CMStatistics 2016: Start Registration
View Submission - CFE
A0365
Title: Improving the UK index flood estimation by catchment characteristics with additive and spatial regression analyses Authors:  Marinah Muhammad - University of Southampton (United Kingdom) [presenting]
Zudi Lu - University of Southampton (United Kingdom)
Abstract: Flood modelling at ungauged catchment has always been a challenging problem. Regionalization is an important procedure widely used with the assumption flood peak can be explained by catchment characteristics. The Flood Estimation Handbook (FEH) model for index flood is a well-established model of this kind in the UK. There are however unexpected or unknown features in flood dataset that need to be taken account of. Two research questions are to be investigated: (1) Is the FEH model reliable in characterising the nonlinear effects of the catchment characteristics for the UK flooding? (2) Could we improve the FEH model with a better accuracy in index flood estimation?Potential nonlinear effects of the covariates are incorporated into the FEH model by an additive regression analysis. Moreover, spatial autocorrelation is examined on the regression residuals, and potential spatial neighbouring effect is incorporated into the FEH regression by spatial econometric models. The results show that: (i) our additive model analysis confirms the nonlinear impact of the catchment characteristics identified by the FEH model is reliable; (ii) the identified statistical significance of spatial autocorrelation indicates the spatial neighbouring effect that is not taken account of in the FEH regression should be well considered; (iii) it has been detected,the FEH regression with spatial error model is a most appropriate alternative to the FEH model, helping to improve index flood estimation.