Work

Bayesian Neural Network-Based Habitat Suitability Prediction for Amphibians Using Remote Sensing Ecological Indices: A Case Study of Pelophylax Nigromaculatus in China

Year: 2026

Type: article

Abstract: Timely and accurate assessment of wildlife habitats is essential for conservation planning, particularly for environmentally sensitive species such as amphibians. This study develops a dynamic habitat modeling framework by integrating Bayesian neural networks with multisource remote sensing data. Using P. nigromaculatus as a case study, we simulated annual habitat distributions from 2016 to 2025 and evaluated the applicability of the proposed approach for large-scale habitat prediction. The model achieved high predictive performance (AUC > 0.9) and demonstrated stable temporal consistency. It showed strong agreement with the maximum entropy model in overall spatial patterns while providing improved delineation of fine-scale habitat patches. In addition, the Bayesian neural network (BNN) framework generated quantitative u+ncertainty estimates ranging from 0 to 0.362, indicating generally low to moderate uncertainty and high confidence across the study period. The selected remote sensing-derived habitat indicators effectively supported interannual habitat simulations from 2016 to 2025, confirming their robustness in driving dynamic habitat prediction. Based on this framework, key environmental drivers, including LCLU, land surface temperature, and normalized difference water index, were identified and their spatial influences on habitat variation were further characterized. Overall, the integration of multisource remote sensing data with BNN models enables accurate and dynamic large-scale habitat modeling for amphibians, providing a robust framework for conservation planning. The proposed approach also offers a transferable methodology for habitat prediction in other amphibian species.

Source: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing

DOI: 10.1109/jstars.2026.3725152

Authors: Kamilan Abulike, Wenfang Zhang, Wenyu Dai, He Wu, Jiechen Wang

Institution: Nanjing University

Language: English

Topic: Species Distribution and Climate Change

Subfield: Ecological Modeling

Field: Environmental Science

Domain: Physical Sciences

SDGs: Life on land, Climate action

Open Access status: gold

Estimated APC paid: $1,800