Scholarlink

<p>Global digital soil mapping products such as SoilGrids are increasingly applied in data-scarce tropical regions, but their suitability for the rapidly urbanizing cities of Africa remains largely untested. This study introduces a convergence-of-evidence framework that evaluates the internal consistency of SoilGrids predictions against independent field vegetation surveys and satellite-derived vegetation indices along urban–peri-urban–rural gradients in Makurdi and Otukpo, Benue State, Nigeria. We extracted five soil properties (soil organic carbon, total nitrogen, bulk density, pH, and cation exchange capacity) from SoilGrids version 2.0 at the 0–5&#xa0;cm depth for 24 administrative wards, and compared the resulting spatial patterns with three independent evidence streams: field-measured plant diversity from 33 stratified plots (1,816 individuals across 52 species), satellite-derived vegetation indices (Normalized Difference Vegetation Index and Enhanced Vegetation Index from Landsat 8), and urbanization metrics (built-up percentage, nighttime light intensity, and population density). Because no direct soil samples were collected, the analysis evaluates the correspondence, rather than the measured accuracy, of SoilGrids predictions relative to independent ecological indicators. Vegetation indices showed highly significant differences among zones (NDVI:, , ; EVI:, , ), and field plant diversity declined consistently from rural to urban areas (Shannon index: 2.79, 2.51, and 2.58 for rural, peri-urban, and urban zones, respectively; ). A multiple regression model using three urbanization predictors explained 86.2 per cent of the variance in NDVI (, ), with population density as the dominant standardized driver (, ). By contrast, SoilGrids-derived soil properties showed no statistically significant zonal differences except for pH (,, ), and five of the most urbanized wards returned missing values because of impervious-surface masking. The divergence between the strong urbanization signals detected by vegetation indicators and the weak signals from SoilGrids indicates that the 250-metre global product does not appear to reproduce urbanization-driven soil gradients in these tropical secondary cities, although this inference concerns predictive consistency and would require direct soil sampling for a definitive accuracy assessment. These findings have methodological implications for how global digital soil mapping products are applied in urban ecological research in data-scarce regions and demonstrate the value of triangulating remotely sensed soil data with independent ecological observations before drawing inferences about urbanization effects on soil systems.</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

Evaluating soilGrids predictions along urban gradients in tropical africa using a convergence-of-evidence approach with field vegetation and remote sensing data from two nigerian cities

  • 期刊论文

  • 发表日期：

    2026年7月25日

查看全文
  • Ochoche Shaibu,
  • Anatoly Aleksandrovich Kirichuk

摘要

Global digital soil mapping products such as SoilGrids are increasingly applied in data-scarce tropical regions, but their suitability for the rapidly urbanizing cities of Africa remains largely untested. This study introduces a convergence-of-evidence framework that evaluates the internal consistency of SoilGrids predictions against independent field vegetation surveys and satellite-derived vegetation indices along urban–peri-urban–rural gradients in Makurdi and Otukpo, Benue State, Nigeria. We extracted five soil properties (soil organic carbon, total nitrogen, bulk density, pH, and cation exchange capacity) from SoilGrids version 2.0 at the 0–5 cm depth for 24 administrative wards, and compared the resulting spatial patterns with three independent evidence streams: field-measured plant diversity from 33 stratified plots (1,816 individuals across 52 species), satellite-derived vegetation indices (Normalized Difference Vegetation Index and Enhanced Vegetation Index from Landsat 8), and urbanization metrics (built-up percentage, nighttime light intensity, and population density). Because no direct soil samples were collected, the analysis evaluates the correspondence, rather than the measured accuracy, of SoilGrids predictions relative to independent ecological indicators. Vegetation indices showed highly significant differences among zones (NDVI:, , ; EVI:, , ), and field plant diversity declined consistently from rural to urban areas (Shannon index: 2.79, 2.51, and 2.58 for rural, peri-urban, and urban zones, respectively; ). A multiple regression model using three urbanization predictors explained 86.2 per cent of the variance in NDVI (, ), with population density as the dominant standardized driver (, ). By contrast, SoilGrids-derived soil properties showed no statistically significant zonal differences except for pH (,, ), and five of the most urbanized wards returned missing values because of impervious-surface masking. The divergence between the strong urbanization signals detected by vegetation indicators and the weak signals from SoilGrids indicates that the 250-metre global product does not appear to reproduce urbanization-driven soil gradients in these tropical secondary cities, although this inference concerns predictive consistency and would require direct soil sampling for a definitive accuracy assessment. These findings have methodological implications for how global digital soil mapping products are applied in urban ecological research in data-scarce regions and demonstrate the value of triangulating remotely sensed soil data with independent ecological observations before drawing inferences about urbanization effects on soil systems.