Assessment of hydrological model performance in Morocco in relation to model structure and catchment characteristics

Abstract

Study region: 30 catchments in Morocco. Study focus: We assessed the KGE performance of eight monthly lumped hydrological models forced by ground-based rainfall observations. We then examined how the performance relates to model complexity and structure, applied exploratory correlation analysis to identify the catchment features (over 200 features were considered) most significantly related to model performance, and investigated how these models respond to three rainfall forcings (ERA5, CHIRPS, and PERSIANN-CDR). New hydrological insights for the region: The findings indicate that no hydrological model outperformed (or underperformed) consistently across all the catchments and that model performance depends more on model structure and hydro-climatic characteristics, particularly those related to calibration and calibration-validation data difference, than on model complexity and non-hydro-climatic features. The linearity between rainfall and runoff was the primary feature influencing model performance. Additionally, besides the expected improvement of model performance when forced with richer rainfall and runoff calibration data in terms of wet and dry years, our results show that this holds true even if the calibration data is only relatively richer than the validation data and that dry periods are more beneficial to model performance than wet periods. Lastly, all the models responded similarly to the different rainfall inputs; each model performed better when using ERA5 than when using CHIRPS and underperformed when using PERSIANN-CDR. The metric that best explained this similarity was the Pearson correlation coefficient between the precipitation products and observed runoff.

Keywords

• Large-sample hydrology
• Rainfall-runoff model
• Hydrological modeling
• Model complexity
• Model intercomparison
• Precipitation products