Due to rapid economic development, high levels of potentially harmful elements and heavy metals are continuously being released into the brown coal mining dumpsites of the Czech Republic. Elevated metal contents in soils not only dramatically impact the soil quality, but also due to their persistent nature and long biological half-lives, contaminant elements can accumulate in the food chain and can eventually endanger human health. Conventional methods for investigating potentially harmful element contamination of soil based on raster sampling and chemical analysis are time consuming and relatively expensive. Visible and Near-Infrared (Vis-NIR) diffuse reflectance spectroscopy provides a rapid and inexpensive tool to simultaneously and accurately predict various soil properties. In this study concentrations of Manganese (Mn), Copper (Cu), Cadmium (Cd), Zinc (Zn), Iron (Fe), Lead (Pb) and Arsenic (As) in soil samples from fields near the brow coal mining dumpsites in the Czech Republic were chemically analyzed and the suitability of Vis-NIR diffuse reflectance spectroscopy for predicting their occurrence was evaluated. Soil spectral reflectance was measured with an ASD FieldSpec 3 spectroradiometer (Analytical Spectral Devices, Inc., USA) under laboratory conditions and the correlations between seven toxic elements and soil diffuse reflectance spectra were studied. Partial Least Square Regression (PLSR) and Support Vector Machine Regression (SVMR) models were constructed to relate soil contaminants data to the reflectance spectral data by applying first and second derivatives preprocessing strategies. Then, the performance of Vis-NIR calibration models was evaluated by Residual Prediction Deviation (RPD) and coefficients of determination (R²). Based on the correlation patterns with reflectance spectra, the seven studied potentially toxic elements were categorized into two or three groups. Moreover, according to the criteria of minimal RPD and maximal R², the first derivative and SVMR models provided more accurate prediction models for soil contaminants than PLSR models which were more feasible to predict the toxic metal levels in agricultural soils. Overall, this study indicated that the Vis-NIR reflectance spectroscopy technique combined with a continuously enriched soil spectral library as well as a suitable chemometric indicator could be a nondestructive alternative for monitoring of the soil environment. Because soil properties in contaminated areas generally show strong variation, a comparatively large number of calibrating samples, which are variable enough and uniformly distributed, are necessary to create more accurate and robust Vis-NIR reflectance spectroscopy calibration models. Future studies with real-time remote sensing data and field measurements are also strongly recommended.Vis-NIR) diffuse reflectance spectroscopy provides a rapid and inexpensive tool to simultaneously and accurately predict various soil properties. In this study concentrations of Manganese (Mn), Copper (Cu), Cadmium (Cd), Zinc (Zn), Iron (Fe), Lead (Pb) and Arsenic (As) in soil samples from fields near the brow coal mining dumpsites in the Czech Republic were chemically analyzed and the suitability of Vis-NIR diffuse reflectance spectroscopy for predicting their occurrence was evaluated. Soil spectral reflectance was measured with an ASD FieldSpec 3 spectroradiometer (Analytical Spectral Devices, Inc., USA) under laboratory conditions and the correlations between seven toxic elements and soil diffuse reflectance spectra were studied. Partial Least Square Regression (PLSR) and Support Vector Machine Regression (SVMR) models were constructed to relate soil contaminants data to the reflectance spectral data by applying first and second derivatives preprocessing strategies. Then, the performance of Vis-NIR calibration models was evaluated by Root Mean Square Error (RMSE) and coefficients of determination in cross-validation (R² cv). Based on the correlation patterns with reflectance spectra, the seven studied potentially toxic elements were categorized into two groups. The prediction accuracy for group I (As, Cd, Fe, Cu, Zn) was higher than that for group II (Mn, Pb). Moreover, according to the criteria of minimal RMSE and maximal R² cv, the first derivative and SVMR models provided more accurate prediction models for soil contaminants than PLSR models which were more feasible to predict the toxic metal levels in agricultural soils. Overall, this study indicated that the Vis-NIR reflectance spectroscopy technique combined with a continuously enriched soil spectral library as well as a suitable chemometric indicator could be a nondestructive alternative for monitoring of the soil environment. Because soil properties in contaminated areas generally show strong variation, a comparatively large number of calibrating samples, which are variable enough and uniformly distributed, are necessary to create more accurate and robust Vis-NIR reflectance spectroscopy calibration models. Future studies with real-time remote sensing data and field measurements are also strongly recommended.