An Analysis On Meteorological Factors And Study On Predicting Air Pollution In Xi'an City

In this paper, the spatial and temporal variable characteristics of air pollution in Xi'an city is analyzed by using data of annual mean concentration of SO₂, NO₂ and PM₁₀ from 1998 to 2004 and daily reports of environmental monitoring from 2001 to 2004. A further statistical analysis is made about the influence of perception, sand-dust storms and other meteorological factors such as wind, thermal inversion features in convection layer as well as the influence of weather prognostic field by using the data of routine surface at the corresponding period, and synoptic chart and meteorological parameter in atmospherical boundary ,which accounts for the change of the concentration of air pollutants, based on which a model is set up for statistical prediction of PM10 concentration by adopting stepwise regression equation. A new PM10 concentration prediction model of artificial neural network (ANN) has been tried to establish based on main principal components. The differences between the two prediction models are also compared in the paper. The results are as follows:i .During 1998—2003,annual mean concentration of SO₂, NO₂ and PM₁₀ presents a trend of fluctuating decrease. PM10 is chief pollutant in Xi'an city. The monthly mean concentrations of the three kinds of pollutants are at high lever in the spring and winter periods and at low lever in the summer and fall periods, and the same trend can also be observed in heating providing period and the period without heating. Spatial distribution is discovered that the maximum of annual mean concentration of SO₂ and NO₂ appears at traffic and commercial area, the maximum of annual mean concentration of PM₁₀ appears at industrial area.ii .The condition of wind and the characteristics of stratification of lower-layer atmospheric temperature are important to air pollution in Xi'an city. The correlation coefficient between temperature vertical decrease rate of low-layer atmosphere and air pollutant concentration is negative. When Xi'an city was controlled by a coldhigh-pressure system or at the backside of cold front, only light pollution might happen there. One the contrary, controlled by homogeneous pressure system or at the foreside of cold front, a moderate or heavy pollution easily happened.iii.Most of sand- dust weather mainly takes place in spring especially in April. They cloud make the average growth rate of PMio concentration to reach 12.1% during March and April. Mean hourly PMio concentration could increase by 0.585 mg-m"J,maximum PMio concentration reached 0.970mg-m'3 in a server sand-dust storm. Snowfall in winter can rain out more efficiently than rainfall in other seasons, the effect of lmm snowfall in winter cleaning out SO2, NO2 and PMio concentration is 4 times, 3 times and 3.78 times of that in summer with same precipitation.iv.Severe sand-dust weather can lead to very heavy air pollution in spring in Xi'an city, and it is obviously different from the characters that the heaviest pollution occurs in December of winter.v .PMio concentration in different seasons in Xi'an city was forecasted by the ways of linear regression prediction equation and BP artificial neural network model based on main principal components. The results indicated that the Back-propagation (BP) neural network model is much better than linear regression prediction equation in historical sample fittings and independent sample test, because PMio concentration prediction model of neural network, condensing forecast information, reducing dimension and eliminating predict noise, could advance the prediction accuracy.