| Abstrakt: |
This research aims to evaluate the vulnerability of aquifers by comparing two approaches of deep learning and machine learning in index calibration. Therefore, by analyzing the inherent vulnerability of the Amol-Babol aquifer with the DRASTIC index, the sensitive areas of the aquifer were identified. The results of the vulnerability index showed that the northwestern part of the aquifer is more sensitive than other areas. Examining the correlation value between nitrate concentration as an effective index with the DRASTIC vulnerability index indicates a value of 24%, which indicated the need for recalibration. Therefore, with two CNN-Harris Hawks and LSTM-MPA methods as deep learning approaches, weighting and index ranks were carried out as decision variables with the aim of maximizing the correlation of nitrate concentration and vulnerability index. The results showed that the CNN-HHO method with a correlation of 0.62 is superior to the LSTM-MPA method with a correlation of 0.59. Vulnerability zones in the assessment conditions showed that the western and northeastern parts have higher vulnerability. On the other hand, the recalibrated weights and ranks indicate an increase in all weights and ranks in recalibration conditions compared to the initial conditions, which was determined after analyzing the optimization approaches. [ABSTRACT FROM AUTHOR] |
