A multi-scale ensemble machine learning framework for assessing human-elephant conflict in the Brahmaputra flood...

Area of Science:

• Ecology
• Conservation Biology
• Wildlife Management

Background:

• Human-elephant conflict (HEC) poses a significant threat to biodiversity and rural livelihoods in Asia.
• The Brahmaputra Flood Plain (BFP) in Assam is a critical habitat for Asian elephants, experiencing increasing HEC.
• Understanding the spatio-temporal dynamics of HEC is crucial for effective conservation strategies.

Purpose of the Study:

• To comprehensively analyze the spatio-temporal dynamics of human-elephant conflict (HEC) in Assam's Brahmaputra Flood Plain (BFP) from 2010 to 2024.
• To integrate human and elephant victimization perspectives for a holistic understanding of conflict risk.
• To develop an adaptable decision-support tool for mitigating HEC and promoting coexistence.

Main Methods:

• Employed a two-way conflict modeling framework analyzing both human victimization of elephants (EV-HI) and elephant victimization of humans (HV-EI).
• Utilized a multi-scale ensemble of five machine learning algorithms (RF, BRT, SVM, CART, MARS) with AUC weighting to create risk maps.
• Integrated time dynamics, conflict type correlations, and bivariate hotspot mapping for a novel analytical framework.

Main Results:

• HEC incidents show strong seasonality, peaking post-monsoon, and have significantly increased since 2017.
• Scale-dependent variables like orchard cover and forest/population ratio predict EV-HI risk.
• Landscape heterogeneity and human footprint predict HV-EI risk, with edge density being a common factor.

Conclusions:

• The study introduces a novel framework for analyzing HEC, considering risks to both humans and elephants.
• Results provide management-friendly insights for targeted interventions like crop-switching and early warning systems.
• The developed decision-support tool aids in strengthening conservation planning and promoting human-elephant coexistence.