Scholarlink

<p>Human–elephant conflict (HEC) is emerging as a growing challenge to biodiversity conservation and rural livelihoods in Asia. Assam’s Brahmaputra Flood Plain (BFP)is an important stronghold of the Asian elephant, wherein this study aims to comprehensively analyse the spatio-temporal dynamics of human–elephant conflict BFP, between 2010 and 2024. Using a two-way conflict modelling framework, we examined both human victimization of elephants (EV–HI) and elephant victimization of humans (HV–EI), and integrated these perspectives to identify spatial intersections and temporal patterns of heightened conflict risk. We designed effective risk maps with AUC weighting based on a multi-scale ensemble of five machine learning algorithms (i.e., RF, BRT, SVM, CART, MARS), optimally weighting predictor spatial scales (1–15&#xa0;km). Conflict incidents exhibited strong seasonality, peaking post-monsoon when major crops such as rice, maize, vegetables, and standing sugarcane are harvested or remain available in the fields. Incidents of conflict have risen significantly after 2017. Spatial analysis revealed that scale-dependent variables, such as orchard cover, forest/population ratio, and edge density, predicted EV–HI risk, whereas landscape heterogeneity, human footprint, and edge density predicted HV–EI risk. We have taken the initiative to combine time dynamics, conflict type correlations, and bivariate hotspot mapping to introduce a new framework that would be concerned not only with risks to human beings but also with risks to elephants. The results are management-friendly and facilitate specific interventions. Our study provides an adaptable decision-support tool that guides crop-switching, early warning alerts, fence evaluations, and corridor restoration. Unlike the usual single-model or one-scale approaches, it integrates multiple perspectives to strengthen conservation planning and promote human–elephant coexistence by protecting both livelihoods and elephant populations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

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

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月25日

查看全文
  • Kajoli Begum,
  • Bishal Kumar Majhi,
  • Mriganka Shekhar Sarkar,
  • Ahanthem Rebika Devi,
  • Wishfully Mylliemngap,
  • Anuradha Reddy

摘要

Human–elephant conflict (HEC) is emerging as a growing challenge to biodiversity conservation and rural livelihoods in Asia. Assam’s Brahmaputra Flood Plain (BFP)is an important stronghold of the Asian elephant, wherein this study aims to comprehensively analyse the spatio-temporal dynamics of human–elephant conflict BFP, between 2010 and 2024. Using a two-way conflict modelling framework, we examined both human victimization of elephants (EV–HI) and elephant victimization of humans (HV–EI), and integrated these perspectives to identify spatial intersections and temporal patterns of heightened conflict risk. We designed effective risk maps with AUC weighting based on a multi-scale ensemble of five machine learning algorithms (i.e., RF, BRT, SVM, CART, MARS), optimally weighting predictor spatial scales (1–15 km). Conflict incidents exhibited strong seasonality, peaking post-monsoon when major crops such as rice, maize, vegetables, and standing sugarcane are harvested or remain available in the fields. Incidents of conflict have risen significantly after 2017. Spatial analysis revealed that scale-dependent variables, such as orchard cover, forest/population ratio, and edge density, predicted EV–HI risk, whereas landscape heterogeneity, human footprint, and edge density predicted HV–EI risk. We have taken the initiative to combine time dynamics, conflict type correlations, and bivariate hotspot mapping to introduce a new framework that would be concerned not only with risks to human beings but also with risks to elephants. The results are management-friendly and facilitate specific interventions. Our study provides an adaptable decision-support tool that guides crop-switching, early warning alerts, fence evaluations, and corridor restoration. Unlike the usual single-model or one-scale approaches, it integrates multiple perspectives to strengthen conservation planning and promote human–elephant coexistence by protecting both livelihoods and elephant populations.