Search

The quality and safety of soil are crucial for ensuring social and economic development and providing contaminant-free food. The availability and quality of soil data, particularly for multiple metals and metalloids, are often insufficient for comprehensive analysis. Soil formation and the distribution of metals are shaped by various factors such as geology, climate, topography, and human activities, making accurate modeling highly challenging. Additionally, agricultural intensification, urban expansion, road construction, and mining activities frequently result in soil pollution, posing serious risks to ecosystems and human health. This study aims to integrate diverse geospatial datasets with machine learning for high resolution soil contamination mapping (10 m spatial resolution) in a major agricultural region of Peruvian highlands. This study mapped 25 elements (Ca, Mg, Sr, Ba, Be, K, Na, As, Sb, Se, Tl, Cd, Zn, Al, Pb, Hg, Cr, Ni, Cu, Mo, Ag, Fe, Co, Mn, V) in the Peruvian Mantaro Valley using a training dataset of 109 topsoil samples combined with various geospatial datasets (remote sensing, climate, topography, soil data, and distance). The model provided satisfactory results in predicting the spatial distribution of the selected elements, with R2 values ranging from 0.6 to 0.9 for most elements. Edaphic, climate, and topographic covariates were the most significant predictors, particularly for croplands near rivers, whereas spectral variables were less important. The results reveal As, Pb, and Cd concentrations significantly above permissible limits, highlighting urgent health risks. These findings suggest that it is feasible to identify polluted soils and improve regulations based on widely available geospatial datasets with minimal training data. The study contributes to the development of models to assess the impact of pollutants on environmental and human health in the short-to-medium term, emphasizing the need for further research on the translocation of toxic metals into food crops and the implications for public health.

The spatial heterogeneity of soil properties has a significant impact on crop growth, making it difficult to adopt site-specific crop management practices. Traditional laboratory-based analyses are costly, and data extrapolation for mapping soil properties using high-resolution imagery becomes a computationally expensive procedure, taking days or weeks to obtain accurate results using a desktop workstation. To overcome these challenges, cloud-based solutions such as Google Earth Engine (GEE) have been used to analyze complex data with machine learning algorithms. In this study, we explored the feasibility of designing and implementing a digital soil mapping approach in the GEE platform using high-resolution reflectance imagery derived from a thermal infrared and multispectral camera Altum (MicaSense, Seattle, WA, USA). We compared a suite of multispectral-derived soil and vegetation indices with in situ measurements of physical-chemical soil properties in agricultural lands in the Peruvian Mantaro Valley. The prediction ability of several machine learning algorithms (CART, XGBoost, and Random Forest) was evaluated using R2, to select the best predicted maps (R2 > 0.80), for ten soil properties, including Lime, Clay, Sand, N, P, K, OM, Al, EC, and pH, using multispectral imagery and derived products such as spectral indices and a digital surface model (DSM). Our results indicate that the predictions based on spectral indices, most notably, SRI, GNDWI, NDWI, and ExG, in combination with CART and RF algorithms are superior to those based on individual spectral bands. Additionally, the DSM improves the model prediction accuracy, especially for K and Al. We demonstrate that high-resolution multispectral imagery processed in the GEE platform has the potential to develop soil properties prediction models essential in establishing adaptive soil monitoring programs for agricultural regions.