Predicting Geographic Distribution and Potential Habitat of Marine Bivalves *

Abstract

Understanding the habitat suitability of the species has been one of the main focus of biodiversity conservation. Species Distribution Modelling (SDM) has great potential to support marine conservation planning. SDM can forecast the optimal conditions for species cultivation, aiding in the prevention of habitat loss and biodiversity degradation. Four habitat suitability models-DNN, MaxEnt, RF, and GBM-were utilized to forecast the distribution of marine bivalves. Also, a stacking ensemble-based estimation model was developed, utilizing the model performance outcomes as input to enhance estimation accuracy. Finally, the habitat suitability results based on the performance evaluation metrics such as the area under the curve (AUC), sensitivity, specificity, the kappa statistic, and the TSS were used to evaluate the predictive performance of the SDMs. The experimentation results indicate that the Ensemble model delivers superior predictive performance as evidenced by its exceptional scores and demonstrated strong predictive capabilities in forecasting the potential habitat of bivalve species. The implemented model achieves impressive AUC values of 0.98, along with Kappa statistics and Specificity scores of 0.96, and Sensitivity and TSS scores of 0.97. These findings underscore the effectiveness of the Ensemble model in accurately predicting bivalve habitat suitability, thus providing valuable insights for marine conservation planning efforts.