Extending the Thornthwaite–Mather Water Budget Model for Surface Runoff Estimation: A Nationwide Assessment in Mainland Portugal

Quantifying water availability at the national scale remains challenging, particularly where runoff observations are scarce or inconsistent. Hydrological modelling offers a practical solution, and monthly water balance models are especially attractive because of their simplicity and modest data requirements. In this study, the performance of the Thornthwaite–Mather Water Budget (TM−WB) model was evaluated for the period of 43 hydrologic years from 1 October 1980 to 30 September 2023, based on twelve watersheds spanning the north–south climatic gradient of mainland Portugal, using precipitation and potential evapotranspiration data derived from the ERA5-Land reanalysis database. Lumped and distributed modelling approaches were tested using contrasting potential evapotranspiration formulations: the data-intensive Penman–Monteith model and the temperature-based Thornthwaite model. The suitability of grid-based available water capacity (AWC) as a substitute for the TM−WB model soil water storage conventional parameter (Smax) was also assessed. The results show that both evapotranspiration models successfully reproduce the spatial and temporal dynamics of monthly surface runoff; however, the choice of the model influences runoff magnitude at the national scale, with the Thornthwaite model generally overestimating runoff relative to Penman–Monteith. Overall, the Penman–Monteith formulation provided superior performance. Application of the TM−WB model across mainland Portugal showed its ability to reproduce the characteristic north-to-southwest-to-east decrease in runoff and yielded mean annual runoff estimates of 317 or 371 mm/year for Penman–Monteith and Thornthwaite models, respectively, consistent with the Portuguese Environment Agency’s (APA) national estimate of 358 mm/year, although this estimate is now considerably outdated. Furthermore, the AWC proved to be a robust and practical substitute for Smax, with a Monte Carlo synthetic generation procedure showing that the runoff estimates are largely insensitive to AWC uncertainty, thus enabling a spatially consistent application of the TM−WB model using globally available soil databases. These findings demonstrate the value of simple water budget models for national-scale water availability assessment, particularly in ground data-limited settings. The simplicity and consistency of the proposed methodology also enable its application at large spatial scales under climatic conditions different from those prevailing today, namely under climate change scenarios.