Globally quantitative analysis of the impact of atmosphere and spectral response function on 2-band enhanced vegetation index (EVI2) over Sentinel-2 and Landsat-8

Abstract

The 2-band enhanced vegetation index (EVI2) was developed as an alternative to the enhanced vegetation index (EVI) for sensors that lack a blue band or for observing bright targets such as thick clouds, snow, and ice. However, research has indicated that EVI2 exhibits inconsistencies with the EVI when applied to non-moderate-resolution imaging spectroradiometer (MODIS) sensors. This study aimed to investigate the sources of these inconsistencies by examining the effects of atmospheric conditions and spectral response functions (differences in band reflectance due to sensor spectral response functions) on Landsat-8 and Sentinel-2 globally and evaluate the stability of EVI2 over snow-covered areas. First, a discrete anisotropic radiative transfer (DART) simulation in a snow-covered area was performed to test the atmospheric resistance and sensitivity of the leaf area index (LAI) to EVI and EVI2. Then, two scenarios were analyzed: an ideal case that investigated the spectral response function effect using simulated band reflectances from the United States Geological Survey (USGS) spectral library and a realistic case that examined the effect of both atmospheric and spectral response function using satellite images from the Google Earth Engine (GEE) globally. Our findings revealed that EVI2 demonstrated greater stability in resisting atmospheric influences and sensitivity to LAI than EVI in snow-covered areas. The blue band exhibited a significantly lower transmittance than the other bands, making it more vulnerable to atmospheric conditions. Furthermore, the discrepancy between the EVI and EVI2 for vegetation may be exacerbated by atmospheric effects, as these tend to overestimate the EVI and underestimate the EVI2. This could be attributed to the net positive atmospheric influence on blue band reflectance. The spectral response function effect analysis indicated that EVI2 values may be lower than EVI for vegetation when using Sentinel-2. This underestimation of EVI2 could be mitigated by modifying the spectral corresponding coefficient (c) to 1.45. Here, we demonstrate the superiority of EVI2 over snow-covered areas and offers reliable EVI2 implementation for Landsat-8 and Sentinel-2 application globally.