Scholarlink

Background to the study <p>Remote sensing, geographic information science, and machine learning are techniques for modeling the forest ecosystem and analyzing its changes, losses, and risk areas. Forest losses threaten the forest ecosystem. It increases emission of greenhouse gas, soil erosion, local climate, water cycle disruption, and biodiversity loss. In light of this, it has become crucial to manage the forest using continuous cover forestry (CCF) as a nature-based solution (NBS), which is a sustainable forest management strategy, to achieve sustainable development.</p> Objectives <p>The study models forest cover losses, assigns weights to thematic variables, and adopts CCF as an NBS to mitigate these losses in Nigeria’s southeastern states during the periods of 2000–2024 and 2030–2050.</p> Methods <p>The study utilized support vector machine (SVM), random forest (RF), and artificial neural network (ANN) as machine learning (ML) classifiers, along with Landsat enhanced thematic mapper plus (ETM+) for 2000 and 2012 and operational land imager (OLI) for 2024 as remote sensing data to map and model forest cover changes in Nigeria’s southeastern states from 2000 to 2024 and through 2030 to 2050. AHP was adopted to assign weight to the study’s thematic variables (soil suitability, slope, and distances to rivers, water bodies, settlements, transport, and agricultural land) for normalization. Accordingly, these variables were integrated into the GIS environment for a weighted overlay analysis to create the final forest cover losses suitability map. The Kappa statistics, agreement/disagreement analysis, and the receiver operating characteristics (ROC) were adopted for validation accuracy.</p> Results <p>The results demonstrate that the forest cover class had declined while the built-up and water body classes had increased progressively throughout the study from 2000 to 2024 for SVM and RF classifiers. Other study classes, including the cultivated and bare land, had also changed for SVM and RF during the study period. The simulation analysis reveals that by 2030–2050, forest cover class will decrease, while other study classes will either increase or decrease and will have a detrimental impact on the forest cover classes.</p> Conclusion <p>The study’s findings demonstrate that Nigeria’s southeastern states have been experiencing forest cover losses during the past 24 years and are anticipated to decrease further from 2030 to 2050. Since the forest class has been found to support sustainable development, it therefore means that the study’s finding will either limit or hinder it; hence, the study recommends CCF as NBS and as crucial measures to combat forest cover losses in order to achieve sustainability of resources in the study area.</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

Machine learning and geospatial modeling of forest loss, drivers, and risk areas: advancing continuous cover forestry as a nature-based solution

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月14日

查看全文
  • Mfoniso Asuquo Enoh,
  • Okwudili John Ugwu,
  • Baywood Ceciliar Nkechi,
  • Njoku Richard Ebere,
  • Okoroji Anthony Chibuzor

摘要

Background to the study

Remote sensing, geographic information science, and machine learning are techniques for modeling the forest ecosystem and analyzing its changes, losses, and risk areas. Forest losses threaten the forest ecosystem. It increases emission of greenhouse gas, soil erosion, local climate, water cycle disruption, and biodiversity loss. In light of this, it has become crucial to manage the forest using continuous cover forestry (CCF) as a nature-based solution (NBS), which is a sustainable forest management strategy, to achieve sustainable development.

Objectives

The study models forest cover losses, assigns weights to thematic variables, and adopts CCF as an NBS to mitigate these losses in Nigeria’s southeastern states during the periods of 2000–2024 and 2030–2050.

Methods

The study utilized support vector machine (SVM), random forest (RF), and artificial neural network (ANN) as machine learning (ML) classifiers, along with Landsat enhanced thematic mapper plus (ETM+) for 2000 and 2012 and operational land imager (OLI) for 2024 as remote sensing data to map and model forest cover changes in Nigeria’s southeastern states from 2000 to 2024 and through 2030 to 2050. AHP was adopted to assign weight to the study’s thematic variables (soil suitability, slope, and distances to rivers, water bodies, settlements, transport, and agricultural land) for normalization. Accordingly, these variables were integrated into the GIS environment for a weighted overlay analysis to create the final forest cover losses suitability map. The Kappa statistics, agreement/disagreement analysis, and the receiver operating characteristics (ROC) were adopted for validation accuracy.

Results

The results demonstrate that the forest cover class had declined while the built-up and water body classes had increased progressively throughout the study from 2000 to 2024 for SVM and RF classifiers. Other study classes, including the cultivated and bare land, had also changed for SVM and RF during the study period. The simulation analysis reveals that by 2030–2050, forest cover class will decrease, while other study classes will either increase or decrease and will have a detrimental impact on the forest cover classes.

Conclusion

The study’s findings demonstrate that Nigeria’s southeastern states have been experiencing forest cover losses during the past 24 years and are anticipated to decrease further from 2030 to 2050. Since the forest class has been found to support sustainable development, it therefore means that the study’s finding will either limit or hinder it; hence, the study recommends CCF as NBS and as crucial measures to combat forest cover losses in order to achieve sustainability of resources in the study area.