Deep Learning Mapping of Road Barriers Affecting Wildlife Movement

Abstract

Introduction: Expansion and improvement of road infrastructure have substantial impacts on the migration behavior of wildlife in terms of acting as barriers, fragmentation of their habitat and loss of ecological connectivity. Precise delineation of road barriers is critical for effective wildlife corridor design and management. The recent advances in deep learning and remote sensing technology can be exploited to address such challenges.

Materials and Methods: A deep learning approach for developing road barrier maps impacting wildlife movements through the use of satellite images, road network information, and observations regarding wildlife movements has been put forth in this study. The gathered data undergo geospatial alignment, enhancement, normalization, and feature extraction processes. The application of a deep learning model using spatial feature analysis for the detection of road barriers, habitat fragmentation assessment, and movement barriers evaluation has been employed.

Results and Discussion: The performance of the proposed framework was better than the traditional deep learning algorithms for barrier detection, movement risk prediction, and habitat connectivity assessment. The proposed method had an accuracy of 96.5% in detecting barriers, 94.8% in predicting movement risk, and 93.7% in habitat connectivity assessment, which were superior to CNN, CNN-LSTM, and ResNet50 models. The findings suggest that the model is able to capture complex spatio-ecological interactions related to wildlife movement constraints.

Conclusion: The proposed framework is an efficient solution for detecting important road barriers and formulating appropriate wildlife management strategies. The combination of deep learning and geospatial data as well as ecological data helps analyze connectivity of the habitats and formulates necessary mitigation plans.