Input data resolution analysis for distributed hydrological modeling

A distributed hydrological model is often needed to analyze spatially variable hydrologic behavior. Such a model can be difficult to set up, especially for an ungauged basin, as it demands a massive amount of data. Moreover, there is an additional challenge of selecting a proper grid resolution as the grid size selection generally leads to predictive uncertainty and also directly determines the amount of work required. In this study, a distributed macro-scale hydrological model, named as the MaScOD model, is applied with a 10-min spatial resolution to the Huaihe River basin, China, to simulate discharge at Bengbu (132,350 km2) and at sub-basins at Wangjiaba (29,844 km2) and at Suiping (2093 km2). A range of input data resolutions are used, from 10 min to 2.5°, based on an experimental hydro-meteorological input data set abstracted from the GAME re-analysis data and the Hubex-IOP EEWB data. Performance of the model is evaluated by comparing observed discharge against simulated discharge for a range of IC-ratio values (the ratio between the input forcing resolution and the Catchment area). Similar results are obtained for all three catchments, despite their different sizes. It is found that improvement in distributed model performance is more pronounced below the IC-ratio 1:10, whereas the rate of improvement is negligible above the IC-ratio 1:20. The IC-ratio range 1:10–1:20 is found to be the optimum performance range considering the data and resource demands of distributed models. This may provide a preliminary criterion for selecting the scale for distributed hydrological modeling in ungauged basins.