USING GRAVITY MODELING TO SEPARATE UNCONSOLIDATED FROM CONSOLIDATED BASIN FILL

An iterative, nonlinear modeling technique has been used to model basin gravity anomalies as a two variable-thickness-layer model of consolidated basin fill overlain by unconsolidated fill (generally the best aquifer). The model used gridded residual gravity anomaly data, that is, anomalies from which the contribution due to bedrock has been removed by subtracting a gravity grid constructed using only bedrock stations plus stations near drillholes that penetrated to bedrock. All available drillhole information is used to constrain the two model surfaces, together with the outcrop edge of basin fill. For the San Pedro and the Santa Cruz basins in southeastern Arizona, drillhole logs were used to define the density-depth functions for each layer. The functions were approximated by constant values for given depth ranges for both the consolidated and unconsolidated units. Bedrock was assumed to have a constant density. Initial grids of depth to bedrock and depth to consolidated layer were obtained by gridding the constraining geologic and drillhole data. Then, for a given grid intersection, the gravity anomaly as a function of depth is calculated from the thickness grids and the density-depth functions for each unit. This gravity-depth function, which is nonlinear, is then used to interpolate a new depth from the difference between the model gravity anomaly and the observed gravity anomaly. This process is iterated first holding the depth to consolidated fill fixed, to improve the depth to bedrock estimate, and then holding depth to bedrock fixed to improve the thickness of unconsolidated fill estimate. About ten complete iterations yielded models that were within data uncertainties in both basins studied. Success of the model is mixed, due mainly to gravity anomaly uncertainties and the heterogeneous nature of the consolidated layer in both basins. However, the estimates are optimum in the sense that they fit both the gravity and drillhole data.