12,000

We have over 12,000 students, from over 100 countries, within one of the safest campuses in the UK

93%

93% of Lancaster students go into work or further study within six months of graduating

Home > Research > Publications & Outputs > Hydrological model calibration using a short pe...
View graph of relations

« Back

Hydrological model calibration using a short period of observations

Research output: Contribution to journalJournal article

Published

Journal publication date15/03/2012
JournalHydrological Processes
Journal number6
Volume26
Number of pages10
Pages883-892
Original languageEnglish

Abstract

To make successful predictions in ungauged basins (PUB) exploring how to facilitate the identification of catchment characteristics from readily available information is as important as improving the theoretical representation of different hydrological processes. Today in most parts of the world, a significant amount of spatial information from GIS overlays and remote sensing imagery is available. However the hydrological information derived from such sources cannot be the complete answer to the PUB problem. Thus, to implement a field measurement program and measure the discharge hydrograph during a short observation period might be a cost-effective strategy to improve predictive capability. In this study, we explore the most effective way to extract the information content from a limited period of observation. At first, the effectiveness of the phase correction technique using Dynamic Time Warping is demonstrated with long periods of observations to reduce the adverse influence of timing error, which is the discrepancy between the time of the real peak discharge and the observed one which may be disinformative for model identification. Then, by using the Monte Carlo approach with short periods of data, both the phase correction technique and the ensemble mean approach, in which a weighted ensemble mean of simulations using acceptable parameter sets is used as a prediction, are compared with conventional methods. We found that the ensemble mean approach without the phase correction technique could give the most robust model identification when the observation period is shorter than 2?weeks. We also found that the phase correction technique could work well with observation periods of 1?year or longer but was not always effective with shorter periods. These results were strongly dependent on the application catchment. Copyright (C) 2011 John Wiley & Sons, Ltd.