Abstract
A new generalised method is presented enabling the use of multiple donor sites when predicting an index flood variable in an ungauged catchment using a hydrological regression model. The method is developed from the premise of having an index flood prediction with minimum variance, which results in a set of optimal weights assigned to each donor site. In the model framework presented here, the weights are determined by the geographical distance between the centroids of the catchments draining to the subject site and the donor sites. The new method was applied to a case study in the United Kingdom using annual maximum series of peak flow from 602 catchments. Results show that the prediction error of the index flood is reduced by using donor sites until a minimum of six donors have been included, after which no or marginal improvements in prediction accuracy are observed. A comparison of these results is made with a variant of the method where donor sites are selected based on connectivity with the subject site through the river network. The results show that only a marginal improvement is obtained by explicitly considering the network structure over spatial proximity. The evaluation is carried out based on a new performance measure that accounts for the sampling variability of the index flood estimates at each site. Other results compare the benefits obtained by adding relevant catchment descriptors to a simple regression model with those obtained by transferring information from local donor sites.
Original language | English |
---|---|
Pages (from-to) | 6646–6657 |
Number of pages | 12 |
Journal | Water Resources Research |
Volume | 50 |
Issue number | 8 |
Early online date | 24 Jul 2014 |
DOIs | |
Publication status | Published - 18 Aug 2014 |
Keywords
- Flood
- Extreme events
- stochastic hydrology
Fingerprint
Dive into the research topics of 'Using multiple donor sites for enhanced flood estimation in ungauged catchments'. Together they form a unique fingerprint.Profiles
-
Thomas Kjeldsen
- Department of Architecture & Civil Engineering - Reader
- Water Innovation and Research Centre (WIRC)
- EPSRC Centre for Doctoral Training in Statistical Applied Mathematics (SAMBa)
- Institute for Mathematical Innovation (IMI)
- Centre for Regenerative Design & Engineering for a Net Positive World (RENEW)
- Centre for Climate Adaptation & Environment Research (CAER)
Person: Research & Teaching, Core staff, Affiliate staff