Spectroscopic estimation of leaf water content in commercial vineyards using continuum removal and partial least squares regression

Abstract

Knowing the water status of grape strains is essential for quality wine. Traditional pressure-chamber methods for estimating water stress are laborious and destructive. We evaluated the effectiveness of spectroscopic methods to estimate water content in commercial vine for three grape varieties (Mencía, Merlot and Tempranillo) from vineyards in El Bierzo (NW Spain). We also determined the spectral range, data fitting method and degree of transformation of spectral data necessary to estimate equivalent water thickness (EWT), from the untransformed spectrum (vegetation indices and full spectra) and from the spectrum transformed using continuum removal (CR) (CR reflectance and CR-derived indices). Partial least squares regression (PLSR) and ordinary least squares regression (OLSR) were used to fit the model. The results depended on the range studied, with the best results obtained for Tempranillo. Continuous stretches of the spectrum produced more suitable EWT models than vegetation indices. The models obtained from the transformed spectrum produced more accurate estimates. The best model was obtained using PLSR in the spectrum transformed by CR in the range 1265nm to 1668nm (R2=6.75 and RMSE=0.014%). We demonstrate that a field spectroradiometer determines vine water status rapidly and without damaging sampled leaves.