The development of a machine-learning approach to construct a field-scale rock-physics transform

Abstract

The ability to relate geophysical measurements to the material properties of the subsurface is fundamental for the successful application of geophysical methods. Estimating the electrical resistivity from material properties can be challenging at many hydrogeologic field sites, which typically lack the spatial density and resolution of the measurements needed to develop an accurate rock-physics relationship. We have developed rock-physics transforms using the machine-learning method of gradient-boosted decision trees (GBDTs). We adopt as our study area the coastal Salinas Valley, in which saltwater intrusion results in changes in resistivity. We use measurements available in boreholes, including salinity and sediment type, to predict the resistivity. In some transforms, we include as predictors in the GBDT algorithm the location of each measurement and the aquifer corresponding to each measurement. We also explore incorporating the predictions of a baseline rock-physics transform as a prior term within the objective function of the GBDT algorithm to guide the predictions made by the machine-learning algorithm. The use of location and aquifer information improves the predictions of the GBDT transform by 28% compared with when location and aquifer information are not included. After the salinity, the easting of each measurement is the most important predictor, due to the spatial pattern of salinity changes in the area. The next most important predictor is the aquifer corresponding to each measurement. The benefit of including the baseline transform in the objective function is greatest for small data sets and when the accuracy of the baseline transform is already high. Finally, using the resistivity predicted by the GBDT, we generate 1D resistivity models, which we use to simulate the acquisition of airborne electromagnetic (AEM) data. In most cases, the 1D resistivity models and the corresponding AEM data match well with the models and data corresponding to the resistivity measured in boreholes.