Infiltration coefficient, percolation rate and depth-dependent specific yields estimated from 1.5 years of absolute gravity observations near a recharge lake in Pingtung, Taiwan

Abstract An efficient underground reservoir management requires reliable hydrogeological parameters, such as infiltration coefficient, percolation rate and specific yield (Sy) values. The cross-well pumping test is a well-known method for determining Sy, but different unsaturated flow models can lead to various Sy results. Heterogeneity in unsaturated and saturated zones can restrict well-dependent methods in estimating these parameters on a regional scale, for which an alternative, gravity-based method is attempted in this study. We established an absolute gravity site at the Wanlong Primary school (WLPS) near a recharge lake (an artificial lake) in the Pingtung Plain and collected 1.5 years of gravity data in 2019–2020. Around a monsoonal rain event in late May 2020, gravity values underwent three stages of changes. The first stage experienced a rapid gravity rise of 18 μgal from infiltrated rain, followed by stable gravity values in the second stage when the infiltrated rain recharged the aquifer, and the third stage showed a persistent gravity increase with rising groundwater level. The gravity-based estimate of infiltration coefficient is 0.84 ± 0.14, and the percolation rate is about 15 m/day, suggesting that WLPS is over a high infiltration region. A depth-dependent Sy profile up to 30 m depth was constructed using gravity and groundwater data. In the dry season, it was found that Sy values are higher at the aquifer depths of 22–28 m than in the shallow part. However, some Sy values determined in the wet seasons are overestimated due to high soil water content (SWC) changes, which induced particularly large gravity changes in the 30-m unsaturated zone of WLPS when the wet season started. This study confirms the advantage of using high precise gravity measurements (at one μgal level) and hydrology records for regional hydrogeological parameter estimation.