QUANTIFYING UNCERTAINTY IN GROUNDWATER RECHARGE USING A STOCHASTIC VADOSE ZONE WATER BUDGET WITH SYNTHETIC RAINFALL AND ACTUAL EVAPOTRANSPIRATION

Water budgets are commonly used to estimate watershed groundwater recharge, but the uncertainty in recharge is less frequently estimated and likely derives mostly from the largest budget components. Uncertainty in precipitation (P) derives in part from the spatiotemporal correlation in rainfall, which may change as rainfall intensity shifts due to climate change. Uncertainty in actual evapotranspiration (AET) is difficult to quantify at the sub-annual time scale but can be estimated at the annual time scale based on empirical patterns relating watershed P, potential evapotranspiration (PET), and AET. In this study, a vadose zone (VZ) water budget over several decades is used to calculate thousands of estimates of average recharge for a typical 80 km² watershed in southern Ontario. Recharge is calculated as net infiltration minus AET, where net infiltration is the difference between P and overland flow. It is assumed that there is no net VZ storage change over this time period. Daily rainfall spatial correlation is quantified based on field data and used to constrain synthetic rainfall estimates, while synthetic annual snowfall estimates are generated based on a local distribution of observations. Synthetic annual AET estimates are derived from a database of water budgets and the Budyko curve. Overland flow as a proportion of total streamflow is estimated using hydrograph separation results and used to constrain synthetic estimates. The results indicate that uncertainty in recharge is around ±10% for the study watershed based on these factors, and the method is useful for estimating the range of plausible recharge values for a watershed.