Estimating catchment-wide total groundwater storage via space-time kriging provides calibration data for catchment-scale groundwater balance models

Sustainable groundwater management requires accurate and reliable prediction of long-term aquifer water balances. This can be achieved with catchment-scale groundwater balance models such as the recently developed Lumped Geohydrological Model (LGhM) (Ejaz et al., 2022, Journal of Hydrology). As a lumped model similar to hydrological rainfall-runoff models, the LGhM offers high computational speed, lower data requirements compared to spatially explicit groundwater flow models, and suitability for uncertainty analysis. Unlike rainfall-runoff models, LGhMs allow simulating total groundwater storage (TGS) by incorporating additional terms for water budget and dedicated, distributed groundwater storage boxes, inspired by the catchment's aquifer characteristics.

Calibration of LGhMs requires both river discharge data and TGS data. LGhMs have shown remarkable performance in synthetic studies (MODFLOW generated data for calibration and validation). The remaining challenge is therefore to obtain TGS data for calibration without full groundwater flow models. Geostatistical methods can help here. They can directly estimate groundwater surfaces from well-based time series, and TGS can then be obtained through spatial aggregation. In this study, we employ space-time kriging to estimate TGS and quantify associated uncertainties. To enhance TGS predictions, we integrate hydrogeological information into the kriging model. These include spatial and temporal trends and soft information inspired by hydrological ideas, such as digital elevation maps, river exchange components, aquifer confinement, and boundary conditions.

The Wairau Plain aquifer in New Zealand serves as the testing ground for this approach, where an existing MODFLOW model provides data for calibration and validation for proof of concept. Once validated, this method can be applied in regions without pre-existing groundwater flow models.