Predicting Short-Term Deformation in the Central Valley Using Machine Learning

Land subsidence caused by excessive groundwater pumping in Central Valley, California, is a major issue that has several negative impacts such as reduced aquifer storage and damaged infrastructures which, turn, produce an economic loss due to the high reliance on crop production. This why it of utmost importance routinely monitor assess surface deformation occurring. Two main goals this paper attempts accomplish are characterization prediction. The first goal realized through use Principal Component Analysis (PCA) applied series Interferomtric Synthetic Aperture Radar (InSAR) images produces eigenimages displaying key characteristics subsidence. Water changes also directly analyzed data from Gravity Recovery Climate Experiment (GRACE) twin satellites Global Data Assimilation System (GLDAS). second accomplished building Long Short-Term Memory (LSTM) model predict short-term after developing InSAR time using LiCSBAS, open-source package. city Madera better results than baseline averaging one dimensional convolutional neural network (CNN) based mean squared error metric showing effectiveness machine learning prediction well potential for incorporation hazard mitigation models. can aid policy makers determining appropriate rate withdrawal while maintaining safety well-being population aquifers’ integrity.