A new four-stage approach based on normalized vegetation indices for detecting and mapping sugarcane hail damage using multispectral remotely sensed data

Abstract

Hailstorms have increased in frequency and intensity over the past decade causing substantial losses in agriculture. Since hailstorms often hit a wide area with no detectable pattern, relying on traditional field-based methods to assess crop damage is difficult. The aim of this study was to develop and test a new four-stage normalized vegetation index approach for detecting the severity of hailstorm damage on sugarcane plants in a large estate in south eastern Zimbabwe. The following six spectral indices were computed for the period before and after a hailstorm event to assess the extent of hailstorm damage on sugarcane: Green Chlorophyll Index (GCI); Normalized Difference Vegetation Index (NDVI); Normalized Difference Senescent Vegetation Index (NDSVI); Red Edge Chlorophyll Index (RECI); Normalized Difference Tillage Index (NDTI); and Modified Soil Adjusted Vegetation Index (MSAVI2). Then, the spectral differences were computed for each index separately and the difference maps are reported as delta (Δ) indices. The results of this study show that within one week and even two weeks after a hailstorm, ΔNDTI, ΔNDVI and ΔRECI were consistently able to detect and characterise the severity of sugarcane damage. When used in partial least squares-discriminant analysis (PLS-DA), ΔNDTI performed best in mapping the severity of crop damage throughout the large estate. ΔNDTI was able to discriminate three different levels of sugarcane damage with an overall accuracy of 90% and a Kappa value of 0.85. Combined these results imply that ΔNDTI computed using multi-spectral datasets within a fortnight after a hailstorm is a promising tool for generating reliable information about the severity of sugarcane damage by hailstorms. Such spatially explicit information is useful for creating customised crop insurance packages sensitive to damage incurred by the farmer.