Informing hydrological models of poorly gauged river catchments – A parameter regionalization and calibration approach

• Published in: Journal of hydrology (Amsterdam) Vol. 587; p. 124999
• Main Authors: Kittel, Cecile M.M., Arildsen, Anne L., Dybkjær, Stine, Hansen, Emilie R., Linde, Ida, Slott, Emma, Tøttrup, Christian, Bauer-Gottwein, Peter
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.08.2020
• Subjects: Calibration; Parameter regionalization; Rainfall-runoff model; Remote sensing; Ungauged river; Calibration; Ungauged river; Rainfall-runoff model; Parameter regionalization; Remote sensing
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2020.124999

•We present a parameter transfer framework for poorly gauged river catchments.•We apply the approach in three African river catchments.•We evaluate spatial patterns of hydrological processes to assess model coherence.•Parameter transfer based on similarity outperforms transfer based on proximity. Development of robust hydrologic-hydrodynamic simulation models is challenging, especially in regions, where in-situ observation data are scarce. Parameter calibration is often a necessary and data-demanding step. Moreover, good calibration performance does not guarantee predictive skill in neighboring geographic regions or new time periods and scenarios. A robust calibration strategy is necessary and must apply an informed parameter regionalization approach, which can transfer parameter values from gauged to ungauged subcatchments, whilst accounting for data availability and quality. In this study, a calibration strategy combining a parameter transfer framework based on catchment similarities with a holistic calibration against hydrological signatures is presented to obtain reliable simulations at all relevant locations within a catchment. The approach is demonstrated for three African river catchments: the Tana, the Upper Niger and the Semliki catchments. A rainfall-runoff model based on Budyko’s concept of limits with up to 11 calibration parameters and a daily simulation time step is used, coupled with a Muskingum routing scheme. Each river catchment is subdivided into so-called hydrological response units (HRU) based on climatic and physiographic characteristics. Multi-mission satellite remote sensing observations are used for the HRU classification: terrain slope derived from the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) and the publicly available European Space Agency (ESA) Climate Change Initiative (CCI) land cover map. Subcatchments within each HRU share the same parameters. The number of units is constrained by the number of in-situ stations available. The calibration strategy is a catchment-scale, multi-objective approach. The method improved the overall model performance compared to a simpler, nearest-neighbor regionalization method. In particular, spatial patterns of the hydrological response are more reasonable and performance is improved at validation stations, highlighting the importance of appropriate parameter regionalization within nested catchments. With this method, a wider range of currently available remote sensing data (elevation, precipitation, land cover, etc.) can be exploited in hydrological model development and calibration.