A Magnetotelluric Investigation of Hawaiian Groundwater Systems in Kona, Hawai‘i Island, Hawai‘i, including Uncertainty Quantification

Abstract

We investigate the spatial distribution of groundwater at high elevations in Hawaiʻi Island's Keauhou aquifer system, where a sparse distribution of water wells indicates anomalously high head levels that are likely caused by a geological structure of unknown nature. To study this system, we have conducted a MT survey on a transect climbing the southwestern flank of the Hualālai shield volcano. We carry-out a 2D deterministic inversion using MARE2DEM (a freely available MT modeling program), producing an image of the bulk electrical resistivity distribution in the subsurface. We explore the use of 1D, station-by-station, Bayesian inversions (using our own software), in an attempt to supplement the limited amount of water wells at higher elevations, by creating `virtual water wells' from MT stations that can be used to infer e.g.~head-level information. Even though we can infer information about the subsurface groundwater distribution with reasonable accuracy, we infer that the vertical resolution of the MT data is most likely insufficient to be used as actual virtual water well head-level data. However, the MT results are anticipated to be useful for roughly validating and constraining hydrological models in terms of e.g.~different hydrogeological regimes. Comparing the 1D inversion results with the 2D results, we find that the 1D inversion results match the 2D results reasonably well for most stations, especially until the depth at which seawater associated resistivities are encountered. Using the uncertainty quantification of the Bayesian inversions, we determine that depths of up to approximately 1 km beneath mean sea level are well constrained by the data. To first order, the 2D image exhibits a layered resistivity structure. Based on literature values for the resistivities of Hawaiian basalts under various saturation conditions, these layers are consistent with (in order of increasing depth), unsaturated Hawaiian basalt, freshwater-saturated Hawaiian basalt, and seawater-saturated Hawaiian basalt. A large head gradient delineating inland high-level heads from coastal low-level heads is also evident in the image as an apparent discontinuity in the freshwater valued resistivities. Resistivities associated with seawater-saturated basalts occur in the high-level aquifer starting at a depth of approximately 1 km beneath mean sea level, suggesting that the high-level aquifer is a vertically extensive body of fresh water that is hydraulically continuous with seawater, most likely via a brackish water mixing zone. In light of the virtual water-well approach, the resistivities from the 2D image are inverted for porosity and saturant salinity, using an Archie's law-based Bayesian inversion routine that we have developed. The resulting salinity profile reflects an increase in salinity with depth, but with very few samples meeting or exceeding the salinity of seawater. Due to the non-identifiability of the Archie's law parameters, incorporating sufficient, high-quality prior information is probably key when trying to obtain useful outcomes, for example by constraining parameter ranges in terms of depth range, using prior porosity information from core samples, or using prior salinity information from water wells in the region.