Short-term stream flow forecasting and downstream gap infilling using machine learning techniques

[摘要] ENGLISH ABSTRACT : Stream flow is an important component in the hydrological cycle and plays avital role in many hydrological applications. Accurate stream flow forecastsmay be used for the study of various hydro-environmental aspects and mayassist in reducing the consequences of floods. The utility of time series recordsfor stream flow analyses is often dependent on continuous, uninterrupted observations.However, interruptions are often unavoidable and may negativelyimpact the sustainable management of water resources. This study proposesthe application of machine learning techniques to address these hydrologicalchallenges.The first part of this study focuses on single station short-term stream flowforecasting for river basins where historical time series data are available. Twomachine learning techniques were investigated, namely support vector regressionand multilayer perceptrons. Each model was trained on historical streamflow and precipitation data to forecast stream flow with a lead time of up toseven days. The Shoalhaven, Herbert and Adelaide rivers in Australia wereconsidered for experimentation. The predictive performance of each modelwas determined by the Pearson correlation coefficient, the root mean squarederror and the Nash-Sutcliffe efficiency, and the predictive capabilities of themodels were compared to that of a physically based stream flow forecastingmodel currently supplied by the Australian Bureau of Meteorology. Based onthe results, it was concluded that the machine learning models have the abilityto overcome certain challenges faced by physically based models and thepotential to be useful stream flow forecasting tools in river basin modelling.The second part of this study investigates the ability of support vector regressionand multilayer perceptron models to infill incomplete stream flow records.The infilling techniques relied upon data from donor stations and rain gaugeswithin close proximity to the station considered for infilling. A case study wasconducted on a channel in the Goulburn basin in Australia. The results showedthe promising role of machine learning applications for the infilling of gaps instream flow records and indicated that data from donor stations contributemore to the success of these models compared to precipitation data.

[发布机构] Stellenbosch University