Comparison of various approaches for estimating leaf water content and stomatal conductance in different plant species using hyperspectral data

Abstract

Water deficit stress is a frequent phenomenon that inhibits plant growth. This study explores the performance of hyperspectral data for estimating the leaf water content of ten tree species under different water conditions. The three most commonly used leaf water content indicators (relative water content, equivalent water thickness, and fuel moisture content) and stomatal conductance were assessed using narrow-band indices (single band, band ratio, band subtraction, and band difference) and multivariate analyses (partial least squares regression (PLSR), support vector regression, artificial neural network, and random forest) within the 350–2500 nm spectral reflectance range. The results indicated that the best bands and band combinations were mainly concentrated in the short-wavelength infrared region, which is sensitive regarding plant water content. Compared to a single band, dual-band indices exhibited better overall performance among the four kinds of indices. Multivariate analyses are more accurate than narrow-band indices. Among these, PLSR is the most robust and can be considered the optimal technique for predicting the water content of all tree species, except for conifer species. However, accurately predicting stomatal conductance is difficult to predict using these methods. This study shows that the PLSR model can accurately estimate leaf water content in multiple tree species, and hyperspectral technology, such as hyperspectral remote sensing, has potential regarding the estimation of leaf water content.