Scholarlink

<p>Aquatic invasive species (AIS) pose a major threat to freshwater biodiversity, ecosystem functioning, and associated socio-economic values worldwide. Effective prevention and management depend on understanding how human activities, environmental conditions, and dispersal pathways interact to shape invasion risk across landscapes. However, establishing predictive links between AIS occurrences and their introduction pathways remains challenging, and invasion risk assessments rarely consider the full spectrum of pathways operating across heterogeneous landscapes. Here we show that combining machine-learning approaches with stakeholder input identifies complementary drivers of AIS invasion risk and reveals consistent spatial hotspots across southern Québec, Canada. Using Boruta feature selection and random forest models applied to occurrence data from 54 AIS taxa, we identified hydrographic connectivity, climate, commercial fisheries, transportation infrastructure, and recreational activities as the most influential predictors of invasion risk, with marked differences among taxonomic groups. Models explained up to 65% of variation in AIS richness and consistently highlighted the St. Lawrence River watershed and adjacent urban regions as areas of highest risk. Stakeholder-informed models emphasized management-relevant vectors such as boat launches, boat cleaning practices, marinas, and ballast water management, revealing additional secondary risk areas beyond major invasion hubs. Although stakeholder-based models showed lower predictive performance, they captured locally relevant pathways not fully represented in spatial datasets. These results demonstrate that integrating quantitative modelling with community knowledge improves the identification of invasion pathways and high-risk areas. This framework provides an immediately applicable tool for prioritizing surveillance and prevention efforts and offers a flexible, transferable approach for freshwater invasion risk assessments in data-limited management contexts.</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

Integrating machine learning and stakeholder insights to identify key vectors of aquatic invasive species and map invasion risk

  • 期刊论文

  • 发表日期：

    2026年6月18日

查看全文
  • Cristina Charette,
  • Andréanne Paris,
  • Jade Montaudié,
  • Lars L. Iversen,
  • Olivier Morissette

摘要

Aquatic invasive species (AIS) pose a major threat to freshwater biodiversity, ecosystem functioning, and associated socio-economic values worldwide. Effective prevention and management depend on understanding how human activities, environmental conditions, and dispersal pathways interact to shape invasion risk across landscapes. However, establishing predictive links between AIS occurrences and their introduction pathways remains challenging, and invasion risk assessments rarely consider the full spectrum of pathways operating across heterogeneous landscapes. Here we show that combining machine-learning approaches with stakeholder input identifies complementary drivers of AIS invasion risk and reveals consistent spatial hotspots across southern Québec, Canada. Using Boruta feature selection and random forest models applied to occurrence data from 54 AIS taxa, we identified hydrographic connectivity, climate, commercial fisheries, transportation infrastructure, and recreational activities as the most influential predictors of invasion risk, with marked differences among taxonomic groups. Models explained up to 65% of variation in AIS richness and consistently highlighted the St. Lawrence River watershed and adjacent urban regions as areas of highest risk. Stakeholder-informed models emphasized management-relevant vectors such as boat launches, boat cleaning practices, marinas, and ballast water management, revealing additional secondary risk areas beyond major invasion hubs. Although stakeholder-based models showed lower predictive performance, they captured locally relevant pathways not fully represented in spatial datasets. These results demonstrate that integrating quantitative modelling with community knowledge improves the identification of invasion pathways and high-risk areas. This framework provides an immediately applicable tool for prioritizing surveillance and prevention efforts and offers a flexible, transferable approach for freshwater invasion risk assessments in data-limited management contexts.