Scholarlink

<p>Analyzing the spatial variability of wetland Soil Organic Carbon (SOC) content is vital for evaluating soil quality and associated factors including climate change resilience, structural stability, nutrient cycling, and achieving carbon neutrality. However, site-specific evaluations remain limited, creating uncertainty in refining global SOC assessments for environmental sustainability. In this study, 60 randomized soil samples (0–20&#xa0;cm) were collected from wetlands in Ibadan, Nigeria and SOC content was assessed using the Walkley–Black method. These samples, in conjunction with environmental covariates derived from index variables including Normalized difference vegetation Index (NDVI), Enhanced vegetation index (EVI), Soil-adjusted vegetation index (SAVI), Digital elevation model (DEM), aspect, curvature, and soil types were integrated into Random Forest (RF) algorithm to digitally predict distribution of SOC. Model performance was evaluated using coefficient of determination (R<sup>2</sup>), root mean squared error (RMSE), and mean absolute error (MAE). The results showed that NDVI, EVI, and SAVI were the most significant predictors of SOC distribution. In contrast, terrain derivatives such as DEM, aspect, curvature, and soil type contributed less to the prediction accuracy. The RF model performed well, yielding an RMSE of 0.35, MAE of 0.30, and R<sup>2</sup> of 0.60. The predicted SOC for the study area showed distinctive spatial pattern. Central areas exhibit higher SOC content, while lower levels are found throughout the areas. This spatial variation provides a nuanced understanding of SOC distribution, which is critical for effective soil management and environmental planning in the studied region. This study underscores the pressing need for reforestation, sustainable land management practices, and a balance between development and environmental sustainability, particularly in regions vulnerable to climate change. Future research should consider deeper soil sampling to capture vertical SOC variability comprehensively. Moreover, exploring alternative statistical approaches, like the variogram model and advanced machine learning models such as deep neural networks, could provide deeper insights and improve SOC prediction accuracy.</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

Random forest algorithm and remote sensing techniques for wetland soil organic carbon prediction towards environmental sustainability

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月8日

查看全文
  • Anthony Olamiji Tobore,
  • Ugonna C. Nkwunonwo,
  • Sadiq Adisa Abdussalaam,
  • Ganiyu Oyerinde,
  • Victor Manna Samson,
  • Adeniran Quam Adeoba,
  • Hakeem Oludare Adedeji,
  • Ayodeji Matthew Odebode,
  • Ayodele O. Adelana

摘要

Analyzing the spatial variability of wetland Soil Organic Carbon (SOC) content is vital for evaluating soil quality and associated factors including climate change resilience, structural stability, nutrient cycling, and achieving carbon neutrality. However, site-specific evaluations remain limited, creating uncertainty in refining global SOC assessments for environmental sustainability. In this study, 60 randomized soil samples (0–20 cm) were collected from wetlands in Ibadan, Nigeria and SOC content was assessed using the Walkley–Black method. These samples, in conjunction with environmental covariates derived from index variables including Normalized difference vegetation Index (NDVI), Enhanced vegetation index (EVI), Soil-adjusted vegetation index (SAVI), Digital elevation model (DEM), aspect, curvature, and soil types were integrated into Random Forest (RF) algorithm to digitally predict distribution of SOC. Model performance was evaluated using coefficient of determination (R2), root mean squared error (RMSE), and mean absolute error (MAE). The results showed that NDVI, EVI, and SAVI were the most significant predictors of SOC distribution. In contrast, terrain derivatives such as DEM, aspect, curvature, and soil type contributed less to the prediction accuracy. The RF model performed well, yielding an RMSE of 0.35, MAE of 0.30, and R2 of 0.60. The predicted SOC for the study area showed distinctive spatial pattern. Central areas exhibit higher SOC content, while lower levels are found throughout the areas. This spatial variation provides a nuanced understanding of SOC distribution, which is critical for effective soil management and environmental planning in the studied region. This study underscores the pressing need for reforestation, sustainable land management practices, and a balance between development and environmental sustainability, particularly in regions vulnerable to climate change. Future research should consider deeper soil sampling to capture vertical SOC variability comprehensively. Moreover, exploring alternative statistical approaches, like the variogram model and advanced machine learning models such as deep neural networks, could provide deeper insights and improve SOC prediction accuracy.