Title: Influence of calibration data on hydrological model prediction

Authors: Shailesh Kumar Singh; Jiaying Liang; Alejandro Chamorro

Addresses: National Institute of Water and Atmospheric Research, 10 Kyle street, Riccarton, Christchurch, New Zealand ' University of Stuttgart, 61 paffenwaldring, 70569 Stuttgart, Germany ' Institute of Landscape Ecology, Justus Liebig University, Heinrich-Buff-Ring 26, 35392, Giessen, Germany

Abstract: Hydrological model performance depends on the information content and hydrological variability in the calibration data. The hypothesis for this study was that a priori predictability of the model performance can be made based on knowledge about the data used for calibration. If the validation time period has similar hydrological variability to the calibration period where the model is expected to perform better and it is termed as the 'interpolation case', otherwise it is 'extrapolation', where a better performance cannot be expected. In this study a geometrical property of data was used to discriminate between the cases of interpolation or extrapolation. The methodology was demonstrated using WaSiM-ETH model in Rems catchment. The results show that the relative residual is very high at extrapolation case as compared to the interpolation case. The result of this study can be used to make a priori estimate of the error in prediction without running the model.

Keywords: hydrological model; interpolation; extrapolation; data depth function; WaSiM-ETH.

DOI: 10.1504/IJHST.2018.093595

International Journal of Hydrology Science and Technology, 2018 Vol.8 No.3, pp.244 - 257

Received: 17 Jan 2017
Accepted: 26 Mar 2017
Published online: 30 Jul 2018 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article