고해상도 시공간 데이터를 이용한 초미세먼지(PM2.5) 인구 노출의 지리정보 통합평가 = Geo-integrated assessment of population exposure to PM2.5 using high-resolution spatiotemporal data

다국어 초록 (Multilingual Abstract)

With the rapid pace of urbanization and industrialization, air pollution has emerged as one of the most serious environmental and public health challenges in modern society. In particular, fine particulate matter (PM2.5) is a major pollutant closely associated with cardiovascular and respiratory diseases, as well as premature mortality. Because these particles have diameters of 2.5 μm or smaller, they can penetrate deep into the alveoli. To accurately assess their health effects, it is essential to measure and analyze regional spatial exposure levels of PM2.5 precisely. However, PM2.5 exposure is a high dimensional problem characterized by complex interactions among multiple environmental factors and spatial structural features. In the Gwangju–Jeonnam region, where the monitoring network is relatively sparse, it is difficult to adequately capture micro-spatial variability and interregional exposure disparities using conventional measurement station based on approaches alone. Therefore, this study aimed to develop a machine learning–based PM2.5 exposure prediction model by integrating diverse environmental and spatial datasets, including population information, and to identify exposure patterns and vulnerabilities among different population groups according to regional and land-use characteristics. In this study, air quality, meteorological, 1 km² population grid, and land-use data from Gwangju Metropolitan City and Jeonnam Province were integrated to produce high-resolution PM2.5 exposure estimates. Spatial inequalities were analyzed using both the prediction model and population-weighted average concentration (PWAC), thereby addressing the limitations of previous region-based studies. The PWAC based analysis revealed that high-concentration exposure events occurred most frequently in winter, least frequently in summer, and peaked during spring and exceeding the annual average standard. Although most daily averages were below the standard, annual mean exposures exceeded it across all regions. Areas displaying both high model predicted PM2.5 concentrations and high population density were clearly observed in downtown Gwangju, commercial districts, and densely populated residential zones. Correlation analysis by land-use type showed positive correlations between PWAC and urban, residential, commercial, and road areas, while forests, grasslands, and waterside regions demonstrated negative correlations. Moreover, a strong similarity was observed between urban form and air pollution cluster structures, confirming that land use and spatial configuration play key roles in PM2.5 exposure. Cluster analysis identified three major groups: urban intensive, mixed agricultural, and natural green space clusters. This study is significant in that it integrates a high resolution PM2.5 prediction model with PWAC to analyze urban land use and population exposure characteristics in the Gwangju and Jeonnam region. However, in certain parts of Gwangju Metropolitan City, model accuracy was limited—the coefficient of determination fell below 0.9 due to wide spacing between monitoring stations. Moreover, emission-related data such as traffic volume and heating fuel usage were not directly incorporated, leading to potential underestimation in high-concentration scenarios. Additionally, because the PWAC analysis relied on static population data, it could not capture real-world movement patterns within residential zones. Land-use and land-cover data were also taken at a single time point (2024), limiting the model’s ability to reflect seasonal variations. Therefore, future research should improve exposure assessment methods by incorporating dynamic population mobility and expanding the analysis to include indoor environments with extended dwelling times, as well as emission source data.

목차 (Table of Contents)

• (Abstract)
• 제 1 장 서 론1
• 1.1 연구배경 및 필요성 1
• 1.2 연구목적 4
• 제 2 장 연구방법5
• 2.1 연구 설계 5
• 2.2 데이터 수집 6
• 2.3 추정 모델 개발 및 검증7
• 2.4 인구 노출 평가 9
• 2.5 지리정보 통합 노출평가 10
• 제 3 장 연구결과 12
• 3.1 데이터 수집 12
• 3.2 추정 모델의 검증 25
• 3.3 인구 노출 평가 27
• 3.4 인구 노출의 지리정보 민감도 분석 31
• 제 4 장 고 찰 35
• 제 5 장 결 론 39
• [참고 문헌] 41
• [그림 차례]
• [그림 1] 연구 개요 5
• [그림 2] 광주광역시 및 전라남도의 대기질측정소 및 기상대 위치 현황도
• ·13
• [그림 3] 월별 초미세먼지(PM2.5) 농도 현황 17
• [그림 4] 계절별 초미세먼지(PM2.5) 농도 현황 17
• [그림 5] 광주광역시 및 전라남도의 2024년 계절별 풍배도 21
• [그림 6] 격자별로 추출된 2024년 토지이용피복지도 24
• [그림 7] 관측된 PM2.5 농도와 예측 농도의 상관성 26
• [그림 8] PM2.5 예측 모델의 잔차분포 26
• [그림 9] 광주광역시 2024년 인구분포28
• [그림 10] 계절별 PM25 농도와 인구가중 분포(PWAC)의 비교30
• [그림 11] 토지이용 유형 및 격자별 PWAC 상관성 분석 32
• [그림 12] 2024년 일별 PWAC와 토지이용 면적 및 격자별 군집화 결과 ··
• ·34