Quantifying leaf-scale variations in water absorption in lettuce from hyperspectral imagery: a laboratory study with implications for measuring leaf water content in the context of precision agriculture

Abstract

Variations in water absorption across lettuce leaves (Latuca sativa L. var. longifolia) were quantified from hyperspectral imagery acquired in the laboratory using selected spectral indices, specifically, the Moisture Stress Index (MSI), the Normalised Difference Water Index (NDWI) and the intensity of specific water absorptions at 970 nm (IA970), 1170 nm (IA1170) and 1775 nm (IA1775). Absorption was separately quantified for the midrib, the green parts of the leaves and for whole leaves. Indices were non-linearly related to water content expressed per weight of wet plant material (g g−1) but linearly to water content per unit area of leaf (g cm−2). Indices were weakly correlated with water content in the stem but strongly correlated with water in the green parts of leaves and in whole leaves. Water content in whole leaves was significantly underestimated (P < 0.01) when it was predicted from a model developed for the green parts of leaves, indicating that water content must be derived from the same leaf component used to derive the predictive model. Some indices (NDWI, MSI, IA1170) highlighted intricate reticulated patterns of water absorption across the leaves but these were poorly defined by other indices (IA970, IA1775). Indices extracted from the leaf along transverse and longitudinal transects were qualitatively similar but quantitative analysis indicated that they were significantly different (P < 0.05). The principal contribution of this study is that it highlights the implications of quantifying leaf water content from hyperspectral imagery acquired at spatial resolutions great enough to resolve individual leaf components.