Conditional parametric models for storm sewer runoff

Harpa Jonsdottir*, H. Aa Nielsen, H. Madsen, J. Eliasson, O. P. Palsson, M. K. Nielsen

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

7 Citations (Scopus)

Abstract

The method of conditional parametric modeling is introduced for flow prediction in a sewage system. It is a well-known fact that in hydrological modeling the response (runoff) to input (precipitation) varies depending on soil moisture and several other factors. Consequently, nonlinear input-output models are needed. The model formulation described in this paper is similar to the traditional linear models like final impulse response (FIR) and autoregressive exogenous (ARX) except that the parameters vary as a function of some external variables. The parameter variation is modeled by local lines, using kernels for local linear regression. As such, the method might be referred to as a nearest neighbor method. The results achieved in this study were compared to results from the conventional linear methods, FIR and ARX. The increase in the coefficient of determination is substantial. Furthermore, the new approach conserves the mass balance better. Hence this new approach looks promising for various hydrological models and analysis.

Original languageEnglish
Article numberW05443
JournalWater Resources Research
Volume43
Issue number5
DOIs
Publication statusPublished - May 2007

Fingerprint

Dive into the research topics of 'Conditional parametric models for storm sewer runoff'. Together they form a unique fingerprint.

Cite this