Application of Integrated SOM- and MOGA-SVM-Based Algorithms to Forecast Groundwater level in Choushui River Alluvial fan, Taiwan.

The over-pumping of groundwater and associated land subsidence in Choushui River alluvial fan, is of serious concern to Taiwan environmental issue. In order to accurately and effectively forecast the groundwater level in this study area, a two-stage approach integrating Self-Organizing Maps (SOM-), Multi-Objective Genetic Algorithm and Support Vector Machine (MOGA-SVM-based) algorithms regarding combinations of meteorological factors in a complex spatial-temporal groundwater system is developed. In the first stage, the SOM-based clustering method is applied to identify separate and meaningful groundwater zones. In the second stage, the temporal analysis model integrating MOGA with SVM is developed to optimize input combinations of meteorological factors. The performance of the MOGA-SVM-based model is better compared with the SVM-based model in the context of both the short lead-time and long lead-time forecasting. For different regions of the alluvial fan, the optimal input combinations are obviously different. It indicates that the MOGA-SVM-based model has the ability to determine the dominant input factors for different regions. To sum up, the integrated SOM- and MOGA-SVM-based algorithms shows its validity in the complex groundwater system which provides a robust prediction of groundwater level change to address water resources management.