Scholarlink

<p>Soil salinization is a major constraint to sustainable agricultural production, particularly in semi-arid regions where evapotranspiration exceeds rainfall. This study applied the Quantile Regression Forest (QRF) method to model and predict soil electrical conductivity (EC) across five depth intervals (0–25, 25–50, 50–75, 75–100, and 100–125&#xa0;cm) in a key agricultural area of Shaanxi Province, China. A comprehensive set of environmental predictors derived from multispectral remote sensing, topographic features, and climatic variables was used to explain the spatial and vertical distribution of salinity. Remote sensing predictors were derived from Landsat-8 OLI imagery (~ 30&#xa0;m spatial resolution) together with terrain attributes extracted from a digital elevation model (DEM, ~ 30&#xa0;m spatial resolution). Principal component analysis identified vegetation indices (NDVI, NDWI), terrain attributes (slope height, stand height), and temperature seasonality as the most influential factors controlling EC variability. The QRF model achieved high accuracy in the surface layer (R2 = 0.81; RPIQ = 3.05) and acceptable performance at deeper layers (R2 = 0.59; RPIQ = 1.96), with cross-validation confirming robust results, particularly for the 0–25&#xa0;cm depth (R2 = 0.86; RPIQ = 5.54). Spatial maps highlighted zones of high EC concentrated in central and northeastern parts of the study area, mostly in low-slope and poorly drained regions, while uncertainty maps revealed areas with complex salinity dynamics. The uncertainty analysis further showed that most observed EC values were captured within the 90% prediction interval, as indicated by the prediction interval coverage probability (PICP). Overall, combining QRF with depth-harmonized environmental covariates provides an effective and uncertainty-aware approach for multi-depth digital soil salinity mapping, offering valuable guidance for precision salinity management and sustainable land use planning in semi-arid agroecosystems.</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 soil salinity across depths using quantile regression forests in Semi-Arid Farmlands of Shaanxi, China

  • 期刊论文

  • 发表日期：

    2026年3月29日

查看全文
  • Lin Lin,
  • Ming Zhang

摘要

Soil salinization is a major constraint to sustainable agricultural production, particularly in semi-arid regions where evapotranspiration exceeds rainfall. This study applied the Quantile Regression Forest (QRF) method to model and predict soil electrical conductivity (EC) across five depth intervals (0–25, 25–50, 50–75, 75–100, and 100–125 cm) in a key agricultural area of Shaanxi Province, China. A comprehensive set of environmental predictors derived from multispectral remote sensing, topographic features, and climatic variables was used to explain the spatial and vertical distribution of salinity. Remote sensing predictors were derived from Landsat-8 OLI imagery (~ 30 m spatial resolution) together with terrain attributes extracted from a digital elevation model (DEM, ~ 30 m spatial resolution). Principal component analysis identified vegetation indices (NDVI, NDWI), terrain attributes (slope height, stand height), and temperature seasonality as the most influential factors controlling EC variability. The QRF model achieved high accuracy in the surface layer (R2 = 0.81; RPIQ = 3.05) and acceptable performance at deeper layers (R2 = 0.59; RPIQ = 1.96), with cross-validation confirming robust results, particularly for the 0–25 cm depth (R2 = 0.86; RPIQ = 5.54). Spatial maps highlighted zones of high EC concentrated in central and northeastern parts of the study area, mostly in low-slope and poorly drained regions, while uncertainty maps revealed areas with complex salinity dynamics. The uncertainty analysis further showed that most observed EC values were captured within the 90% prediction interval, as indicated by the prediction interval coverage probability (PICP). Overall, combining QRF with depth-harmonized environmental covariates provides an effective and uncertainty-aware approach for multi-depth digital soil salinity mapping, offering valuable guidance for precision salinity management and sustainable land use planning in semi-arid agroecosystems.