Identification of robust hyperspectral indices on forest leaf water content using PROSPECT simulated dataset and field reflectance measurements

Abstract In this study, we aim at finding efficient and robust hyperspectral indices for estimating forest leaf water content parameters (equivalent water thickness, EWT and fuel moisture content, FMC), which are useful for the understanding of terrestrial ecosystem functioning and evaluating fire risk. The most efficient hyperspectral indices have been identified (both on the context of index types and wavelength domains) using both a simulated dataset generated from the calibrated leaf reflectance model, PROSPECT, and datasets of field measured reflectance. Results indicate that high precision can be obtained using the type of vegetation index of two wavelength bands based on reflectance derivatives to estimate both parameters, with overall R 2 and RMSE of 0·60 ∼ 0·75 and 0·0009 ∼ 0·0012 g cm −2 for EWT, 0·63 ∼ 0·87 and 0·12 ∼ 0·20 g g −1 for FMC, respectively. The best indices identified in this study for vegetation water status in temperate deciduous forests were dSR (1510, 1560) for EWT and dSR (2110, 2260) for FMC, with widths of wavebands (Δλ) be variable up to 50 nm for both dSR indices. Despite the obvious discrepancies found in fit when applying the identified indices to different datasets, the indices identified in this study are applicable to various species (Dataset III), various phenological stages, different sites (Dataset I) and various leaf anatomies (Dataset II), and therefore suitable for an all inclusive wide range of application especially in temperate deciduous forests. Copyright © 2011 John Wiley & Sons, Ltd.