In many areas water allocations are determined by recharge estimates, either directly or indirectly through models. For a water balance component that can’t be directly measured it is imperative that the uncertainty in recharge estimates is quantified. It is often recommended that multiple methods should be used as a measure of uncertainty in the recharge but this will only produce multiple recharge estimates. Different methods of estimating recharge can be estimating different quantities of water. This can be a challenge if relying on different estimation methods to provide uncertainty but can also be an opportunity for constraining recharge estimates if recharge is estimated probabilistically. Baseflow in streams is an estimate of groundwater discharge and must be equal to or less than the groundwater recharge. The chloride mass balance gives an estimate of the net recharge (includes ET from groundwater) and must be equal to or greater than baseflow and also equal to or less than gross recharge (water that reaches the water table). The water table fluctuation method is estimating gross recharge and must be equal to or greater than the net recharge but also equal to or less than the excess water, which is calculated from remotely sensed ET subtracted from rainfall (excess water also contains the runoff component). If each of these water components can be estimated probabilistically then they can be jointly constrained using a rejection sampling approach. Using examples from recent projects we can show that unconstrained modelled recharge estimates are highly uncertain, the range can extend over two orders of magnitude. The chloride mass balance can be estimated probabilistically with a range over a factor of 3. By jointly constraining multiple recharge estimates the uncertainty can be reduced closer to a factor of 2.