Detection and mapping of vegetation stress using AVIRIS-NG hyperspectral imagery in coal mining sites

Abstract

This paper primarily focuses on identifying and mapping vegetation stress caused by the impact of coal mining using airborne hyperspectral image (AVIRIS-NG) at a fine scale level and is validated using spectroscopic field data. In this work, we have calculated and tested vegetation stress-affected narrow banded vegetation indices (VIs) based on Separability (S) and Coefficient of discrimination (R2) statistical tests for the identification of suitable vegetation stress indices. We have considered the highest S and R2 values of stress indices for weighted and vegetation combined pixels’ analysis. The final weighted combination index has been used for vegetation stress detection and mapping in coal mining sites. The outcome has been validated using field-based healthy and stress-affected plant spectral data and compared to ENVI software's agriculture stress tool (AST) based vegetation stress result. Some indices exhibited higher performance for vegetation stress analysis (NDVI, NSI-2, SR-1, SWST-3, SRWI, NWI-2, CSVI, HMSSI, and ARI-1), according to statistical results (S and R2) of VIs. Based on the vegetation stress index, high-stress zones are located in and around the mines whilelow-stress regions are positioned farther from the mines. The findings additionally demonstrated that the VIs-based stress result (AUC-0.69) fits the data more accurately than the ENVI's AST result (AUC-0.61). The VI’s-based stress index had a clearly negative correlation with the distance of vegetation from the mines, which could be verified using the on-field NIR spectral measurements. Therefore, effective vegetation health monitoring and planning in coal mining sites might benefit from these results .