At the underground station of Matsushiro, a decrease in gravity is observed when it rains. This is a consequence of the Newtonian attraction exerted by the water that lies above the gravimeter. After the rainfall, gravity returns to the original level at an approximately constant rate. Imanishi et al. (2006) developed a simple numerical model to describe this short-term effect of underground water on gravity. This was a purely empirical model, and it has been left unexplained why gravity changes linearly with time and what determines its rate. In this paper, we analyze the data of drip rate inside the tunnel, as well as soil moisture at the surface of the hill, in order to elucidate the relation between the different hydrological processes taking place in the local system of Matsushiro. As a result, we reach a physical picture on the budget of the groundwater that (i) the soil layer on the top of the hill can be interpreted as a reservoir of water, (ii) the bedrock of the hill acts as a conduit of water, (iii) the mass of water in the soil, which causes gravity changes, is discharged downward at a quasi-steady rate, and (iv) evapotranspiration has a relatively minor effect at average. This gives a qualitative explanation to the linear feature of the gravity recovery. Quantitatively, the observed gravity rate is predicted to a good precision by estimating the vertical flow rate of water inside the hill. The estimated flow rate is discussed from a hydrological viewpoint in terms of viscous flows in tubes or fractures. © 2012 Elsevier Ltd.