 This study utilizes inSAR processing and machine learning methods to model and predict future land subsidence behavior in the vicinity of Lake Ermia, Iran. Sentinel-1 data over 56 months and small baseline subsets, SBAS, inSAR methods were used to identify regions with high rates of subsidence. Environmental factors affecting subsidence were also considered using TIM, GRACE, and MODIS satellite data. Several machine learning methods were implemented to determine the relationship between land deformations and environmental variations, including multilayer perceptron, MLP, convolutional neural network, CNN, and long short-term memory, LSTM, networks. A weighted ensemble was constructed by blending the forecast of the three models, which outperformed the single models and reached an IMSE, MAE, and SD of 8.2 millimeters, 6.4 millimeters, and plus or minus 5.2 millimeters, respectively. The study suggests that an ensemble model can improve land deformation anticipation by leveraging the strengths of networks in various conditions. This article was authored by Ali Radman, Media Kunzadeh, and Benyamin Hosseini.