Land Subsidence in Western US Mapped Using Machine Learning

Land subsidence caused by groundwater extraction has numerous negative consequences, such as loss of groundwater storage, damage to infrastructure and in many cases arsenic contamination. Understanding the magnitude, timing, and locations of land subsidence, as well as the mechanisms driving it, is crucial to implementing mitigation strategies, yet the complex, nonlinear processes causing subsidence are difficult to quantify. Physical models relating groundwater flux to aquifer compaction exist, but require substantial hydrological datasets and are time consuming to calibrate. Land deformation can be measured using InSAR and GPS, but the former is computationally expensive to estimate at scale and is subject to tropospheric and ionospheric error, and the latter leaves many spatial gaps. In this study, we apply for the first time a machine learning approach that quantifies relationships of various gridded input data, including evapotranspiration, land use, and sediment thickness, with land subsidence. We use existing land subsidence datasets from InSAR and GPS to train the model. We apply this method over the Western United States. This approach allows us to both map the spatial distribution of land subsidence over a larger region than has previously been possible, and to quantify the relationship between various input data and land subsidence due to groundwater extraction. We find that sediment thickness and cultivated area have the strongest relationship to subsidence. Subsidence is concentrated in the Central Valley of California, with significant additional subsidence areas in cultivated areas of the Basin and Range province, as well as the San Luis and Albuquerque Basins. Agricultural area accounts for the majority of land subsidence, followed by urban areas. This study demonstrates that widely available ancillary data can be used to estimate subsidence over large spatial scales if sufficiently large training datasets are available. This approach can enhance InSAR-based deformation studies by approximating the groundwater contribution to deformation, and identifying periods or areas of interest where InSAR processing would be most beneficial.