The Meteorological Causes Of Ozone Pollution In Baotou And Its Prediction

Surface O₃ pollution has become the most important atmospheric environmental problem in summer in most regions in China,and has adverse effects on human health,crop growth,and climate change.With the implementation of strict control measures for PM_2.5,the O₃ pollution in Baotou may further aggravate.The topographic features of high north and low south,the special terrain that the Yinshan Mountains traverse,and the geographical location of the junction between the monsoon area and the non-monsoon area make the O₃ pollution of Baotou more susceptible to meteorological conditions.Clarifying the relationship between ozone concentration and meteorological conditions such as meteorological factors and synoptic situation,and establishing the daily concentration forecast equation based on the relationship is vital for the early waning of ozone pollution,and that will also provide a scientific basis and guidance for the fine management of atmospheric environment and decision-making.This paper firstly used correlation analysis to discuss the relationship between O₃ concentration and meteorological elements in Baotou,and used the Hysplit model to analyze the transport characteristics and potential source of O₃qualitatively.Secondly,the Lamb-Jenkinson method was applied to classify the weather conditions and the influence mechanism of different synoptic patterns on O₃ in Baotou was explored.Finally,six kinds of machine learning models were established by using meteorological data to forecast the maximum daily eight-hour average concentration of O₃(MDA8 O₃)in Baotou.The main conclusions are as follows:(1)From 2014 to 2020,O₃ concentrations in Baotou showed an overall increasing trend,with the 90th percentile,mean and 5th percentile of MDA8 O₃ increasing by 9,13 and 4?g/m~3,respectively.The monthly change of O₃ concentrations showed an inverted"V"type,with the days exceeded the national standard mainly concentrated in May-August.O₃ concentration was significantly positively correlated with daily maximum air temperature(r=0.62,p<0.01),negatively correlated with air pressure(r=-0.18,p<0.01),and had no significant correlation with relative humidity,wind speed and precipitation.However,according to the O₃ concentration and the exceeding standard rate corresponding to different intervals of each meteorological element,it can be seen that the relative humidity and wind speed are too high or too low,which are not conducive to the occurrence of O₃ pollution.The vast majority(92%)of O₃ pollution occurred on days without precipitation.When the wind direction is southwest-south-southeast,O₃ pollution is prone to occur.(2)There is a significant difference between the air mass trajectories on the O₃clean days and those on the exceeding standard day.72.9%of the air mass trajectories on the clean day come from the north,with a relatively long transport distance,fast moving speed and clean underlying surface,while 88.5%of the air mass trajectories on the polluted days come from the south,whose transport distance is short,the moving speed is slow,and the underlying surfaces passing through are mostly heavy industrial cities.The areas with high WPSCF and WCWT values on over-standard days and clean days are mainly in the south of Baotou,and the main potential source areas include Ordos,Yulin,Luliang,Yan'an,Linfen,Tongchuan,Yuncheng,Jincheng and other places.(3)Synoptic type over Baotou was identified by L-J method from May to August,we found that S,SW,and C types accounted for 50.2%of all synoptic types,and the corresponding O₃ concentration mean and over-standard rate were high,151?140?g/m~3 and 38.8%?24.8%respectively,which were prone to O₃ pollution and were called O₃ pollution-prone synoptic types.The N,NE and NW types have a low frequency of occurrence,with an accumulation of 15.2%,and the corresponding O₃concentration is the lowest,ranging from 102 to 112?g/m~3.These three types have no ozone over-standard days and are called as clean synoptic types.(4)Under the N,NE and NW synoptic patterns,Baotou is controlled by the low pressure trough at 500h Pa,affected by the dry and cold northwesterly air flow at middle layer significantly,mostly located in front of the cold high pressure at surface,where the local temperature is low,there is a strong northerly wind,the diffusion condition is good with low return index value,long transmission distance and high boundary layer height,which is not prone ozone pollution.The air masses affecting Baotou mainly pass through Bayannaoer and Outer Mongolia in the N and NE types,while the air masses in the NW type mainly come from Alxa and Bayannaoer.Under the S,SW and C synoptic patterns,there is a high pressure shallow ridge at 500h Pa over Baotou,and Baotou is affected by the warm center or warm advection at 850h Pa.The recirculation index,transport distance and boundary layer height are all at the medium level,and the diffusion conditions are good.These types are not only favorable for local ozone generation,but also favors the accumulation of O₃ and its precursors,which were transported to Baotou with the southerly air flow,in the front of the Yinshan mountain,which makes the O₃ concentration increasing.In the C type,the air mass affecting Baotou mainly passes through the surrounding Ordos,Wuhai and Shizuishan.Compared with the C type,the SW type also passes through Yulin,Yan'an,Qingyang,Xianyang,Tongchuan and Weinan,while S also passed through Xinzhou,Luliang,Linfen and Yuncheng in Shanxi.(5)Based on the meteorological data and the previous day's O₃concentration,six machine learning models were established to simulate and retrospectively forecast the MDA8 O₃ in Baotou.Among the six models,the extra tree regression model(ETR)has the best performance in all stages with good fitting effect on O₃ concentration and small deviation between forecast and true values.The R~2,RMSE,MAE,and MAPE of the ETR model from May to August are 0.68,17.6,13.8,and 11.3%,respectively.Its forecast accuracy for O₃ exceeding the standard day is relatively good,with a forecast score(TS)of 0.52,a false rate of 8%,and a missed rate of 45%.The R~2,RMSE,MAE,and MAPE of the Light GBM model(LGBR)are 0.64,18.6,13.9,and 11.5%,respectively,which are slightly worse than that of the ETR model,but the forecast accuracy for O₃ exceeding the standard day is the best.Its TS score is highest at 0.62,false alarm rate and missed rate are lowest at 7%,35%respectively.The linear regression model(LR)has the worst overall performance,with R~2,RMSE,MAE,and MAPE of 0.53,21.2,16.8,and 13.9%,respectively.Its TS score is lowest at 0.30,while its false rate and missed rate are both the highest,at 30%and 65%,respectively.In general,the machine learning forecast model based on meteorological parameters can make good forecasts of Baotou O₃ concentrations and exceedance days.