Abstract: The annual variation, diurnal variation characteristics of atmospheric pollutant and its relationship with meteorological factor were studied by using the monitoring data of two major air pollutants （PM_2.5 and PM₁₀） from environmental monitoring stations in Tianmen city from January 1, 2017 to May 30, 2020. Results showed that the relationship between PM_2.5 and PM₁₀ concentrations and daily average air temperature increased first and then decreased. However the concentration of particulate matter below 10 ℃ increased with the ascend of air temperature, decreased with the increased of air temperature; Precipitation had obvious removal effects on PM_2.5 and PM₁₀ pollutants, meanwhile the concentration of PM_2.5 and PM₁₀ decreased by 0.72 μg/m³ and 1.22 μg/m³ respectively with each increase of 1 mm. The relative humidity was 30%~70%, the concentrations of PM_2.5 and PM₁₀ increased with the increase of relative humidity; When the relative humidity was 70%~100%, the concentration decreases with the increase of relative humidity; The concentrations of PM_2.5 and PM₁₀ decreased significantly with the increase of wind speed. Compare and analyze the prediction ability of multiple nonlinear regression prediction model, multiple linear regression model and adaptive linear neural network model, the multiple nonlinear regression model established by the daily mass concentration of PM_2.5, PM₁₀ and the daily temperature, precipitation and wind speed was suitable for the prediction of the mass concentration of particulate pollutants in Tianmen city.

Key words: PM_2.5, PM₁₀, meteorological conditions, multiple nonlinear regressions, Tianmen city