Lucerne (alfalfa, Medicago sativa L.) is grown extensively worldwide owing to its high forage biomass production and nutritional value. Although this crop is characterised as being tolerant to drought, its production and persistence are affected by water stress. Selection of genotypes of high yield potential and persistence after a long period of drought is a major objective for lucerne-breeding programmes in Mediterranean environments. This selection could be enhanced and accelerated by the use of physiological and productivity traits and their estimation through remote-sensing methods. A set of nine cultivars of lucerne from Australia and the USA were assessed in four locations in Mediterranean central-south Chile. Several physiological and productivity traits were evaluated: forage yield (FY), stomatal conductance (gs), water potential (WP), leaf area index (LAI), nitrogen (N) content, and isotope composition (δ¹³C and δ¹⁸O) of the dry matter. Spectral-reflectance data were used to estimate the traits through spectral-reflectance indices (SRIs) and multivariate regression methods. For the SRI-based estimations, the R² values for each assessment were <0.65. However, traits such as LAI, WP, gs, and N content showed higher R² values when data from the different assessments were combined. Regression-based estimation showed prediction power similar to or higher than the SRI-based approaches. The highest R² value was for δ¹³C (0.78), but for most traits the combination of data from different assessments led to higher trait estimation, with respective R² values for LAI, FY, WP and gs of 0.67, 0.71, 0.63 and 0.85. Among regression methods, the best estimation was achieved by using support vector machine regression. The use of spectral-reflectance data collected at field level and multivariate regression models has great potential to estimate physiological and productivity traits in lucerne under water deficit and could be useful in lucerne-breeding programmes.