The estimation of recharge and boundary flux is an important problem in deterministic groundwater modelling since these quantities are often difficult to measure directly in the field. A new method (inferred recharge) is proposed for this estimation, based upon water level and transmissivity observations; it is particularly applicable to those semiarid regions where a steady state formulation can be used and recharge is confined to certain known areas. The inferred recharge method which assumes that recharge occurs in these areas only is compared with an alternative method which does not use this information, and is found to be superior. In the example of the inferred recharge method applied to the Oman coastal plain, random normal errors have been added to both water levels and transmissivities in simulation experiments to assess the effect of errors in the data. The results of the simulations have been used to test the reliability of variance estimates derived from the theory. In addition, statistical tests have been used to examine the inferred recharge results.