Quantifying uncertainty in groundwater recharge due to spatiotemporal rainfall and temporal evapotranspiration variability

• Stochastic vadose zone water budget used to estimate mean recharge and uncertainty. • Spatial rainfall variability characterized by local network of rain gauges (3 yr). • Actual evapotranspiration developed from empirical dataset for similar watersheds. • Budyko curve framework used to generate annual actual evapotranspiration. • Most of the uncertainty in recharge derives from rainfall variability not actual ET. The sustainable management of public supply wells relies to a significant degree on groundwater recharge estimates. Accuracy of these estimates will depend on the uncertainty within the largest components of the water budget, including precipitation and evapotranspiration. Quantifying this uncertainty and understanding the effect it may have on regional water balances is challenging. To examine the relative contribution of spatiotemporal rainfall variability (SRV) and annual actual evapotranspiration (AET) variability to groundwater recharge uncertainty, a method was developed to calculate a watershed stochastic vadose zone water budget within a Monte Carlo framework. The method incorporates rainfall time series generated through a semi-parametric approach that is constrained by observed local spatial rainfall correlation coefficients. Stochastic annual AET estimates are generated based on Penman-Monteith potential evapotranspiration (PET) estimates and observed variation about the Budyko curve for selected US MOPEX watersheds with PET ¯ / P ¯ ratios similar to the study area. Overland flow is estimated using streamflow records and hydrograph separation results for the study watershed. The method was applied to the Alder Creek watershed (78 km 2 ) in southern Ontario, Canada, over a 46-year period. Results suggested that 84 % of the uncertainty in recharge was related to SRV while 16 % was related to AET. This method could be used to estimate uncertainty in recharge as a context for numerical groundwater modelling and to project changes in this uncertainty based on possible climate-change induced reductions in rainfall correlation.