Combinando árboles de decisión y métodos geoestadísticos para estimar la relación del mercurio total con propiedades químicas del suelo en el distrito minero de San Joaquín, Querétaro, México

Abstract

The degree of contamination of a site, due to the deposition of mercury from mine tailings, is directly related to soil properties since these determine the mobility of pollutants. The objective of this work was to determine the relationship between soil properties and total mercury, using classification and regression tree analysis, and geostatistical methods. A total of 102 samples were collected to measure total mercury and 10 soil-properties analysis. The regression analysis with decision trees showed four relationships between total mercury and chemical properties: 1) pH > 8.5: total mercury > 64 mg/kg; 2) organic matter > 5.5%: total mercury > 19 mg/kg; 3) bicarbonates > 256 mg/kg: total mercury > 77 mg/ kg; and 4) redox potential > 171 mV: total mercury < 49 mg/kg. In addition, a relationship was found between the regression decision trees and the maps elaborated with the geostatistical methods. For example, in places with higher concentrations of total mercury, there is a greater amount of organic matter, pH, and bicarbonates; however, there was higher redox potential in the lower concentrations of total mercury. In conclusion 1) high organic matter contents favor the retention of mercury, 2) neutral-alkaline pH favors low mobility of mercury, 3) bicarbonates favors mercury stabilization, and 4) intermediate redox potential provides soil stability to mercury pollution.