OCO-2 위성자료와 딥러닝을 이용한 한반도 XCO2 모니터링과 예측 (Monitoring and Forecasting XCO2 Using OCO-2 Satellite Data and Deep Learning)

서지정보

· 발행기관 : 한국대기환경학회
· 수록지 정보 : 한국대기환경학회지 / 40권 / 5호 / 572 ~ 584페이지
· 저자명 : 이권호, 김관철

초록

Carbon dioxide (CO2) is a greenhouse gas that contributes to global scale climate change, but its spatial and temporal variations at the regional scale are not well understood. This study aims to investigate the long-term spatiotemporal variations of atmospheric CO2 concentrations over the Korean Peninsula, utilizing satellite-observed XCO2 data from September 2014 to December 2023. Analysis reveals that the average XCO2 concentration in the Seoul metropolitan area (416.34±2.03 ppm) is notably higher than in other regions, highlighting the uneven distribution of emissions. Furthermore, seasonal patterns and long-term trends indicate a consistent year-on-year increase in XCO2 levels. To enhance our understanding of future CO2 trends, we employed a CNN-LSTM deep learning model, which yielded promising accuracy metrics (MSE=3.10 ppm, RMSE=1.73 ppm, MAE=1.49 ppm, MAPE=0.36%). The model’s predictions suggest a continued rise in XCO2 concentrations, projecting a peak of 432 ppm in 2024~2025. Without substantial efforts to mitigate carbon emissions, further increases in CO2 levels are anticipated. This research underscores the urgent need for effective climate policies to address rising greenhouse gas concentrations.

영어초록

Carbon dioxide (CO2) is a greenhouse gas that contributes to global scale climate change, but its spatial and temporal variations at the regional scale are not well understood. This study aims to investigate the long-term spatiotemporal variations of atmospheric CO2 concentrations over the Korean Peninsula, utilizing satellite-observed XCO2 data from September 2014 to December 2023. Analysis reveals that the average XCO2 concentration in the Seoul metropolitan area (416.34±2.03 ppm) is notably higher than in other regions, highlighting the uneven distribution of emissions. Furthermore, seasonal patterns and long-term trends indicate a consistent year-on-year increase in XCO2 levels. To enhance our understanding of future CO2 trends, we employed a CNN-LSTM deep learning model, which yielded promising accuracy metrics (MSE=3.10 ppm, RMSE=1.73 ppm, MAE=1.49 ppm, MAPE=0.36%). The model’s predictions suggest a continued rise in XCO2 concentrations, projecting a peak of 432 ppm in 2024~2025. Without substantial efforts to mitigate carbon emissions, further increases in CO2 levels are anticipated. This research underscores the urgent need for effective climate policies to address rising greenhouse gas concentrations.

참고자료

· 없음