Determination of critical aquifer subsidence points by DFFITS-COOK distance method using Importance sampling reliability analysis

Abstract

The aim of this study is to investigate aquifer structure subsidence by groundwater level drawdown. In this study, a performance function matrix was developed in the first-order reliability method (FORM) by the importance sampling in order to investigate the regional resilience in a saturated aquifer. The study region is located in southwestern Iran. The changes in the layers of vertical structure destruction are an important factor affecting the plain's recharge capacity. Preparing the subsidence prediction layers due to the MODFLOW finite-difference model and criterion for its changes was related to five scenarios of recharge changes in the aquifer water level from 2008 to 2019 and also from 2019 to 2029. The results indicated that failure probability in Izeh aquifer varies from 0 to about 89 %. Considering structure type, most of the internal plain areas have high local resilience. For the aquifer boundary strips, the highest resilience is only in the southwestern part of aquifer. The high-risk border areas have a wide fitness on porous boundary strips as the main recharge sources. The reason for severe spatial changes in the failure probability of reliability was investigated by out-of-date analysis methods. The anomaly pattern was used in three methods of defining decision function to determine out-of-point under Robust Covariance, One-Class SVM and Isolation Forest. DFFITS and DFFITS methods (difference in fits) with COOK Distance were used to confirm the existence of 106 abnormal cells in the two-dimensional digit of baseline indices mean. Results showed that the most critical areas can be found in the central parts around the two lakes. The plain's permeable boundaries, the parts between the northwest and the center of the aquifer, and the northern aquifer's limited areas are other abnormal regions.