Localised impact of different meteorological forcing data and scales in high-resolution hydrological modelling

Hydrological modelling is increasingly adopting hyper-resolution approaches in an attempt to provide more accurate and high-resolution representations of local hydrological dynamics. For arid regions and areas expected to be disproportionately affected by a changing climate this may prove particularly useful to make highly localised water allocation decisions. However, the benefits of increasing spatial resolution remain uncertain, as hydrological models are highly sensitive to meteorological forcing data, which constitute one of the main sources of model uncertainty.  

A fundamental question is whether finer resolution forcing data meaningfully improve model performance and reliability, or instead amplify uncertainty at a greater computational cost. Addressing this question is critical, as hydrological simulations and future scenario development increasingly form the basis for infrastructure planning and water-related decision-making that will impact land use policies and communities' livelihoods.  

Here, we explored the effect of four different resolution meteorological forcing data on the performance of the 1km grid cell CWatM hydrological model for the Ebro basin (approx. 80,000 km2) in Spain. The meteorological datasets have a resolution ranging from 1 arcmin x 1 arcmin (EMO-1, approx. 1.4 km x 1.4 km at 41º latitude), over 5 km x 5 km (EMO-5, approx. 3.6 arcmin x 3.6 arcmin at 41º latitude), to 0.1º (MSWX, E-OBS, approx. 8.4 km x 8.4 km).   While traditional model performance evaluations often only assess streamflow performance, we also assessed simulated reservoir volume and inflow results and modelled irrigation amounts throughout the basin. For this, we used traditional validation data obtained through the gauging network of the  Sistema Automático de Información Hidrológica (SAIH), as well as irrigation amounts estimated through satellite imagery, hereby testing a novel method for validation that could also be utilised in less well-gauged basins.  

The results provide twofold insights and contributions to the current debate on hyper-resolution hydrological modelling. On the one hand, this research addresses the perceived necessity of high-resolution data to produce reliable results for future scenario development, especially since up-to-date high-resolution climate projection data at this level of detail are not widely available. On the other hand, while higher resolution meteorological forcing data can provide highly localised information on water allocation impacts, utilisation of hyper-resolution data must also be seen considering practicality and computational effort. Hydrological modelling results in highly gauged basins can reliably be validated with a wealth of data, whereas remote sensing products provide a feasible alternative for high-resolution hydrological modelling as validation tools in less well-gauged basins. These insights can enable water-allocation decision makers locate areas of highest impact per allocated water unit finding trade-offs for maintaining the local water cycle.