the commonly used threshold value of 0.4, used to differentiate vegetated from non-vegetated areas, does not necessarily yield to accurate estimates when compared to the National Land Cover Dataset (NLCD). A statistical criterion—determined based on probability distributions of VC—used to filter pixels with low Vegetation Index (VI) values during the time of peak-vegetation as non-vegetated. The same approach implemented to determine vegetated areas during the low-vegetation season. These pixels were left out of analysis—regardless of the VI value—and their VC condition was determined based on the group of filtered data they belong to. Then, VC was calculated for the remaining pixels and results were integrated with those of the filtered areas. Application of the suggested technique improved VC estimations. The errors of VC estimations based on NLCD-VI comparison were reduced by 0.03%, 3%, and 5% for EVI and 17%, 6%, and 7% for NDVI in LF, LEFW, and LSJR basins, respectively.