Using Self-Organizing Maps to find spatial relationships between wildlife-vehicle crashes and land use classes

Area of Science:

• Environmental Science
• Wildlife Ecology
• Transportation Ecology

Background:

• Road construction and expansion significantly impact the environment, affecting vegetation and animal distribution.
• Habitat fragmentation from roads increases wildlife-vehicle collision (WVC) risks.
• Understanding spatial patterns of WVC is crucial for mitigation.

Purpose of the Study:

• To analyze spatial patterns of wildlife-vehicle crashes using spatial analysis and machine learning.
• To establish relationships between WVC patterns and landscape features.
• To identify distinct spatial patterns for different animal types.

Main Methods:

• Utilized spatial analysis and machine learning tools, specifically Self-Organizing Maps (SOM), an artificial neural network (ANN).
• Reorganized multi-dimensional data based on similarity to identify spatial patterns.
• Analyzed relationships between ecological data and WVC events.

Main Results:

• Wildlife-vehicle crash events are spatially clustered and not uniformly distributed along highways.
• Distinct spatial patterns were identified for different animal types.
• Most collisions occurred near forest areas and water bodies, less so near sugarcane fields, forestry, or built environments. Diverse landscapes showed a considerable number of collisions.

Conclusions:

• SOM effectively identified distinct spatial patterns in wildlife-vehicle collisions based on animal type and landscape context.
• Proximity to forest areas and water bodies are significant factors in WVC.
• Findings highlight the importance of landscape heterogeneity in WVC and inform targeted mitigation strategies.