Stormwater infiltration and focused groundwater recharge in a rain garden: Numerical modeling and field experiment

Traditional stormwater management, which relies heavily on detention storage, does not mitigate groundwater depletion resulting from groundwater pumping and loss of groundwater recharge. There has been an increasing interest in the use of alternative practices, such as rain gardens, that encourage infiltration of stormwater. A rain garden is a landscaped garden in a shallow depression of relatively small area that receives the stormwater from an impervious surface. Rain gardens can be particularly effective when infiltration is focused to maximize recharge. A numerical model was developed (RECHARGE) that can be applied in the design and evaluation of rain gardens. To continuously simulate recharge, runoff and evapotranspiration during wet and dry periods, the Richards Equation was coupled with a surface water balance in the model. Water flow through the rain garden soil is modeled over three layers: a root zone, a middle storage layer of high conductivity, and a lower layer that represents the subsoil at the site. For the climate of southern Wisconsin, simulation results show that very high recharge rates are possible in the April-September rainy season (over twice the natural annual rates). Additionally, a rain garden with an area of approximately 10% of the contributing impervious area maximizes groundwater recharge; the recharge benefit decreases for larger rain garden areas due to evapotranspiration. Increasing the depression depth increases recharge but also increases ponding times, potentially affecting plant survival. Model results indicate that the feasibility of a rain garden depends heavily on the saturated hydraulic conductivity of the underlying soil, which can override the effect of a deeper soil storage zone. We installed an experimental rain garden and validated the RECHARGE model with three experimental runs resembling typical recharge events. We also developed a simpler Green-Ampt model, which yields similar results to the Richards Equation model, with an order of magnitude less computation time.