The influence of land cover-related changes on the NDVI-based satellite agricultural drought indices

Abstract

Drought is a natural climatic event that often causes sharp declines in agricultural production. In recent years, drought indices based on remote sensing products have been developed on the premise that photosynthetic rate of vegetation slows down under drought/water stress, and this can be accurately tracked by the satellite data and methods. The Normalized Difference Vegetation Index (NDVI) is the most popular index with the long historical record to monitor terrestrial vegetation state around the world. It has been suggested that drought-induced NDVI decline can be confused with non-drought-related NDVI decline (e.g., fire, flood, land cover/land use change, pest infestation) in the NDVI-based drought method. To investigate the effect of land cover-related changes on the NDVI-based drought indices, we selected the Vegetation Condition Index (VCI), a popular NDVI-based drought index. We found that deforestation (i.e., land cover change) is falsely classified as drought in the VCI method, hence producing the false drought signals during the non-drought years. However, these false drought signals can be eliminated from drought maps with the help of spatial filters (e.g., median filter) because they are small scale signals relative to the study area size. Furthermore, the rotation of agricultural crops (e.g., crop rotation between corn and soybean) reduces the accuracy of drought reporting when crop rotation is not considered in the drought index computation because crop rotation annually accounts for over 50% agricultural land in Iowa, it has a significant impact on the NDVI-based drought methods. In conclusion, it can be said that the influence of land cover-related changes on the NDVI-based drought indicators is proportional to the size of non-drought related changes relative to the study area.