Research papersEstimating initial conditions for groundwater flow modeling using an adaptive inverse method

- Estimation of initial conditions for groundwater flow problems.
- Adaptive parameterization for estimation of initial conditions.
- Objective function formulation adapted for estimation of initial conditions.

Due to continuous increases in water demand, the need for seasonal forecasts of available groundwater resources becomes inevitable. Hydrogeological models might provide a valuable tool for this kind of resource management. Because predictions over short time horizons are foreseen, the reliability of model outputs depends on accurate estimates of the initial conditions (ICs), as well as the estimated parameter values, boundary conditions and forcing terms (e.g., recharge, as well as sinks and sources). Here, we provide an inverse procedure for estimating these ICs. The procedure is based on an adaptive parameterization of the ICs that limits over-parameterization and involves the minimization of an ad hoc objective function. The quasi-Newton algorithm is used for the minimization, and the gradients are computed with an adjoint-state method. Two test cases based on a real aquifer that are designed to evaluate the capability of the method were addressed. It is assumed that the boundary conditions, hydraulic parameters and forcing terms are known from an existing hydrogeological model. In both test cases, the proposed method was quite successful in estimating the ICs and predicting head values that were not used in the calibration. 50 calibrations for each test case have been performed to quantify the reliability of the predictions.