In and around Otomeyama Park, Shinjuku Ward, Tokyo Metropolis, we manually measured natural spring water flow in April 2009–January 2016 and monitored groundwater level in September 2011–May 2015. We adopted a tank model to simulate the groundwater level by using precipitation data near the park. As a result, the simulation was very reasonable with a high NSE (Nash-Sutcliffe Efficiency) value of 0.993. Moreover, we tried to estimate the recharge area of the spring water by using Radar-raingauge Analyzed Precipitation on the assumption that the tank model can be adapted to wide range. First, we selected five candidates for the recharge area around the park, and we got the average precipitation for each of them. Then we calculated groundwater levels by inputting the five average precipitation values into the tank model. For the calculation, the parameters of the tank model were fixed. We compared calculated and measured groundwater levels, and found the area was about 10 km² for which the NSE was the highest and RMSE (Root Mean Squared Error) was the lowest. Also, we analyzed the water quality of the spring. Observed SiO₂ concentration and electric conductivity showed only slight differences on heavy-rainfall days and sunny days. This suggests that relatively old water stored was pushed out by newly recharged water during heavy rainfall events.