ESTIMATING THE PROBABILITY OF ALTERNATIVE GEOLOGIC MODELS USING GRAVITY DATA WITH DISTANCE-BASED KERNELS

We propose a method of leveraging gravity modeling to estimate the probability of different alternative conceptual geologic models of the subsurface. While geophysical modeling is often used to provide a model, or series of models, of the subsurface geologic structure, rarely are probabilities assigned to the conceptual geologic models resulting from these investigations. For example, consider a region where groundwater flow is of interest, and a fault is suspected, but not certain, to exist in the subsurface. Answering the question “What is the likelihood of a contaminant traveling a certain distance within a given period of time?” requires an estimate of the probability that the fault exists. To answer such questions we use a forward modeling approach that approximates the probability of occurrence of different alternative conceptual geologic models yielding gravity anomalies that match the observed. We recognize that geologic sedimentary and structural features produce heterogeneous changes in density across a range of scales. These heterogeneities cannot be mapped directly but may significantly impact the resulting gravity anomaly. We therefore use Monte Carlo approach to simulate the heterogeneities and populate alternative conceptual geologic models with geostatistically-generated realizations of the subsurface distribution of density. We forward calculate the gravity anomalies of the proposed models, then use distance-based kernel methods to map the input conceptual geologic models to the output gravity anomalies, and estimate the probabilities of each conceptual geologic model by comparing the calculated gravity anomalies with the observed. Working in two dimensions, we develop this method through the investigation of a gravity profile across the Vaca Fault, a north-trending fault on the western edge of the northern central valley, California. Current geologic mapping and cross-sections cannot resolve the dip of the Vaca Fault, and gravity was proposed as a method of investigating fault dip. Our investigations reveal that the Vaca Fault has approximately a 70% chance of being shallow to moderately dipping, and approximately 30% chance of being steeply dipping or overturned, based on the gravity data.