Spatial Simulation And Exposure Risk Assessment Of PM_2.5 Concentration In The Yangtze River Delta Based On Random Forest Model

Land use regression?LUR?model is the commonly used method to simulate the distribution of PM_2.5,but this method takes the multivariate linear regression modeling,not taking into account the complex relations between variables and PM_2.5concentrations of nonlinear,and easy to appear multicollinearity.In order to improve the accuracy of the simulation,this paper uses the random forest algorithm training model,the Yangtze river delta to simulate the spatial distribution of PM_2.5.And Referring to the existing PM_2.5 exposure risk assessment methods and the previously obtained PM_2.5 concentration spatial distribution data,we assess the exposure risk of PM_2.5 in the Yangtze river delta.Research shows that:?1?In this paper,we use the method of random forest improved regression model of land use model.To evaluate the simulation results,we adopt the correlation coefficient?R?,Root Mean square Error?RMSE?,Mean Absolute Error?MAE?and Index of Agreement?IA?.The larger R and IA,the better effect of the model;the smaller RMSE and MAE,the better effect of the model.Testing the effect of this model,the IA,MAE,RMSE and R of this model are 0.854,4.757,5.871,0.831 by the test sets.To illustrate the efficiency and accuracy of the model,this paper compared the model we adopted with multivariate regression of traditional land use regression model and SVM improved land use regression model.Using multivariate regression of traditional land use regression model in common use,IA,MAE,RMSE and R of test sets are 0.702,5.862,7.58,0.647,respectively.Using the method of SVM improved land use regression model test set of the model calculation of IA,MAE,RMSE and R are 0.825,5.521,6.871,0.825.By contrast,due to the larger test set IA and R,and the smaller MAE and RMSE,the random forest algorithm in PM_2.5concentration space simulation effect is better.?2?Average 2015 PM_2.5 concentrations of the Yangtze river delta shows the trend that lower from north to south,east to west;Cities in Jiangsu Province of PM_2.5concentration significantly higher than in Zhejiang Province and Shanghai,especially in Taizhou,Wuxi,Yangzhou,Changzhou and other Southern Jiangsu,central Jiangsu;The highest concentrations of PM_2.5 cities in Zhejiang Province are Huzhou,Shaoxing,Jiaxing,the best is in zhoushan and Taizhou;Hangzhou and Ningbo while the city's average value is not high,but also exist obvious high value area.?3?Based on the concentration of PM_2.5 air quality and PM_2.5 exposure risk population distribution of the population,the population PM_2.5 exposure risk of Yangtze river delta region from south to north,from east to west to present the phenomenon of increasing gradient.from.The downtown area is high value area population PM_2.5 exposure risks.Taking Hangzhou as an example,the population of PM_2.5 exposure risk is roughly with concentric circles structure.Center to the edge,the population of PM_2.5 exposure risk is weaker.weaker risk is in southwest area.Present continuous high risk area,downtown in the city center of vice city with patches of high risk area.Inside the city has some axis of relatively high risk area,such as belt along the main traffic routes and Qiantang River.While based on weighted average PM_2.5 exposure population risk,we can evaluate the effects of air pollution on public health between the units.Taizhou,Wuxi and Changzhou as the biggest risk value,should be given more to prevention and control,in addition,Hangzhou,Ningbo,Shaoxing,Shanghai and other places of the population is highly concentrated in PM_2.5 concentrations high value area,also need to focus on.?4?Concentration of PM_2.5 air quality,utilizing the exposure intensity,the population of the Yangtze river delta region exposure risk can be a preliminary analysis of spatial pattern;By the weighted average of PM_2.5 population exposed to the risk assessment for each administrative unit of PM_2.5,comparing the population exposure risk screening of PM_2.5 population exposure risk focus on prevention and control;Using the population exposure intensity can be further monitoring key prevention and control the focus of the urban area,suitable for urban area within the exposure risk assessment.Comprehensive above index for atmospheric pollution control and the improvement of living environment has important practical significance.In general,this paper has the following research features:the simulation accuracy is higher by combining AOD data with common land use regression data;In the traditional LUR model,the random forest algorithm is incorporated into the method,and it has some innovation in the method.It is more scientific to analyze the risk of PM_2.5 population exposure by combining the existing multiple evaluation methods.