Scholarlink

Background <p>African animal trypanosomosis (AAT) and human African trypanosomosis (HAT) are transmitted and spread primarily by tsetse flies (<i>Glossina</i> spp.) in sub-Saharan Africa. The animal disease poses significant challenges to agropastoral systems, including in Kenya, where 38 out of 47 counties are infested with eight species of <i>Glossina</i>. Climate change and human activities can also aggravate these infestations, putting rural-scale farmers who rely on agropastoral systems at a greater risk. Geographical gaps in existing entomological datasets limit a comprehensive understanding of tsetse fly distribution across the country, especially amid rapid landscape dynamics.</p> Methods <p>This study aimed to predict the spatial distribution of tsetse flies habitat in Kenya using recent entomological data (i.e., tsetse fly occurrence records), satellite-derived environmental variables, landscape structure, demographic indicators, and species-distribution modeling techniques. We applied four machine learning (ML) algorithms—random forest (RF), support vector machines (SVM), maximum entropy (MaxEnt), and generalized linear models (GLM)—to predict tsetse flies habitat suitability. Additionally, we developed ensemble models that combine the predictive power of the four algorithms. Predictions were made at the genus level (<i>Glossina</i> spp.) and the species level for one priority species (<i>Glossina pallidipes</i>).</p> Results <p>The models performed well with true skill statistic (TSS) and area under the curve (AUC) metric measures of 0.67 and 0.88 for <i>Glossina</i> spp. and 0.85 and 0.96 for <i>G. pallidipes</i>, respectively. The predictions indicated an estimated potential suitable area of about 26% of Kenya for <i>Glossina</i> spp. and 9% for <i>G. pallidipes</i>. Tsetse fly habitat&#xa0;suitability was positively correlated with increased sheep density, normalized difference vegetation index, and soil moisture. However, suitability declined when the maximum land surface temperature (LST) exceeded 40&#xa0;°C and elevation increased above 400&#xa0;m.</p> Conclusions <p>These findings can help improve the targeting and, hence, the cost-effectiveness of surveillance and ultimately support an evidence-based progressive control of tsetse flies infestation in Kenya.</p> Graphical Abstract <p></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

Species distribution modeling to predict tsetse fly (Glossina spp.) habitat suitability in Kenya

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月24日

查看全文
  • Raphael Mongare,
  • Stella Gachoki,
  • Elhadi Adam,
  • Emily Kimathi,
  • Antoine M. G. Barreaux,
  • Giuliano Cecchi,
  • Seth Onyango,
  • Nancy Ngari,
  • Daniel Masiga,
  • Elfatih M. Abdel-Rahman

摘要

Background

African animal trypanosomosis (AAT) and human African trypanosomosis (HAT) are transmitted and spread primarily by tsetse flies (Glossina spp.) in sub-Saharan Africa. The animal disease poses significant challenges to agropastoral systems, including in Kenya, where 38 out of 47 counties are infested with eight species of Glossina. Climate change and human activities can also aggravate these infestations, putting rural-scale farmers who rely on agropastoral systems at a greater risk. Geographical gaps in existing entomological datasets limit a comprehensive understanding of tsetse fly distribution across the country, especially amid rapid landscape dynamics.

Methods

This study aimed to predict the spatial distribution of tsetse flies habitat in Kenya using recent entomological data (i.e., tsetse fly occurrence records), satellite-derived environmental variables, landscape structure, demographic indicators, and species-distribution modeling techniques. We applied four machine learning (ML) algorithms—random forest (RF), support vector machines (SVM), maximum entropy (MaxEnt), and generalized linear models (GLM)—to predict tsetse flies habitat suitability. Additionally, we developed ensemble models that combine the predictive power of the four algorithms. Predictions were made at the genus level (Glossina spp.) and the species level for one priority species (Glossina pallidipes).

Results

The models performed well with true skill statistic (TSS) and area under the curve (AUC) metric measures of 0.67 and 0.88 for Glossina spp. and 0.85 and 0.96 for G. pallidipes, respectively. The predictions indicated an estimated potential suitable area of about 26% of Kenya for Glossina spp. and 9% for G. pallidipes. Tsetse fly habitat suitability was positively correlated with increased sheep density, normalized difference vegetation index, and soil moisture. However, suitability declined when the maximum land surface temperature (LST) exceeded 40 °C and elevation increased above 400 m.

Conclusions

These findings can help improve the targeting and, hence, the cost-effectiveness of surveillance and ultimately support an evidence-based progressive control of tsetse flies infestation in Kenya.

Graphical Abstract