Vegetation Water Content Estimation Using Vegetation Indices of Low Resolution Satellite Data

This study investigates the potential of MODIS data for estimating Vegetation Water Content (VWC) using Vegetation Indices (VI) that work in VIS/NIR and SWIR spectral region and improving their performances with appropriate band selection. In fact the aim of this study is raising the possibility of the practical use of MODIS imageries which are available easily، for free and with high temporal resolution for estimation VWC parameter. The ground truth data which were used for validation، were collected from 31 stations in an agricultural area in the State of Iowa، USA. Among them، 21 stations belonged to corn and 10 stations to the soybean fields. Two sets of indices that work in the VIS-NIR and NIR-SWIR spectral region were calculated. By using the cross-validation the corn data were divided to 4 clusters and the soybeans data were divided in 2 clusters. Each time one cluster considered as test data and the other clusters as training. And thus by using the training data، scatter plots between VWC reference values and calculated VWC values using vegetation indices were sketched and linear regression was established. Then by using test data and R2 statistic the results were validated. According to the results، indices which use NIR-SWIR bands show better performance than indices that use VIS-NIR spectral range. In this study bands 5، 6 and 7 of MODIS sensors were used to calculate NDWI index that bands 5، 6 produced better results than band 7. Therefore، band 6 was used for the rest of indices. The results also showed that the experiment on the soybean fields were less reliable than corn fields because of low density and accuracy in soybean field data. The high potential of MODIS data to assess VWC was confirmed by comparing the results of indices and field data.