학술논문
국내 미세먼지 오염도에 영향을 미치는 요인에 대한 분석
Analysis of Factors Influencing PM10 Pollution in Korea
Analysis of Factors Influencing PM
Document Type
Article
Author
Source
한국환경경제학회 하계학술대회논문집. Aug 27, 2018 2018:779
Subject
Language
Korean
Abstract
본 연구는 미세먼지(PM10)의 발생 및 확산과 관련된 요인들에 의사결정나무 분석을 적용하여 미세먼지 오염도에 미치는 영향력을 정량적으로 평가한다. 설명변수로는 대기오염물질 농도 및 배출량, 기상기후, 사회지리, 중국 대기질과 관련한 변수를 사용하며, 반응변수는 국내 시군구별 월별 미세먼지 평균농도이다. 각 모델의 설명변수를 다르게 하여 예측 정확도를 기준으로 변수의 영향력을 분석하였으며, 평균제곱오차를 감소시키는 효과를 기준으로 개별 변수의 중요도를 측정하였다. 마지막으로 민감도 분석을 통해 개별 변수의 증감이 미세먼지 오염도에 미치는 영향을 정량적으로 평가하였다. 분석 결과 황사 현상 및 중국내 주요 도시의 대기질과 관련한 변수가 미세먼지 예측 모델의 정확도를 높이고 중요한 변수이며, 국내 미세먼지 오염도에 2차 미세먼지가 미치는 영향력이 높음을 파악하였다.
This study quantitatively evaluates the influence of factors related to generation and diffusion of PM10 on PM10 pollution by applying decision tree analysis. We use factors related to air pollutant concentration and emission, weather, social geography and China air quality as the explanatory variables in tree-based models. The response variable is monthly average PM10 concentrations in each city and district. The influence of variables on the prediction accuracy of the models was analyzed by explaining the explanatory variables of each model, and the importance of individual variables was measured based on the effect of reducing the mean squared error. Finally, the sensitivity analysis was used to quantitatively evaluate the effect of the increase and decrease of individual variables on the PM10 pollution. As a result of this study, it was found that the variables related to the yellow dust phenomena and the air quality of the main cities in China are important for improving the accuracy of the PM10 prediction models and that secondary generated PM10 has a high influence on PM10 pollution in Korea.
This study quantitatively evaluates the influence of factors related to generation and diffusion of PM10 on PM10 pollution by applying decision tree analysis. We use factors related to air pollutant concentration and emission, weather, social geography and China air quality as the explanatory variables in tree-based models. The response variable is monthly average PM10 concentrations in each city and district. The influence of variables on the prediction accuracy of the models was analyzed by explaining the explanatory variables of each model, and the importance of individual variables was measured based on the effect of reducing the mean squared error. Finally, the sensitivity analysis was used to quantitatively evaluate the effect of the increase and decrease of individual variables on the PM10 pollution. As a result of this study, it was found that the variables related to the yellow dust phenomena and the air quality of the main cities in China are important for improving the accuracy of the PM10 prediction models and that secondary generated PM10 has a high influence on PM10 pollution in Korea.