همایش بین المللی هوش مصنوعی

صفحه اصلی / دومین همایش بین المللی هوش مصنوعی

Integrating Physicochemical Analysis and Deep Learning for SGD Detection: A Case Study of the Northern Persian Gulf

نویسندگان :

Ali Lotfi¹ Razieh Lak² Saeedreza Kheradpishe³ Mohsen Ehteshami Moein abadi⁴ Yaser Nikpeyman⁵

1- دانشگاه شهید بهشتی 2- سازمان زمین شناسی 3- دانشگاه شهید بهشتی 4- دانشگاه شهید بهشتی 5- دانشگاه شهید بهشتی

کلمات کلیدی :

تخلیه آب زیرزمینی به دریا،یادگیری ماشین،خلیج فارس،داده‌های فیزیکوشیمیایی

چکیده :

Submarine Groundwater Discharge (SGD) is a hidden yet critical component of the global water cycle, facilitating the transport of nutrients and chemical compounds into coastal ecosystems. This study aims to identify potential SGD zones in the northern Persian Gulf using a hybrid approach that integrates expert hydrogeological analysis with machine learning algorithms. By utilizing physicochemical seawater data—including temperature, salinity, dissolved oxygen, and chlorophyll—collected from the region, a framework was developed to automatically detect hydrochemical anomalies. The methodology involved a multi phase analysis using supervised (Gradient Boosting) and unsupervised (SOM and K-Means) models, culminating in a deep learning model (LSTM) trained on synthetic data to address data scarcity. The results demonstrate that the developed models can identify high-probability SGD hotspots with high accuracy, particularly in the Strait of Hormuz and along the Bushehr coast. Furthermore, a significant spatial correlation was observed between these identified zones and major tectonic structures, such as the Qatar–Kazerun fault. These findings provide a scalable and intelligent pathway for monitoring coastal water resources in the Persian Gulf.