AI-powered hybrid zoning: Optimizing pristine conservation in Abbas abad wildlife refuge

This study advances conservation zoning in the Abbas Abad Wildlife Refuge (AWR), a critical arid ecosystem in central Iran, by integrating artificial intelligence (AI) with a hybrid pixel- and object-based methodology and the Ordered Weighted Averaging (OWA) method. Targeting the pristine conservation zone, we utilized Landsat-۹ imagery (۳۰m resolution) to generate ۲۲۹ optimally sized segments, ensuring ecological accuracy and computational efficiency. Habitat suitability for nine key species, including Persian leopard, Asiatic cheetah, Urial sheep, and Asian houbara, was modeled using MaxEnt, revealing high-elevation zones as vital habitats for biodiversity preservation. Multi-Criteria Decision-Making (MCDM) with OWA, paired with AI models (BRT, ANN, CART, XGBoost, and RF), facilitated precise spatial assessment. The Random Forest (RF) model outperformed others, achieving an overall accuracy of ۰.۹۳ and a Kappa coefficient of ۰.۹۱, validated by ۶۰ ground points that confirmed robust alignment with field observations. The OWA method's low inconsistency coefficient (<۰.۱) and emphasis on fauna criteria produced reliable suitability maps. Unlike traditional pixel-based approaches, which often suffer from salt-and-pepper noise, our hybrid methodology delivered cohesive zoning outcomes, enhancing the management of complex arid ecosystems. This AI-driven framework provides a scalable, innovative model for conservation planning in protected areas, effectively balancing biodiversity preservation with practical implementation. Future research could integrate dynamic environmental factors, such as seasonal climate variations, involve local communities in MCDM processes to align with socio-economic needs, and explore deep learning techniques, like convolutional neural networks, to further enhance classification accuracy, setting a new standard for sustainable environmental management in arid regions.

کلیدواژه ها:

Conservation zoning ، Artificial intelligence (AI) ، Ordered weighted averaging (OWA) ، Hybrid pixel-object approach ، Abbas abad wildlife refuge (AWR)