Sugarcane drought detection through spectral indices derived modeling by remote-sensing techniques

Several indices based on satellite images have been explored to monitor agricultural drought. Despite the existence of some drought indices, no drought monitoring system for sugarcane exists. In this sense, drought detection could be useful tool to quantify losses and help with action plans. This study investigates the Landsat image potential for sugarcane drought detection by assessing the relationship between vegetation and agricultural drought indices (normalized difference vegetation index (NDVI), vegetation condition index (VCI), normalized difference water index (NDWI), global vegetation moisture index (GVMI), and normalized difference infrared index (NDII)). Two new indices combining near-infrared (NIR) and short-wave infrared (SWIR) bands are proposed for sugarcane drought detection. All indices were individually and collectively compared with soil water deficit and water surplus, simulated by the climatological soil–water balance (CSWB) model. A significant correlation between spectral indices and water balance results, specifically for NDVI and VCI indices (~ 30%), was observed. The drought detection system identification was developed by cluster analysis classifying the pixels into three distinct groups (drought, intermediate drought, and non-drought) to later be used in the discriminant analysis. This methodology showed to have an accuracy rate of 65%. However, the discriminant analysis approach was better suited for sugarcane drought monitoring when compared with individual spectral indices.