Optimizing mangrove afforestation site selection in gulf cooperation council nations using remote sensing and machine learning

Mangrove forests are vulnerable coastal ecosystems that provide multiple ecosystem services and act as blue carbon sinks. Mangroves in the Gulf Cooperation Council (GCC) countries of the Arabian peninsula (Bahrain, Kuwait, Oman, Qatar, Saudi Arabia and the UAE) have faced pressure from numerous anthropogenic factors including population growth, dredging and reclamation in lagoonal habitats, pollution and rapid urban development. Given the inconsistent outcomes of past mangrove ARR (Afforestation-Reforestation-Revegetation) efforts, our research aimed to identify high-potential ARR sites in the GCC using remote sensing. We identified eight factors related to mangrove ARR outcomes through correlation analyses: elevation, soil pH, median precipitation, median and minimum land surface temperature (LST), soil salinity, soil texture and distance from urban areas. To predict mangrove suitability, we compared the Random Forest (RF), XGBoost (XGB), Support Vector Machines (SVM) and Naive Bayes (NB) classification algorithms. The RF model performed best with an F1-score of 0.96, ROC-AUC of 0.99 and PR-AUC of 0.75. Variable importance analysis revealed that median LST, soil texture and median precipitation were the most influential variables. Favorable conditions for mangrove establishment included median temperatures of 32–37 °C, minimum temperatures around 27 °C, clayey soils, and monthly rainfall above 10 mm. Other suitable characteristics included lower elevation, greater distance from urban areas, slightly acidic to neutral pH, and moderate-to-high soil salinity. Our findings show that there is a large opportunity for mangrove afforestation in the GCC and also proposes a framework to identify optimal sites for mangrove growth, which can improve ARR success and support biodiversity and blue carbon goals in the Arabian peninsula.