The Pollution Situation Of Atmospheric Particulate Matter In Nanjing Area And Its Relationship With Meteorological Factors

Haze is a more serious phenomenon of air pollution.Although the cause of the haze is very complex, but the main culprit is inhalable particles PM10 and PM2.5.Its evolution is closely related to meteorological factors and other pollution gases.In order to better reflect its changing trend, the characteristics of the particulate matter analysis, cause analysis and simulation prediction are carried out in this paper.1. Through collecting PM2.5 and PM10 mass concentration of 13 monitoring points which can represent all districts of Nanjing area, analysis of Nanjing particulate matter in space and time,from the spatial distribution of the whole, Nanjing urban pollution levels greater than the outskirts of the city and is larger than that of the suburban areas; little difference in PM2.5 in 13 monitoring during the study period, the difference of PM10 is obvious. In Spring the most serious monitoring point of PM2.5 pollution is Maigaoqiao, the Olympic Sports Center is the most serious monitoring points of PM10 pollution, in winter Olympic Sports Center are the most serious of the monitoring points of both PM2.5 and PM10 pollution. Time distribution of PM10 and PM2.5 month average concentration were the highest in May and lowest in January, standard ratio of the number of days is the opposite:in May, the highest and lowest in January; The 24 hours variation curve of particulate matter concentrations is roughly two peaks and two valleys, and winter concentrations of particulate matter changing fluctuation curve and bimodal double Valley feature is more obvious; PM2.5 and PM10 in correlationin the same season is very good,the winter of the correlation coefficient is 0.948, the spring is 0.897.Almost no correlation between PM2.5 and PM10 across a quarter.2.Through data collection and analysis of PM2.5 and single meteorological factors and other pollution gas correlation, the results show that PM2.5 is most polluted when the relative humidity is up to about 75%; PM2.5 and wind speed were negatively correlated, and the instantaneous effect of daily maximum wind speed is greater than the lasting effect of average daily wind speed;The amount of PM2.5 clearance increases with the increase of rainfall quantity which is between 1-5 mm and The amount of PM2.5 and initial concentrations of PM2.5 were positively correlated, with increasing of the initial concentration, removals is greater; PM2.5 and CO, NO2, SO2 and other pollution gases was significantly positive correlation, but O3 and visibility showed a negative correlation.3. Based on the data analysis, the seasonal stepwise regression analysis model was established quantitative analysis of the causes of PM2.5, quantitative regression model for Winter is y1=-18.103+105.811 xco-3.236x rainfall-2.755X maxmum temperature +2.167x minimum temperature; quantitative regression model for spring is y2=-62.713+52.343xco-0.664x rainfail+0.584x re,ative humidity-0.208XO3(8)+0.220XNo2; regression model simulation results show that significant factors affecting PM2.5 concentration in winter are carbon monoxide, Tmin, Tmax, rainfall; The significant factors that affect the PM2.5 concentration in spring are CO, rainfall, humidity, O3(8h),NO2; Therefore, it is important to effectively control the concentration of these factors to reduce the concentration of PM2.5.4. BP neural network was used to establish a simulation method to predict the change of PM2.5 concentration, With the aid of MATLAB toolbox, Call some function to establish BP neural network in the winter and spring season, Screening of other pollutants and meteorological factors in the strong correlation with PM2.5 as the input neurons then training and simulation.The validation of BP neural network of PM2.5 concentration simulation model predicted that the true values are in good agreement with the simulation data. The prediction accuracy of the winter simulation model is 80.12%, and 78.99% for the prediction accuracy of the spring simulation model, absolute error valuing at -20μg.cm03-20μg.cm-3, relative error within the-30%-30%, Only individual points out of this range. By comparing the predicted value with the real value, it is proved that the method can better predict the PM2.5 concentration in a certain error range.