Assessing the value of high-resolution data and parameter transferability across temporal scales in hydrological modeling: a case study in northern China

Abstract

The temporal resolution of input data and the computational time step are crucial factors affecting the accuracy of hydrological model forecasts. This study presents a four-source hydrological model tailored to the runoff characteristics of the mountainous areas in northern China. Using this model, along with meteorological and hydrological data from seven catchments of varying sizes in northern China, we investigate the impact of different input data resolutions and computational time steps on simulation accuracy as well as the transferability of parameters across different timescales. The results show that (1) the proposed model performs well across different spatial and temporal scales, with an average Nash-Sutcliffe efficiency (NSE) for daily and hourly flow forecasts of 0.93 and 0.85, respectively. (2) For daily streamflow simulations, there was significant improvement in model performance when the data resolution was increased from 24 to 12 h; however, beyond the 12 h resolution, the improvement became negligible. For hourly streamflow simulations, the enhancement in overall flood process accuracy becomes insignificant when the resolution exceeds 6 h, although higher resolutions continue to improve the precision of peak flow simulations. (3) When the computational time step is fixed (e.g., 1 h), model parameters are transferable across different data resolutions; parameters calibrated with daily data can be used in models driven by sub-daily data. However, parameters are not transferable when the computational time step varies. Therefore, it is recommended to utilize a smaller computational time step when constructing hydrological models even in the absence of high-resolution input data. This strategy ensures that the same simulation accuracy can be achieved while preserving the transferability of model parameters, thus enhancing the robustness of the model.