Scholarlink

<p>Desert locusts threaten vast regions of agriculture, food security and economic stability. The devastating 2019–2020 outbreak highlighted the critical need for early detection through reliable habitat modelling. This study adopts a framework that employs <i>Maxent</i> to model desert locust habitat suitability for the year 2019 and project it to the year 2020, identifying key environmental drivers and evaluating short-term interannual transferability as an alternative to traditional longer baseline periods. The optimized model achieved an area under the receiver operating characteristic curve value of 0.7396, with land surface temperature and normalized difference vegetation index emerging as the strongest predictors, followed by soil type, soil moisture, precipitation and temperature. When projected to 2020, the model showed that 81.34% of occurrence points fell within the predicted suitable regions, indicating good agreement with independent records. Model evaluation using pseudo-absences produced an overall accuracy of 75.10%, a sensitivity of 63.98%, a specificity of 86.21%, and a True Skill Statistic (TSS) of 0.5019, indicating moderate but meaningful temporal transferability. Compared with the default Maxent configuration, the optimized model provided a better balance between predictive accuracy and generalizability. These findings underscore the importance of careful model calibration/tuning, dynamic environmental variables and high-resolution data for enhancing predictive capacity, as well as the critical need for validation with independent datasets when projecting to future time periods. Such improved habitat modelling enables proactive intervention strategies to mitigate locust outbreaks and protect agricultural livelihoods in vulnerable regions.</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 Maxent-based framework for modelling desert locust habitat suitability: balancing accuracy and temporal transferability

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • N. A. Anjita,
  • Biplab Banerjee,
  • Indu Jayaluxmi

摘要

Desert locusts threaten vast regions of agriculture, food security and economic stability. The devastating 2019–2020 outbreak highlighted the critical need for early detection through reliable habitat modelling. This study adopts a framework that employs Maxent to model desert locust habitat suitability for the year 2019 and project it to the year 2020, identifying key environmental drivers and evaluating short-term interannual transferability as an alternative to traditional longer baseline periods. The optimized model achieved an area under the receiver operating characteristic curve value of 0.7396, with land surface temperature and normalized difference vegetation index emerging as the strongest predictors, followed by soil type, soil moisture, precipitation and temperature. When projected to 2020, the model showed that 81.34% of occurrence points fell within the predicted suitable regions, indicating good agreement with independent records. Model evaluation using pseudo-absences produced an overall accuracy of 75.10%, a sensitivity of 63.98%, a specificity of 86.21%, and a True Skill Statistic (TSS) of 0.5019, indicating moderate but meaningful temporal transferability. Compared with the default Maxent configuration, the optimized model provided a better balance between predictive accuracy and generalizability. These findings underscore the importance of careful model calibration/tuning, dynamic environmental variables and high-resolution data for enhancing predictive capacity, as well as the critical need for validation with independent datasets when projecting to future time periods. Such improved habitat modelling enables proactive intervention strategies to mitigate locust outbreaks and protect agricultural livelihoods in vulnerable regions.