Best hyperspectral indices for tracing leaf water status as determined from leaf dehydration experiments

Leaf water status information is highly needed for monitoring plant physiological processes and assessing drought stress. Retrieval of leaf water status based on hyperspectral indices has been shown to be easy and rapid. However, a universal index that is applicable to various plants remains a considerable challenge, primarily due to the limited range of field-measured datasets. In this study, a leaf dehydration experiment was designed to obtain a relatively comprehensive dataset with ranges that are difficult to obtain in field measurements. The relative water content (RWC) and equivalent water thickness (EWT) were chosen as the surrogates of leaf water status. Moreover, five common types of hyperspectral indices including: single reflectance (R), wavelength difference (D), simple ratio (SR), normalized ratio (ND) and double difference (DDn) were applied to determine the best indices. The results indicate that values of original reflectance, reflectance difference and reflectance sensitivity increased significantly, particularly within the 350-700 nm and 1300-2500 nm domains, with a decrease in leaf water. The identified best indices for RWC and EWT, when all the species were considered together, were the first derivative reflectance based ND type index of dND (1415, 1530) and SR type index of dSR (1530, 1895), with R2 values of 0.95 (p < 0.001) and 0.97 (p < 0.001), respectively, better than previously published indices. Even so, different best indices for different species were identified, most probably due to the differences in leaf anatomy and physiological processes during leaf dehydration. Although more plant species and field-measured datasets are still needed in future studies, the recommend indices based on derivative spectra provide a means to monitor drought-induced plant mortality in temperate climate regions.