Machine-Learning Characterization of Tectonic, Hydrological and Anthropogenic Sources of Active Ground Deformation in California

Tectonic, hydrological and industrial processes coexist in the dynamic natural environments. However, our knowledge of ground deformation associated with tectonic, hydrological and anthropogenic processes and their interactions remains limited. California represents a natural laboratory that hosts the San Andreas fault system, Central Valley and other aquifer systems, and extensive human extraction of natural resources. The attendant multi-scale ground deformation that has been mapped using Copernicus Sentinel-1 Synthetic Aperture Radar (SAR)-satellite constellation from four ascending and five descending tracks during 2015–2019. We consider the secular horizontal surface velocities and strain rates, constrained from GNSS measurements and tectonic models, as proxies for tectonic processes, and seasonal displacement amplitudes from interferometric SAR (InSAR) time series as proxies for hydrological processes. We synergize 23 types of multidisciplinary datasets, including ground deformation, sedimentary basins, precipitation, soil moisture, topography, and hydrocarbon production fields, using a machine learning algorithm–random forest, and we succeed in predicting 86%–95% of the representative data sets. High strain rates along the SAF system mainly occur in areas with a low-to-moderate vegetation fraction (∼0.3), suggesting a correlation of rough/high-relief coastal range morphology and topography with the active faulting, seasonal and orographic rainfall, and vegetation growth. Linear discontinuities in the long-term, seasonal amplitude and phase of the surface displacement fields coincide with some fault strands, the boundary zone between the sediment-fill Central Valley and bedrock-dominated Sierra Nevada, and the margins of the inelastically deforming aquifer in the San Joaquin Valley, suggesting groundwater flow interruptions, contrasting elastic properties, and heterogeneous hydrological units. Plain Language Summary Although scientific advances have been achieved in every individual geoscience discipline with more extensive and accurate observations and more robust models, our knowledge of the Earth complexity remains limited. The tectonic, hydrological and anthropogenic processes interact in the highly populous California, which have contributed to multi-scale ground deformation. Here we rely on remotely sensed ground deformation products and locations of oil and gas fields as proxies for tectonic, hydrological and anthropogenic processes. The training model of the random forest algorithm has a good performance in predicting the representative multidisciplinary datasets and reveals their intrinsic similarities and relative importance in the predictions. We also note compelling spatial correlation between the long-term and seasonal displacement discontinuities, the high-strainrate fault zones, and a narrow range of vegetation fraction, as well as the margins of the heterogeneous structures.