Regional groundwater flow models play an important role in decision making regarding water resources; however, the uncertainty embedded in model parameters and model assumptions can significantly hinder the reliability of model predictions. One way to reduce this uncertainty is to collect new observation data from the field. However, determining where and when to obtain such data is not straightforward. There exist a number of data‐worth and experimental design strategies developed for this purpose. However, these studies often ignore issues related to real‐world groundwater models such as computational expense, existing observation data, high‐parameter dimension, etc. In this study, we propose a methodology, based on existing methods and software, to efficiently conduct such analyses for large‐scale, complex regional groundwater flow systems for which there is a wealth of available observation data. The method utilizes the well‐established d‐optimality criterion, and the minimax criterion for robust sampling strategies. The so‐called Null‐Space Monte Carlo method is used to reduce the computational burden associated with uncertainty quantification. And, a heuristic methodology, based on the concept of the greedy algorithm, is proposed for developing robust designs with subsets of the posterior parameter samples. The proposed methodology is tested on a synthetic regional groundwater model, and subsequently applied to an existing, complex, regional groundwater system in the Perth region of Western Australia. The results indicate that robust designs can be obtained efficiently, within reasonable computational resources, for making regional decisions regarding groundwater level sampling.

Groundwater management and allocation planning involves a rigorous assessment of the performance of operational decisions such as extraction/injection rates on community and environmental objectives. Maximizing performance through numerical optimization can be essential for high‐value resources and is often computationally infeasible due to long simulation model run times combined with nonconvex objectives and constraints. In order to mitigate these drawbacks, surrogate models can be used in place of complex models during the optimization process. There exist a number of machine learning techniques that can be used to develop a data‐driven surrogate model. However, the curse of dimensionality, common to groundwater management, limits the use of these techniques due to the necessity for large training data sets. Even though it is now possible to handle large data sets, the generation of these data sets themselves remains computationally prohibitive as they require numerous simulations to produce accurate surrogates. In this study, we integrate a dimensionality reduction method using truncated singular value decomposition to reduce the number of decision variables, thereby reducing the size of the training data set needed. Correspondingly, we demonstrate a simple technique for acquiring an approximate minimax Latin Hypercube design from within the subspace. We also implement a novel technique for adaptive resampling through particle swarm optimization in order to maintain accuracy of the surrogate model throughout the optimization process. The resulting accurate surrogate model for the Perth regional aquifer system of Western Australia runs in a matter of seconds. Adopting this approach can produce timely solutions, making formal optimization tractable for practitioners.