표제지

목차

Abstract 11

제1장 서론 14

1.1. 연구 배경 및 필요성 14

1.2. 연구의 목적 17

1.3. 연구의 범위 19

제2장 연구 방법 21

2.1. 기상 모델 21

2.2. 배출량 모델 23

2.2.1. 국내 배출량 산정 24

2.2.2. 자연 배출량 산정 34

2.2.3. 황사 배출량 산정 37

2.3. 화학수송 모델 43

2.3.1. CMAQ 43

2.3.2. CAMx 48

2.4. 앙상블 모델 51

2.5. 모델링 정합도 분석방법 63

2.5.1. 통계학적 분석 63

2.5.2. 예보적중률 분석 67

제3장 연구결과 70

3.1. 결정론적 모델의 구성 70

3.1.1. F-01 예보 모델 71

3.1.2. F-02 예보 모델 71

3.1.3. F-03 예보 모델 75

3.1.4. F-04 예보 모델 76

3.1.5. F-05 예보 모델 78

3.2. 앙상블 모델의 구성 79

3.2.1. AF-01 모델 81

3.2.2. AF-02 모델 81

3.2.3. AF-03 모델 82

3.2.4. AF-04 모델 83

3.2.4. AF-05 모델 84

3.3. 모델 예측 정확도 분석결과 86

제4장 결론 124

참고문헌 127

〈Table 1〉 YSU scheme upgrade note in WRF update 22

〈Table 2〉 Emission source classification in the CAPSS 25

〈Table 3〉 Emission informations provided in the CAPSS 27

〈Table 4〉 Emission input data necessary for chemical transport model 27

〈Table 5〉 Characteristics of BEIS 3.14 and MEGAN 2.04 36

〈Table 6〉 The composition of the soil texture in the dust source region 39

〈Table 7〉 Monthly variations of the threshold wind speed in each soil-type region 40

〈Table 8〉 The dust occurrence probability density function with the R² value, free NDVI value (FNV) and upper limit value of NDVI for dust... 42

〈Table 9〉 A list of chemical species in CMAQ v4.6 and v5.0 45

〈Table 10〉 CMAQ v5.0 release note 47

〈Table 11〉 A list of chemical species in CAMx 50

〈Table 12〉 Uncertinties of chemical transport modeling associated with main input variables 55

〈Table 13〉 Statistics to evaluate model performance 63

〈Table 14〉 Verification statistics used to evaluate category forecasts 69

〈Table 15〉 A list of prognostic forecast models proposed in this study 70

〈Table 16〉 Model version and options in F-01 and F-02 forecasting modeling 72

〈Table 17〉 Model version and options in F-03 forecasting modeling 75

〈Table 18〉 Model version and options in F-04 forecasting modeling 77

〈Table 19〉 Model version and options in F-05 forecasting modeling 78

〈Table 20〉 A list of ensemble methods proposed in this study 80

〈Table 21〉 Classified 4 categories based on previous day's wind direction for AF-03 ensemble forecast model 82

〈Table 22〉 Classified 5 categories based on previous day's average wind speed for AF-04 ensemble forecast model 83

〈Table 23〉 Classified 5 categories based on previous day's average humidity for AF-05 ensemble forecast model 84

〈Table 24〉 A list of air quality monitoring stations in Seoul 87

〈Table 25〉 A list of air quality monitoring stations in Incheon 88

〈Table 26〉 Statistical analysis of hourly PM₁0 forecasting in Seoul(이미지참조) 91

〈Table 27〉 Statistical analysis of daily PM₁0 forecasting in Seoul(이미지참조) 91

〈Table 28〉 Statistical analysis of hourly PM₁0 forecasting in Incheon(이미지참조) 92

〈Table 29〉 Statistical analysis of daily PM₁0 forecasting in Incheon(이미지참조) 92

〈Table 30〉 Statistical analysis of hourly PM₁0 forecasting in Seoul(이미지참조) 107

〈Table 31〉 Statistical analysis of daily PM₁0 forecasting in Seoul(이미지참조) 107

〈Table 32〉 Statistical analysis of hourly PM₁0 forecasting in Incheon(이미지참조) 108

〈Table 33〉 Statistical analysis of daily PM₁0 forecasting in Incheon(이미지참조) 108

〈Table 34〉 Hit rate analysis of hourly PM₁0 forecasting in Seoul(이미지참조) 109

〈Table 35〉 Hit rate analysis of daily PM₁0 forecasting in Seoul(이미지참조) 110

〈Table 36〉 Hit rate analysis of hourly PM₁0 forecasting in Incheon(이미지참조) 110

〈Table 37〉 Hit rate analysis of daily PM₁0 forecasting in Incheon(이미지참조) 111

 The East Asia region is the world's most populous area with a rapidly growing economy resulting in large air pollutant emissions. The urban and industrial development of China in this region leads to substantial increase of anthropogenic emission and it causes heavy and complex air pollution problems in China and neighboring countries. The air quality in the Seoul Metropolitan Area (SMA), Korea has deteriorated due to emissions in the SMA itself as well as the growing influence from China. The large anthropogenic emission from megacity clusters in North and East China such as the Beijing-Tianjin-Hebei province, the Yangtze River delta and Shenyang located in eastern China can be transported to the SMA by the prevailing northwestern winds in Winter and Spring season. Moreover, dust emission from deserts in northwestern China can be also moved to the Korean peninsula causing the dust event from Winter to early Summer.

It's thus necessary to promote the method of protecting the people's living environment by publicizing the possibility of a pollutant event with high concentration of fine PM (Particulate Matter) in advance before air quality gets worse in order to minimize its consequential health damage. The biggest issue in the advance alarming of air quality is how to secure the accuracy in the air quality forecasting; actually, a lot of research work is under way in an effort to improve the accuracy in the air quality forecasting at home and abroad.

This study was carried out to improve the accuracy of forecasting model using various prognostic and assemble approaches based on the previous research works. The 5 prognostic forecast models were proposed using various combinations of WRF updated versions as a meteorological model, emission processing methods using SMOKE, and different chemical transport models of CMAQ and CAMx. The 5 ensemble forecast models were also proposed based on the following configurations: ① a method of simply doing an arithmetical mean of 5 prognostic model outcomes, ② a super-ensemble method in which the weighted values of the prognostic models are applied, ③ a method of the expanded super-ensemble in which the weighted values of the prognostic models are calculated by group classified by the main wind direction of the previous day, ④ a method of the expanded super-ensemble in which the weighted values of the prognostic models are calculated by group classified by the wind speed of the previous day, ⑤ a method of an expanded super-ensemble in which the weighted values of the prognostic models are calculated by group classified by the humidity of the previous day.

The forecasted results from the proposed 5 prognostic models (F01 to F05) and 5 ensemble models (AF01 to AF05) were compared with air quality and meteorological observations in the SMA. The period of forecast performance test was from January to March in 2013.

The results of comparison in daily PM10 concentration in the Seoul metropolis shows that AF-04 to which the weighted values were applied based on wind speed has the highest accuracy rate with 76.4% while the other ensemble models are 66.2 - 75.2% accuracy rate. In the false warning rate, the ensemble models are from 30.7% to 47.0%, showing an exceptionally improvement comparing to a single prognostic model whose false warning rate stayed between 47.6% and 56.6%.

The comparison shows that the ensemble models with weighted values on meteorological conditions, such as wind direction, wind speed and humidity improve the accuracy in the forecasting. This study also prove that the reliability in the air quality forecasting can be improved employing the ensemble approach rather than using a single prognostic model and the ensemble methods in determining weighted values are important.