A machine learning-based framework for spatial soil health assessment integrating remote sensing indices and terrain data
· flipboard.com
Soil health (SH) plays a fundamental role in sustaining agricultural productivity and ecological stability, particularly in semi-arid environments where soils are highly vulnerable to degradation. This study developed a data driven framework for evaluating and spatially predicting SH by integrating field based soil indicators with the vegetation health index (VHI), surface water condition index (SWCI), and digital elevation model (DEM) data. A systematic sampling design with a 3 × 3 km grid was applied across an agricultural region in northwestern Iran, leading to the collection of 368 surface soil samples (0–30 cm). Key physical, chemical, and biological soil properties were measured and used to construct soil health indices (SHIs) based on both a total data set (TDS) and a minimum data set (MDS) derived through Principal component analysis (PCA). The integrated health index (IHI) and Nemoro quality index (NQI) were calculated to quantify soil condition. The results indicated considerable spatial variability in soil properties across the study area. PCA identified a limited number of influential indicators that effectively represented the overall variability of soil conditions. Soil texture particles, carbonate content, soil organic carbon (SOC), and biological activity emerged as major contributors influencing SH status. Random forest regression (RFR) model incorporating VHI, SWCI, and DEM variables successfully predicted the spatial distribution of SHIs. Among the evaluated models, the IHI-TDS model showed the strongest predictive performance during validation, with a coefficient of determination (R²) of 0.72, a root mean square error (RMSE) of 0.23, and a residual prediction deviation (RPD) of 3.04, indicating good reliability for spatial prediction. The generated maps with a spatial resolution of 30 m revealed clear patterns of SH variation across the landscape and identified areas with relatively lower SH and higher prediction uncertainty. These findings highlight the effectiveness of integrating field measurements, vegetation health information, and terrain data to improve regional SH assessment and spatial decision making in semi-arid agricultural systems.
Soil health (SH) plays a fundamental role in sustaining agricultural productivity and ecological stability, particularly in semi-arid environments where soils a...