Estimation of foliage moisture content using near infrared reflectance spectroscopy

This study explored the relationships between the initial moisture content of leaves and their spectral properties when dry. Foliage moisture content (FMC in % dry-weight (DW)) was measured in 864 samples of eight common Mediterranean tree and shrub species from two sites during the summers of 2001 and 2002. All samples were dried, ground and scanned using a near infrared reflectance spectrophotometer. There were close relationships between FMC and foliage absorbance in the 400–2500 nm spectral range. Calibrations involving search for predictive relationships between spectral data and FMC were performed using partial least squares (PLS) regression that took into account all the spectral information. Calibration on FMC in each species from each year was predictive when a sufficiently large range of FMC values were available (R 2 = 0.93–0.99, standard error of cross-validation (SECV) = 2–7%). Moreover, calibration on FMC in all species combined, for each year or for both years together, was also predictive (R 2 = 0.92–0.95, SECV = 7%). It was thus possible to estimate the initial moisture content of a fresh foliage sample from its spectral characteristics when dried, whatever the species. However, mean foliage spectral absorbance changed from 1 year to the other. Consequently, FMC in leaf samples from each year was not estimated with accuracy when using the calibration established on samples from the other year but was accurately estimated when using the calibration established on samples from both years together. These close relations between dried foliage spectral data and FMC meant that some biochemical properties of the leaves associated with the changes in leaf water status caused changes in spectral reflectance which was preserved in the dried samples.