Characterization of Traffic-Related Air Pollution along Highways Using Ground and UAV Observations
지상 및 드론 관측 기반 고속도로 교통 기인 대기오염 특성 분석
- 주제(키워드) Highway , Roadside , Air Pollution , Unmanned Aerial Vehicle (UAV) , Machine Learning
- 주제(DDC) 628
- 발행기관 아주대학교 일반대학원
- 지도교수 이재영
- 발행년도 2026
- 학위수여년월 2026. 2
- 학위명 석사
- 학과 및 전공 일반대학원 환경공학과
- 실제URI http://www.dcollection.net/handler/ajou/000000035860
- 본문언어 영어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
Traffic-related air pollution along highways exhibits strong spatiotemporal variability driven by traffic volume, travel speed, and site-specific spatial characteristics. To address the limitations of fixed ground monitoring in capturing these dynamics, this study applies an integrated framework combining ground-based measurements, unmanned aerial vehicle (UAV) profiling, traffic data, and machine learning analysis. In Chapter 1, short-term and long-term measurements were conducted sequentially to identify representative highway sections and monitoring sites. Short-term measurements of CO, NO2, O3, and PM2.5 were performed across four functional highway environments—parking areas, service areas, roadside sections, and tollgates—along major expressways. When short-term observations were evaluated together with emission estimates and traffic characteristics, the Gyeongbu Expressway showed the most pronounced emission signature, with estimated NOₓ emissions up to 25–30% higher than other routes, and measured pollutant–traffic correlations that were consistently stronger; for example, Route A (Gyeongbu) exhibited correlation coefficients of up to 0.4–0.5 for CO and PM2.5, which were noticeably higher than those observed along the other routes. Consequently, the Gyeongbu Expressway was selected for long-term intensive monitoring. Long-term analysis revealed that NO2 concentration was 397.7–888.7% higher than nearby Air Quality Monitoring Stations (AQMS), indicating strong primary emission influence, whereas O3 and PM2.5 showed smaller differences (typically 10–40%), reflecting secondary formation and regional background contributions. Machine learning analysis further confirmed that roadside locations most directly capture traffic-induced air quality variations, and UAV measurements showed spatial concentration gradients consistent with near-road emission effects. In Chapter 2, the spatial distribution of roadside air pollutants was identified using UAV measurements and their concentrations were predicted using machine learning models. Pollutant concentrations generally decreased with increasing altitude and distance from the roadway, with NO2 and O3 exhibiting similar spatial trends. NO2 concentrations showed a pronounced peak at 7 a.m., corresponding to periods of heavy traffic emissions. Seasonally, O3 concentrations were higher in summer due to enhanced solar radiation, while PM2.5 levels also increased during summer, with the influence of long-range transport confirmed through Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) trajectory analysis. The UAV-to-ground concentration ratios for each pollutant were predicted using machine learning models, among which the Categorical Boosting (CatBoost) model was selected as the final model based on its superior performance (R2 = 0.65–0.95). SHapley Additive exPlanation (SHAP)-based feature importance analysis revealed a negative relationship between traffic speed and CO concentrations. The predicted spatial concentration distributions of roadside pollutants were consistent with known general concentration patterns for each pollutant. Overall, the proposed UAV-integrated machine learning framework provides a novel and effective approach for high-resolution, three-dimensional estimation of roadside air quality. Keywords: Highway, Roadside, Air Pollution, Unmanned Aerial Vehicle (UAV), Machine Learning
more목차
Chapter 1. Spatiotemporal Analysis of Traffic-Related Air Pollution along a Highway 1
1. Introduction 2
2. Methodology 4
2.1 Study Design 4
2.2 Measurement Setup 10
2.3 Data Collection 11
3. Results and Discussion 12
3.1 Emission-Based Comparison of Major Highway Characteristics 12
3.2 Integrated Analysis of Roadside Air Pollution Characteristics 21
3.2.1 Long-Term Concentration Characteristics of Highway Locations 21
3.2.2 Traffic-Related Diurnal Variation of Roadside Concentrations 27
3.2.3 Machine Learning-Based Prediction of Roadside Concentrations 30
3.2.4 Spatial Concentration Distribution of Roadside Based on UAV Measurements 36
4. Conclusions 39
Chapter 2. Altitude-Resolved Prediction of Roadside Air Pollution Using UAV Measurements 42
1. Introduction 43
2. Methodology 46
2.1 Measurement Site 46
2.2 Instrumentation 46
2.2.1 Sensor 46
2.2.2 UAV Platform 47
2.3 Experimental Setup and Study Design 49
2.4 Data Analysis 51
2.4.1 Data Collection 51
2.4.2 UAV/Ground Prediction Modeling 52
3. Results and Discussion 55
3.1 Spatial Concentration Distribution Characteristics 55
3.1.1 Spatial Concentration Profile 55
3.1.2 HYSPLIT-Based Long-Range Transport Trajectory Analysis 58
3.2 Prediction using Machine Learning Methods 61
3.3 Spatiotemporal Characteristics of Predicted UAV/Ground 68
4. Conclusions 73
Chapter 3. Conclusions and Future Works 75
1. Conclusions 76
2. Limitations and Future Works 78
References 80
Appendix 89
국문 초록 99
