Remote sensing equivalent water thickness of grass treated with different fertiliser regimes using resample HyspIRI and EnMAP data

Abstract

The aim of this study was to test the spectral settings of the forthcoming EnMAP and HyspIRI hyperspectral imagers in estimating equivalent water thickness of native grass grown under different management practices, using two regression ensembles (partial least squares regression (PLSR) and sparse partial least squares regression (SPLSR). Results showed that the HyspIRI spectral settings better characterised Equivalent water thickness (EWT) of grass grown under different fertiliser regimes in relation to the EnMAP spectral settings. Specifically, t-test analysis indicated there were significant differences (α = 0.05) between the (pooled) relative root mean square errors derived using the HyspIRI spectral settings when compared to those exhibited by EnMAP. When all data was combined, HyspIRI spectral data settings exhibited relatively higher accuracy in estimating grass EWT with a RMSE of 3.039 cm−2 (RRMSE = 31.7%) and an R-square of 0.59 when compared to EnMAP which exhibited a RMSE of 3.216 g cm−2 (RRMSE = 33.6%) and an R-square of 0.54. The near infrared and shortwave infrared bands were selected as the most optimally variables in estimating grass water content across different grassland management practices using the SPLSR. These findings are critical for monitoring grasslands condition at landscape scale. Overall this information is critical for determining and understanding their moisture retention capacity across different land management practices and in light of climate variability and change.