Modeling of particulate matter concentrations from hauling operations in quarries using decision tree approach

Abstract Hauling operations are a primary contributor to the release of particulate matter (PM) in open-pit mining. While most existing concentration models focus on coal and iron mines, this study provides a novel contribution by developing PM estimation equations tailored to gypsum and limestone quarries near Sivas, Türkiye. Field measurements of PM fractions (TSP, PM 10 , PM 2.5 , PM 1 ) and thermal comfort parameters were collected simultaneously along truck routes. Roadbed samples were also analyzed for moisture and silt + clay content under laboratory conditions. A total of 3928 measurements were obtained using roadside monitoring stations. Both SPSS and WEKA (Waikato Environment for Knowledge Analysis) were used to create prediction models. While regression models in SPSS produced low adjusted coefficients of determination, the M5P decision tree algorithm in WEKA demonstrated significantly higher model fit statistics. This confirmed the algorithm’s capacity to handle complex, nonlinear relationships between PM concentrations and independent variables. The concentration of particulate matter is significantly influenced by both meteorological factors (air temperature, relative humidity, dew point, wind speed and air pressure) and operational parameters (truck mass, speed, number of wheels, roadbed moisture and silt + clay content). The integration of more atmospheric variables and vehicle specifications into the modeling process supports methodological improvements in the assessment of air quality in mining. The roadbed moisture, silt + clay content, and wind speed are the variables with the greatest influence on PM concentrations. The M5P decision tree algorithm, which is rarely used in PM concentration models, provides an innovative approach to developing local concentration estimates for non-coal surface mining.