HR: 1340h
AN: H23F-1673    [Abstracts]
TI: Forecasting Groundwater Level Fluctuations In a Costal Aquifer Using Support Vector Machine
AU: * Yoon, H
EM: oolahee1@snu.ac.kr
AF: School of Earth and Environmental Sciences, Seoul National University, Rm 505 Bld 25-1 Seoul National University, Seoul, 151-741, Korea, Republic of
AU: Jun, S
EM: skybeast@hanmail.net
AF: School of Earth and Environmental Sciences, Seoul National University, Rm 505 Bld 25-1 Seoul National University, Seoul, 151-741, Korea, Republic of
AU: Lee, K
EM: kklee@snu.ac.kr
AF: School of Earth and Environmental Sciences, Seoul National University, Rm 505 Bld 25-1 Seoul National University, Seoul, 151-741, Korea, Republic of
AB: Precise prediction of groundwater level fluctuations has been an important and challenging topic in hydrology. In coastal aquifer the groundwater level is influenced by a tide level as well as a precipitation, which renders the prediction more difficult. Support vector machine (SVM), a novel data-driven and artificial intelligence-based model, shows remarkable prediction performances for non-linear systems in many disciplines. Recently, researches using SVM for the prediction of water resource variables are increasing. In this study we developed a SVM based time series model, then we applied it to forecasting the groundwater level at the coastal aquifer of Mangsang in the western side of East Sea, Korea. We especially focused upon assessing the influence of input vector organizations on model performances. The results show that input vectors including past data of the groundwater level raise the model performance notably, with correlation coefficient over 0.9 in this case. This model performance is comparable or even superior to that of artificial neural network models or linear time series models. Sensitivity analyses for input vector sizes and prediction lag times emphasize that the input vector organization is necessary for a SVM time series model application to hydrologic fields.
DE: 1829 Groundwater hydrology
DE: 1869 Stochastic hydrology
SC: Hydrology [H]
MN: 2007 Fall Meeting