Abstract

Hydrological connections between aquifers and boreal mires need to be better understood for protection of this type of wetland. Three-dimensional (3D) models have so far been sparsely used for such systems. This study investigated the effect of parameterisation with global sensitivity analysis on groundwater-surface water (GW-SW) interactions in a boreal esker-aapa mire system. Sensitivity analysis by the elementary effect (Morris) method was applied to a 3D steady-state hydrological model built with the fully-integrated HydroGeoSphere code. Parameter sensitivity with respect to various model outputs was explored, providing comprehensive insights into the most hydrologically relevant parameters. The results indicated that depending on model outputs the most influential model parameters varied. They also revealed existence of feedback in terms of interdependence of parameters between the esker aquifer and surrounding aapa mires. The properties of the mire (or peatland) landscape affected groundwater levels in the unconfined aquifer and, conversely, esker-mire interactions depended on esker hydraulic characteristics. This implies that accurate representation of both systems is required and that reliable determination of GW-SW interactions may be impeded by parameter interactions. In this study, the van Genuchten functions were used to represent hydraulic properties of unsaturated flow domain and the results revealed that formal sensitivity analysis methods based on random sampling might not be appropriate for evaluating parametric sensitivity. If the van Genuchten parameters ranges are large, randomly sampled values may not always produce physically realistic water retention characteristics. Overall, our results demonstrate that the computationally efficient elementary effect method is a suitable tool for investigating the parametric sensitivity of integrated models, enhancing modelling of boreal groundwater-dependent ecosystems.