<abstract xmlns="http://www.w3.org/1999/xhtml"><p> It is widely acknowledged that in the hydrological and meteorological communities, there is a continuing need to improve the quality of quantitative rainfall and river flow forecasts. A hybrid (combined deterministic-stochastic) modelling approach is proposed here that combines the advantages offered by modelling the system dynamics with a deterministic model and a deterministic forecasting error series with a data-driven model in parallel. Since the processes to be modelled are generally nonlinear and the model error series may exhibit nonstationarity and heteroscedasticity, GARCH-type nonlinear time series models are considered here. The fitting, forecasting and simulation performance of such models have to be explored on a case-by-case basis. The goal of this paper is to test and develop an appropriate methodology for model fitting and forecasting applicable for daily river discharge forecast error data from the GARCH family of time series models. We concentrated on verifying whether the use of a GARCH-type model is suitable for modelling and forecasting a hydrological model error time series on the Hron and Morava Rivers in Slovakia. For this purpose we verified the presence of heteroscedasticity in the simulation error series of the KLN multilinear flow routing model; then we fitted the GARCH-type models to the data and compared their fit with that of an ARMA - type model. We produced one-stepahead forecasts from the fitted models and again provided comparisons of the model’s performance.</p></abstract>

It is widely acknowledged that in the hydrological and meteorological communities, there is a continuing need to improve the quality of quantitative rainfall and river flow forecasts. A hybrid (combined deterministic-stochastic) modelling approach is proposed here that combines the advantages offered by modelling the system dynamics with a deterministic model and a deterministic forecasting error series with a data-driven model in parallel. Since the processes to be modelled are generally nonlinear and the model error series may exhibit nonstationarity and heteroscedasticity, GARCH-type nonlinear time series models are considered here. The fitting, forecasting and simulation performance of such models have to be explored on a case-by-case basis. The goal of this paper is to test and develop an appropriate methodology for model fitting and forecasting applicable for daily river discharge forecast error data from the GARCH family of time series models. We concentrated on verifying whether the use of a GARCH-type model is suitable for modelling and forecasting a hydrological model error time series on the Hron and Morava Rivers in Slovakia. For this purpose we verified the presence of heteroscedasticity in the simulation error series of the KLN multilinear flow routing model; then we fitted the GARCH-type models to the data and compared their fit with that of an ARMA - type model. We produced one-stepahead forecasts from the fitted models and again provided comparisons of the model’s performance.

Hybrid Forecasting of Daily River Discharges Considering Autoregressive Heteroscedasticity

Slovak Journal of Civil Engineering

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

It is widely acknowledged that in the hydrological and meteorological communities, there is a continuing need to improve the quality of quantitative rainfall...

Hybrid Forecasting of Daily River Discharges Considering Autoregressive Heteroscedasticity

Published Online: Jul 01, 2017

Page range: 39 - 48

DOI: https://doi.org/10.1515/sjce-2017-0011

Keywords
Morava and Hron Rivers, KLN multilinear flow routing model, Simulation error, GARCH model.

© Faculty of Civil Engineering, Slovak University of Technology

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Hybrid Forecasting of Daily River Discharges Considering Autoregressive Heteroscedasticity

Published Online: Jul 01, 2017

Page range: 39 - 48

DOI: https://doi.org/10.1515/sjce-2017-0011

KeywordsMorava and Hron Rivers, KLN multilinear flow routing model, Simulation error, GARCH model.

© Faculty of Civil Engineering, Slovak University of Technology

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Keywords
Morava and Hron Rivers, KLN multilinear flow routing model, Simulation error, GARCH model.