Scholarlink

<p>Protected areas are critical for global biodiversity conservation, yet they are increasingly threatened by human activities such as settlement growth and farmland expansion. Conventional methods of monitoring these land-use changes often rely on labor-intensive field surveys or coarse-resolution satellite data, which can be slow, costly, and insufficient for detecting fine-scale encroachments. This study employs high-resolution satellite imagery (0.3–0.55&#xa0;m) and deep learning to develop an automated framework for detecting and quantifying settlement and farmland expansion within Bale Mountain National Park (BMNP), Ethiopia. High-resolution Pleiades imagery from 2019 to 2024 enabled the identification of small, fragmented settlements and farmland plots that are typically missed by coarser datasets. Deep learning–based segmentation and detection models were applied to map built-up areas and farmland and to assess human-induced land-use changes within the park boundaries. Results indicate a nearly tenfold increase in settlement detections between 2019 and 2024 (from 388 to 3,506), while farmland expanded from 10.69&#xa0;km² to 15.16&#xa0;km², representing a 42% increase. These findings reveal intensifying anthropogenic pressure on BMNP and highlight the effectiveness of AI-assisted high-resolution monitoring for early detection of encroachment. Key limitations include the reliance on limited temporal snapshots and potential classification uncertainties in spectrally heterogeneous landscapes. Despite these constraints, the proposed approach provides a scalable and timely decision-support tool for park authorities, enabling targeted enforcement, improved land-use planning, and proactive conservation management in ecologically sensitive protected areas. </p> Graphical Abstract <p></p> <p>The graphical abstract presents a deep learning framework to accurately detect and quantify settlement and farmland encroachment in Bale Mountain National Park (BMNP), Ethiopia. Utilizing high-resolution Pléiades satellite imagery (0.55&#xa0;m and 0.30&#xa0;m), our approach integrates a Mask R-CNN model for fine-scale settlement detection and a SAM-LoRA model (for farmland segmentation) into a cohesive workflow, as shown in the diagram. This advanced method overcomes the limitations of traditional monitoring by enabling precise, scalable, and near real-time assessment. The results reveal intensifying threats to the protected area: between 2019 and 2024, settlement detections increased nearly tenfold, and farmland expanded by 42% (from 10.69 km<sup>2</sup> to 15.16 km<sup>2</sup>). This integration of high-resolution imagery with AI offers a powerful tool for evidence-based and proactive conservation interventions, significantly enhancing the ability to detect early signs of anthropogenic pressure.</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

Mapping Threats To Protected Areas: Machine Learning–Based Detection of Land Encroachment in Bale Mountains National Park, Ethiopia

  • 期刊论文

  • 发表日期：

    2026年3月31日

查看全文
  • Yacob T. Tesfaldet,
  • Getachew Mulualem,
  • Nji T. Ndeh,
  • Alireza Daneshkhah

摘要

Protected areas are critical for global biodiversity conservation, yet they are increasingly threatened by human activities such as settlement growth and farmland expansion. Conventional methods of monitoring these land-use changes often rely on labor-intensive field surveys or coarse-resolution satellite data, which can be slow, costly, and insufficient for detecting fine-scale encroachments. This study employs high-resolution satellite imagery (0.3–0.55 m) and deep learning to develop an automated framework for detecting and quantifying settlement and farmland expansion within Bale Mountain National Park (BMNP), Ethiopia. High-resolution Pleiades imagery from 2019 to 2024 enabled the identification of small, fragmented settlements and farmland plots that are typically missed by coarser datasets. Deep learning–based segmentation and detection models were applied to map built-up areas and farmland and to assess human-induced land-use changes within the park boundaries. Results indicate a nearly tenfold increase in settlement detections between 2019 and 2024 (from 388 to 3,506), while farmland expanded from 10.69 km² to 15.16 km², representing a 42% increase. These findings reveal intensifying anthropogenic pressure on BMNP and highlight the effectiveness of AI-assisted high-resolution monitoring for early detection of encroachment. Key limitations include the reliance on limited temporal snapshots and potential classification uncertainties in spectrally heterogeneous landscapes. Despite these constraints, the proposed approach provides a scalable and timely decision-support tool for park authorities, enabling targeted enforcement, improved land-use planning, and proactive conservation management in ecologically sensitive protected areas.

Graphical Abstract

The graphical abstract presents a deep learning framework to accurately detect and quantify settlement and farmland encroachment in Bale Mountain National Park (BMNP), Ethiopia. Utilizing high-resolution Pléiades satellite imagery (0.55 m and 0.30 m), our approach integrates a Mask R-CNN model for fine-scale settlement detection and a SAM-LoRA model (for farmland segmentation) into a cohesive workflow, as shown in the diagram. This advanced method overcomes the limitations of traditional monitoring by enabling precise, scalable, and near real-time assessment. The results reveal intensifying threats to the protected area: between 2019 and 2024, settlement detections increased nearly tenfold, and farmland expanded by 42% (from 10.69 km2 to 15.16 km2). This integration of high-resolution imagery with AI offers a powerful tool for evidence-based and proactive conservation interventions, significantly enhancing the ability to detect early signs of anthropogenic pressure.