Classification of water level fluctuation data in wells using linear regression models and genetic algorithm

Wakamatsu, Naonori*; Watanabe, Kunio*; Takeuchi, Shinji; Saegusa, Hiromitsu

A method to evaluate similarities of water level fluctuation between wells is proposed. Linear regression models with independent variable for meteorological condition such as rainfall and atmospheric pressure etc. are developed, and well similarity is estimated from model parameters (regression coefficients and model fitness) calculated by Genetic Algorism. The method was applied to the twelve wells in Tono area, central Japan. Although groundwater level fluctuation is primarily affected by rainfall and pumping conditions, different geological conditions would cause different types of water level response to the factors. Models using preceding rainfalls and atmospheric pressure and models using water level in other wells suggested that water level fluctuation data of the wells are classified into groups which reflect difference in pressure propagation for rain infiltration among the geological units.