Remote sensing-based estimation of evaporation losses from agricultural reservoirs across Morocco

Distributed agricultural reservoirs improve irrigation flexibility, yet their collective evaporative losses remain poorly represented in water accounting. Quantifying these losses is essential for assessing the water costs of decentralized storage in water-scarce regions. Here, we integrate satellite observations, field-based water-balance estimates, and meteorological reanalysis to assess evaporation from agricultural reservoirs across Morocco’s 20 watersheds during 2020-2025. Annual open-water surfaces were mapped from Sentinel-2 imagery using complementary spectral indices, temporal recurrence, spatial filtering, and visual quality control. Three meteorological formulations were evaluated against evaporation estimates from an instrumented reservoir in Benguerir. Penman-Monteith estimates were empirically bias-corrected using the field observations and applied nationally with ERA5-Land forcing. More than 70,000 agricultural reservoirs were identified. Estimated annual evaporation losses increased from 142.1 in 2020 to 204.7 in 2025, a 44% increase. Across watersheds, evaporation volumes were strongly associated with cumulative reservoir surface area (R² = 0.96), whereas southern watersheds exhibited the highest evaporation-to-precipitation ratios. The national increase persisted across all 10,000 Monte Carlo realizations representing evaporation-model residual uncertainty. Estimates remain conditional on single-site calibration, annual water-surface mapping, and exclusion of reservoirs smaller than 1,000 m². These findings reveal the cumulative evaporative cost of distributed agricultural storage and demonstrate why farm-scale reservoirs warrant explicit representation in basin and national water budgets. The framework links local observations to national assessments to inform evaporation-loss management in data-scarce, water-stressed regions.