Effect of Distinct Evaluation Objectives on Different Precipitation Downscaling Methods and the Corresponding Potential Impacts on Catchment Runoff Modelling

As an essential tool to bridge the gap between climatic output and hydrological input, the precipitation spatial downscaling (PSD) method exhibits divergent performance in terms of different evaluation objectives. This study compared and analysed the performance of the three PSD methods using two evaluation objectives - precipitation (P) and possible effective precipitation (PEP), and evaluated the effect of different downscaled precipitation on regional flow simulation by constructing an ideal model and a real case based on the Vertical Mixed Runoff (VMR) model. Results show that when the evaluation objective changes from P to PEP, the Artificial Neural Network replaces the Weather Research & Forecasting Model as the dominant PSD model. In spatial analysis, the statistical PSD models perform significantly better when using PEP, compared to P. In temporal analysis, the PEP biases are more stable compared to the P biases from the same PSD model. The validation in the ideal case and the actual basin further proves that taking PEP as the evaluation objective can improve the reliability of the PSD method selection for hydrological research.