Quantifying Geophysical Inversion Uncertainty Using Airborne Frequency Domain Electromagnetic Data—Applied at the Province of Zeeland, the Netherlands

Abstract

An accurate understanding of the fresh‐saline distribution of groundwater is necessary for effective groundwater management. Airborne electromagnetic (AEM) surveys offer a rapid and cost‐effective method with which to map this, offering valuable additional information about the subsurface. To convert AEM data into electric conductivity and ultimately groundwater salinity, an inversion is undertaken. A number of algorithms are available for this purpose; however, these are affected by significant uncertainty, owing to inherent nonunique characteristics of this process. The most commonly used inversion codes in hydrogeophysical studies were quantitatively tested using frequency domain AEM and ground data from the province of Zeeland, the Netherlands. These include UBC1DFM code and quasi‐2D laterally constrained inversions. Following an investigation of inversion parameter settings, data were inverted for four inversion methods and interpolated into 3‐D volumes. Using geological data and empirical electrical conductivity and water salinity relationships, each inversion was converted into groundwater electrical conductivity and split into fresh‐brackish‐saline regions. For groundwater volume estimates out of a total volume of 2.8 billion m3, a fresh groundwater estimate could differ by as much as 178 million m3, depending on the inversion used. The primary factor here was the choice of model smoothness, which was shown to affect the thickness of the brackish interval. Fresh‐brackish‐saline interfaces were consistently mapped with an accuracy of ~3 m, the brackish being the most accurately resolved. The few layer method was less successful at resolving smoothly varying salinity distributions but more successful at mapping the brackish interface at greater depth.

Automatic identification of fresh–saline groundwater interfaces from airborne electromagnetic data in Zeeland, the Netherlands

“…Instead, we analyse resistivity–depth functions derived from already existing smooth inversion models because we do not focus on optimizing the inversion as discussed by King et al . ().…”

Coastal groundwater systems: mapping chloride distribution from borehole and geophysical data

“…Such a solution is always a simplified image of the subsurface. Comparison studies (e.g., Delsman et al 2018 andKing et al 2018), however, demonstrated that the majority of models were able to reveal the principle layered resistivity features. A further challenge occurs if the subsurface is strongly heterogeneous in the lateral direction.…”

Probabilistic Categorical Groundwater Salinity Mapping From Airborne Electromagnetic Data Adjacent to California's Lost Hills and Belridge Oil Fields

“…Furthermore, resistivity structure is estimated from AEM data through inverse modeling, leading to uncertainty in how well resistivity itself is known. This is a particular concern for the narrow range of resistivity values pertinent to mapping saline and brackish groundwater, and different deterministic inversion algorithms have been shown to have measurable impacts on salinity mapping outcomes (King et al, 2018). The approach presented in this paper uses a Bayesian inversion (Minsley, 2011) to develop statistical models of resistivity derived from AEM data and local correlations between resistivity and salinity to quantify the joint uncertainty in both the interpretational relationship and in how well resistivity is known, following recent work by Christensen et al (2017) using synthetic AEM data to predict lithologic classes.…”