Inter-comparison of lumped hydrological models in data-scarce watersheds using different precipitation forcing data sets: Case study of Northern Ontario, Canada

Abstract

Study region: Big East River and Black River watersheds in Northern Ontario, Canada as snow-dominated, data-poor case studies. Study focus: In this study, seven lumped conceptual models were thoroughly compared in order to determine the best performing model for reproducing different components of the hydrograph, including low and high flows in data-poor catchments. All models were calibrated using five various objective functions for reducing the effects of calibration process on models’ performance. Additionally, the effects of precipitation, an important factor, particularly in data-scarce regions, were assessed by comparing two precipitation input scenarios: (1) low-density ground-based gauge data, and (2) the Canadian Precipitation Analysis (CaPA) data. The final goal of this study was to compare the effects of using either the Degree-Day or SNOW17 snowmelt estimation methods on the accuracy of streamflow simulation. New hydrological insights: The results indicate that, in general, MACHBV is the best performing model at simulating daily streamflow in a data-poor watershed, and both SACSMA and GR4J can provide competitive results. Additionally, MACHBV and GR4J are superior to the other conceptual models regarding high flow simulation. Moreover, it was found that incorporating the more complex SNOW17 snowmelt estimation method did not always enhance the performance of the hydrologic models. Finally, the results also confirmed the reliability of the CaPA data as an alternative forcing precipitation in the case of low data availability.

Keywords

• Model inter-comparison
• Simple conceptual models
• Data scarce regions
• Canadian Precipitation Analysis (CaPA)
• Snowmelt estimation
• Canada