PhD defence: Facilitating and enabling large-scale, hyper-resolution, groundwater modeling with distributed-memory parallel computing

Water managers and policymakers worldwide face the major challenge of securing the availability of fresh groundwater under excessive groundwater extraction and climate change. To this end, they need future projections of groundwater resources computed with numerical groundwater models. Such computer models must have a sufficiently high spatial resolution, which often results in the problem of long runtimes and large memory requirements when applied at larger scales.

To solve this problem, this thesis investigated the method of distributed-memory parallel computing. MODFLOW, the world-wide standard for groundwater simulation, was parallelized as simulation code, for which experiments were conducted on the national computer cluster of the Netherlands. Two real-world existing groundwater models were considered, which still use uniform computational grids, as well as two new applications that use more flexible computational grids.

For existing models, the (quantitative integrated) National Hydrological Model of the Netherlands was considered and the (qualitative fresh-salt) groundwater model for the Sand Engine. For both models, large speedups were obtained (speedup of 20 and 90) for a quite limited number of processor cores (60 and 250 respectively). These large speedups show that parallelization can greatly increase the practical applicability of these existing groundwater models.

As a first new application, GLOBGM was developed, the world's first time-dependent global groundwater model with a resolution of 30 arc seconds (~ 1 km). In addition to parallel simulation, the size of this model also required parallel preprocessing of input data. With quite limited computing power (up to 400 cores with a speedup of 250), 60 years could be computed in parallel in one night. This makes that this global model can also be used by modelers who do not have access to very large computer clusters.

As a second new application, a multi-resolution groundwater model was explored for the Netherlands to incorporate regional-scale models in the National Hydrological Model. Nationwide grid refinements up to a regional-scale resolution of 12.5 m were considered, resulting in a model with more than one billion grid cells. For this model, it was shown that with quite limited computing power (up to 500 processor cores with a speedup of 325), 8 years could be computed in parallel in 2 days. This shows that these very large models are already within range with today's computers.