Subsidence poses a significant management challenge, causing damage to infrastructure, energy transmission lines, buildings, soil stability, and leading to the formation of sinkholes. This study employed the Multilayer Perceptron (MLP) neural network to evaluate and model the extent of subsidence in the Dehgolan Plain aquifer, located in Kurdistan Province, Iran, between March 23, 2022, and September 24,2023. A subsidence model was constructed using groundwater level data, changes in transmissivity, alluvial thickness, and results from radar interferometry. Regression analysis comparing predicted and observed values confirmed the model's high accuracy in forecasting subsidence. Furthermore, the model successfully estimated missing subsidence rates. The maximum subsidence calculated using radar interferometry over the 552-day period was 154 mm, while the maximum uplift was 16 mm. In comparison, the MLP model estimated a maximum subsidence of 145 mm and a maximum uplift of 12 mm. Subsidence was found to be more pronounced in the western and central regions of the plain compared to the eastern areas. Considering the ongoing progression of subsidence in the Dehgolan Plain aquifer, it is imperative to implement strategies to reduce the over-extraction of groundwater and establish continuous monitoring systems.
