Prediction of the concentration of cadmium in agricultural soil in the Czech Republic using legacy data, preferential sampling, Sentinel-2, Landsat-8, and ensemble models

Inversion of heavy metal content in soil using hyperspectral characteristic bands-based machine learning method

The global concern regarding the adverse effects of heavy metal pollution in soil has grown significantly. Accurate prediction of heavy metal content in soil is crucial for environmental protection. This study proposes an inversion analysis method for heavy metals (As, Cd, Cr, Cu, Ni, Pb) in soil based on hyperspectral and machine learning algorithms for 21 soil reference materials from multiple provinces in China. On this basis, an integrated learning model called Stacked RF (the base model is XGBoost, LightGBM, CatBoost, and the meta-model is RF) was established to perform soil heavy metal inversion. Specifically, three popular algorithms were initially employed to preprocess the spectral data, then Random Forest (RF) was used to select the best feature bands to reduce the impact of noise, finally Stacking and four basic machine learning algorithms were used to establish comparisons and analysis of inversion model. Compared with traditional machine learning methods, the stacking model showcases enhanced stability and superior accuracy. Research results indicate that machine learning algorithms, especially ensemble learning models, have better inversion effects on heavy metals in soil. Overall, the MF-RF-Stacking model performed best in the inversion of the six heavy metals. The research results will provide a new perspective on the ensemble learning model method for soil heavy metal content inversion using data of hyperspectral characteristic bands collected from soil reference materials.

Digital exploration of selected heavy metals using Random Forest and a set of environmental covariates at the watershed scale

The current study was established for predicting some selected heavy metals (HMs) including Zn, Mn, Fe, Co, Cr, Ni, and Cu, by applying random forest (RF) and a set of environmental covariates at watershed scale. The objectives were to find out the most effective combination of variables and controlling factors on the variability of HMs in a semiarid watershed in central Iran. One hundred locations were selected in the given watershed in the hypercube manner and soil samples from a surface 0-20 cm depth and concentration of HMs and some soil properties were measured in the laboratory. Three scenarios of input variables were defined for HMs prediction. The results revealed that the first scenario (remote sensing + topographic attributes) explained about 27-34% of the variability in HMs. Inclusion of a thematic map to the scenario I, improved the prediction accuracy for all HMs. Scenario III (remote sensing data+ topographic attributes + soil properties) was the most efficient scenario for prediction of HMs with R² values ranging from 0.32 for Cu to 0.42 for Fe. Similarly, the lowest nRMSE was found for all HMs in scenario III, ranging from 0.271 for Fe to 0.351 for Cu. Among the soil properties, clay content and magnetic susceptibility were the most important variables, and also some remote sensing data (Carbonate index, Soil adjusted vegetation index, Band2, and Band7) and topographic attributes (mainly control soil redistribution along the landscape) were the most efficient variables for estimating HMs. We concluded that the RF model with a combination of remote sensing data, topographic attributes, and assisting of thematic maps such as land use in the studied watershed could reliably predict HMs content.