Developing of an artificial intelligence model for predicting algal blooms in the Oman Sea an‎d the Persian Gulf based on remotely sensed climatology data

9/17/2026 12:00:00 AM

Forecasting Harmful Algal Blooms (HABs) is critical for the sustainable management of marine ecosystems in the Persian Gulf an‎d Gulf of Oman. This study develops a novel artificial neural network (ANN) model to forecast monthly HAB occurrences using freely available, Level 3 an‎d 4 remotely sensed data, including chlorophyll-a (Chl-a), sea surface temperature (SST), salinity, an‎d wind speed. The core innovation is the development an‎d validation of novel statistical-dynamic features, including spatial an‎d temporal stan‎dard deviations (SD) an‎d a spatial gradient feature (SSGF), designed to quantify environmental stability an‎d dynamics. The multi-layer perceptron model, trained on 66% an‎d tested on 34% of the data, achieved an overall accuracy of 88.7% in binary classification (bloom/non-bloom) with an Area Under the ROC Curve (AUC) of 0.901. Feature importance analysis revealed that the novel features, particularly those capturing thermal stability an‎d spatial heterogeneity, were among the strongest predictors after the current monthʹs Chl-a concentration. A key finding of this research is the definitive role of environmental dynamics an‎d stability as critical precursors to bloom events. The results demonstrate that blooms are highly dependent not just on the absolute values of climatic parameters, but also on their trends, stability, an‎d spatial patterns. Ultimately, this study presents an effective, low-cost, an‎d operational approach for HAB forecasting, providing a valuable decision-support tool for regional environmental stakeholders.

شبكه عصبي مصنوعي , ، كلروفيل-آ , ، سنجش از دور , ، يادگيري ماشين , شكوفايي جلبكي مضر

Artificial Neural Network , HABs , Climate Analysis , Machine Learning , Chl-a , Remote Sensing

Mohammad Hosein Seyed-Djawadi

Dr. Siadatmoosavi - Dr. Haghayeghi