The distributed parameter theory presented by McLaughlin and Wood (this issue) is used to evaluate the accuracy of a groundwater flow model. The special case investigated assumes that the model's prediction errors are due primarily to data limitations. In this case, approximate prediction error moments (mean and covariance) are obtained by solving two sets of coupled partial differential equations which have the same basic structure as the original flow equation. Solutions can be obtained with spectral, Green's function, or numerical techniques, depending on the assumptions made. Two examples are solved using a finite element approach. The first (two‐dimensional) example confirms steady state infinite domain results obtained with spectral methods. The second (three‐dimensional) example investigates the influence of spatial variability, sampling strategy, and suboptimal estimation on model prediction accuracy.
All Science Journal Classification (ASJC) codes
- Water Science and Technology