Integrating remote sensing and ion balance to predict yield losses under saline irrigation in rice

Abstract

Rice cultivation in the Guadalquivir River marshes of southern Spain is increasingly constrained by irrigation water salinity, exacerbated by drought and seawater intrusion. This study assessed the agronomic, physiological, and spectral responses of indica and japonica cultivars under commercial farming conditions across a natural salinity gradient (mean electrical conductivity of irrigation water ranging from 3.1 to 6.9 dS m⁻¹). Field measurements included yield, growth traits, and leaf ion concentrations, complemented with Sentinel-2 vegetation indices and integrated using Generalized Additive Models (GAMs). Rice yield declined steeply with salinity, with up to 70 % losses between 3 and 7 dS m⁻¹ . Rice grown in medium-salinity fields maintained Na/K ratios comparable to low-salinity fields, suggesting that compensatory K⁺ uptake mitigated yield penalties. By contrast, high salinity led to marked ionic imbalance, particularly in japonica cultivars, which consistently exhibited higher Na/K ratios than indica. Spectral data revealed that broad-band greenness indices (NDVI, GNDVI, EVI, SAVI, NDRE) effectively captured early osmotic effects (<60 DAS), while MCARI uniquely detected late-stage ionic stress during reproductive phases. GAM analysis confirmed two phenological windows of higher sensitivity to salinity—vegetative establishment and reproductive development—while demonstrating the predictive utility of combined physiological and spectral indicators (LOOCV R² = 0.867). These findings underscore the need for growth phase-specific management and the potential of integrating physiological and remote sensing data to support adaptation strategies in Mediterranean rice systems.