Determining the water status of Satsuma mandarin trees [Citrus Unshiu Marcovitch] using spectral indices and by combining hyperspectral and physiological data

This study investigated the water relations of Satsuma mandarin trees [Citrus Unshiu Marcovitch] and drought stress indicators used for irrigation scheduling in orchards namely: (1) the midday leaf water potential (MLWP), (2) midday stem (or xylem) water potential (MSWP), (3) predawn leaf water potential and (4) the leaf water content. Remote sensing spectral indices were applied to predict these indicators for trees subjected to different drought stress regimes. Continuous measurements of the MLWP and the MSWP on individual trees during cloudless days showed large fluctuations in the MLWP of up to 2.0MPa and less variation in the MSWP of ∼0.30MPa. The large variability in the MLWP was directly related to stomatal oscillations characteristic of most citrus species and the MSWP measurements were more representative of the tree water status. Spectral indices derived from canopy reflectance data of mature citrus trees in the orchard showed poor correlations with both the MSWP and MLWP. However, using spectral indices from the leaf reflectance of young potted citrus trees showed that the water index (WI) and a narrow-band spectral ratio of the reflectance at 960 and 950nm wavelengths gave the best predictions of the MSWP with R ² =0.77 and 0.79, respectively but only for severely stressed trees. All the indices failed to predict the water potentials of trees with mild or no drought stress (R ² <0.20), although the leaf water content predictions were accurate for both stressed and non-severely stressed trees. Integrating the hyperspectral estimates of leaf water content with the transpiration and soil water potential data in a simple dynamic tree-level water balance model yielded more accurate estimates of the MSWP of non-severely stressed trees. This suggests that in the absence of robust spectral indices for predicting plant water status, as is the case at present, combining hyperspectral remote sensing and in situ data in physiological models potentially yields useful information for irrigation management in orchards.