A system identification methodology for distributed‐parameter model building is described and compared with other methodologies for the modeling of groundwater reservoirs. The methodology is then applied to the analysis of New Zealand's Hutt Valley‐Port Nicholson groundwater reservoir data. It is established that there are too few measurement wells to allow for the identification of a model suitable for forecasting reservoir performance. Application of the methodology does, however, indicate where additional wells should be drilled, so that such a model could be identified. Certain field parameters are identified to within 100%, and this accuracy is found to be acceptable, provided all the important features of the reservoir are represented in the mo