Keywords
ENSO; Long-range Streamflow Prediction; Input Variable Selection; Data-driven Models; Hydrological Modelling
Location
Session G2: Data Mining for Environmental Sciences (s-DMTES IV)
Start Date
17-6-2014 9:00 AM
End Date
17-6-2014 10:20 AM
Abstract
Medium-to-long range streamflow predictions provide a key assistance in anticipating hydro-climatic adverse events and prompting effective adaptation measures. In this context, recent modelling efforts have been dedicated to seasonal and inter-annual predictions based on the teleconnection between at-site hydrological processes and large-scale, low-frequency climate fluctuations, such as El Niño Southern Oscillation (ENSO). This work proposes a novel procedure for first detecting the impact of ENSO on hydro-meteorological processes at the basin scale, and then quantitatively assessing the potential of ENSO indexes for building medium-to-long range streamflow prediction models. Core of this procedure is the adoption of the Iterative Input variable Selection (IIS) algorithm, which is employed to find the most relevant determinants of streamflow variability and derive predictive models based on the selected inputs. The procedure is tested on two different case studies, the Columbia River (US) and the Williams River (Australia), whose sensitivity to ENSO fluctuations has been documented in previous studies. Results show that IIS outcomes for both case studies are consistent with the results of previous analyses conducted with state-of-the-art detection methods, and that ENSO indexes can effectively be used in both regions to enhance the accuracy of streamflow prediction models.
Included in
Civil Engineering Commons, Data Storage Systems Commons, Environmental Engineering Commons, Hydraulic Engineering Commons, Other Civil and Environmental Engineering Commons
Quantifying ENSO impacts at the basin scale using the Iterative Input variable Selection algorithm
Session G2: Data Mining for Environmental Sciences (s-DMTES IV)
Medium-to-long range streamflow predictions provide a key assistance in anticipating hydro-climatic adverse events and prompting effective adaptation measures. In this context, recent modelling efforts have been dedicated to seasonal and inter-annual predictions based on the teleconnection between at-site hydrological processes and large-scale, low-frequency climate fluctuations, such as El Niño Southern Oscillation (ENSO). This work proposes a novel procedure for first detecting the impact of ENSO on hydro-meteorological processes at the basin scale, and then quantitatively assessing the potential of ENSO indexes for building medium-to-long range streamflow prediction models. Core of this procedure is the adoption of the Iterative Input variable Selection (IIS) algorithm, which is employed to find the most relevant determinants of streamflow variability and derive predictive models based on the selected inputs. The procedure is tested on two different case studies, the Columbia River (US) and the Williams River (Australia), whose sensitivity to ENSO fluctuations has been documented in previous studies. Results show that IIS outcomes for both case studies are consistent with the results of previous analyses conducted with state-of-the-art detection methods, and that ENSO indexes can effectively be used in both regions to enhance the accuracy of streamflow prediction models.
