Deep Learning Identification of Priority Corridors for Large Mammal Conservation

Abstract

Introduction: Habitat fragmentation, land use change, and human activities that affect the natural movement of animals have become the cause of many threats to large mammal populations. It is necessary to identify important wildlife corridors as a way to ensure habitat continuity and the movement of species. Conventional techniques to determine wildlife corridors rely mainly on field work and ecological modeling and can be limited in analyzing the correlation between various environmental variables. Materials & Methods: The present study offers a method based on deep learning for corridor identification for large mammals. The method incorporates the data from remote sensing, mammal occurrence data, environmental factors, vegetation information, topographic factors, and human interference index. The obtained data is then preprocessed and subjected to feature extraction and integration, and is used for analysis through deep learning models. The proposed method ranks corridor locations based on habitat suitability and landscape connectivity. Results: The outcomes show that the proposed deep learning framework has been able to capture the intricate relationships between the ecological parameters and animal movement. The model performed well when it came to classifying corridors with 95.2% accuracy, 94.7% precision, 93.9% recall, 94.3% F1-Score, and 0.96 AUC, which is superior to traditional methods. The prioritized corridors were linked to dense vegetative coverage, low human disturbance, and connectivity to protected areas. Conclusion: The presented framework is a good example of an effective data-driven technique for large mammal corridor identification and conservation. Through the combination of ecological and geographical information, the model contributes to more effective decision-making in regard to habitat and biodiversity preservation as well as sustainable wildlife conservation. Future research may involve the use of real-time data and multiple species movements.