This paper presents the regionalisation of the three parameter event-based PQRUT model, which is used for design flood analyses. The PQRUT model is used for the analysis of peak flows for which a sub-daily temporal resolution is required. The availability of high-resolution discharge data and disaggregated precipitation data have made it possible to re-evaluate the regional regression equations currently in use. We also assess whether the model parameters show spatial dependency. Event-based calibration was performed for the 45 highest flood events for each of 55 selected catchments across Norway, representing peak flows generated predominantly by rainfall. Due to the geographical heterogeneity of most areas in Norway, a statistically significant homogeneous region was only identified for catchments in southeastern Norway. Multiple linear regression and weighted regression were, therefore, used to develop a single set of equations, applicable to the entire country. The results for the weighted regression indicate a decrease in the median Kling–Gupta efficiency from 0.64 to 0.51 for calibration and regionalisation, respectively. The results also suggest that regression methods may perform better than methods based on spatial proximity in regions with varying topography when a parsimonious model is used.
- PQRUT model
- spatial autocorrelation
- ungauged catchments
- First received 26 March 2015.
- Accepted in revised form 19 December 2015.
- © IWA Publishing 2016