Features of spatiotemporal groundwater head variation using independent component analysis

The effect of external stimuli on a groundwater system can be understood by examining the features of spatiotemporal head variations. However, the head variations caused by various external stimuli are mixed signals. To identify the stimuli features of head variations, we propose a systematic approach based on independent component analysis (ICA), frequency analysis, cross-correlation analysis, well-selection strategy, and hourly average head analysis. We also removed the head variations caused by regional stimuli (e.g., rainfall and river stage) from the original head variations of all the wells to better characterize the local stimuli features (e.g., pumping and tide). In the synthetic case study, the derived independent component (IC) features are more consistent with the features of the given recharge and pumping than the features derived from principle component analysis. In a real case study, the ICs associated with regional stimuli highly correlated with field observations, and the effect of regional stimuli on the head variation of all the wells was quantified. In addition, the tide, agricultural, industrial, and spring pumping features were characterized. Therefore, the developed method can facilitate understanding of the features of the spatiotemporal head variation and quantification of the effects of external stimuli on a groundwater system.