HYGRID-M: A Grid-Based Water Balance Model for Drought Management at River Basin District Scale

Climate change is worsening the imbalance between water availability and demand, making water scarcity a critical global issue. Understanding the water balance is essential for sustainable water management. Existing grid-based and lumped water balance models provide monthly estimates of actual evapotranspiration (AET) and runoff (Q), but their assumption of uniform soil moisture storage and simplified evapotranspiration parameterizations limits their ability to capture spatial heterogeneity, reducing accuracy in drought-prone Mediterranean basins. This study introduces HYGRID-M (HYdrological GRIDed - Monthly), a grid-based water balance model with 1-km spatial grid and monthly temporal resolution for estimating the water balance components in River Basin Districts (RBDs). HYGRID-M was applied to the Southern Apennines District (Italy) over the period 2000–2023 in configurations: 1) uniform vs spatially heterogeneous soil conditions, and 2) a generalized vs a regionalized Hargreaves coefficient. This experimental setup allowed for assessing the soil variability effects and evapotranspiration parameterization on model performance. The model outputs were validated through a multi-source comparison: 1) AET estimates against three global products: GLASS, ETMonitor, and MOD16; 2) Q estimates against water balance-based observations at pilot basins and compared with the BIGBANG model. The annual average AET estimates from HYGRID-M, under both uniform and heterogeneous soil conditions, were consistent with global datasets, whereas BIGBANG produced lower values. The heterogeneous soil configuration provided the most accurate AET estimates, highlighting the importance of accounting for spatial variability in soil properties. Variations in the runoff coefficient β or in the Hargreaves parameterization (generalized vs regionalized) had only minor effects on AET performance. Q estimates showed good agreement with BIGBANG and reproduce observed flow variability at the basin scale, while β remained a key parameter for runoff accuracy across basins. Overall, HYGRID-M performed well in estimating AET and Q at both RBD and sub-unit scales, highlighting the importance of soil heterogeneity, multi-dataset AET validation, and the key role of β in simulating Q.