미세먼지예보 정확도 향상에 관한 연구

Abstract

Along with the industrial development, environmental pollution emerged as one of the most critical social problems that needs be resolved urgently. In particular, air pollution, one of the major factors that affects the human life most significantly, possess high uncertainties and its estimates vary greatly depending on implementation results of reduction measures or weather changes. Air pollution is reported to have negative impacts on human life. Especially the weak and the elderly are mainly affected by air pollution. Consequently it is understood that a common ground was created to realize that air pollution is a problem that must be resolved in order to provide a healthy and pleasant life for the humankind. Since 2005, the city of Seoul has implemented forecasting/alarming system to protect citizens' health from PM10. Its accuracy rate hovers around 72%, and some people raised the necessity to improve the confidence in the system. The study has analyzed impacts and intensity of weather that affect PM10 concentration based on PM10 forecast conducted by the city of Seoul in order to identify ways to improve the accuracy of PM10 forecast. Variables that influence PM10 concentration include not only velocity and direction of the wind and rainfalls, but also those including secondary particulate matter, which were identified to greatly influence the concentration in complicated manner as well. In addition, same variables were found to have different impacts depending on seasons and conditions of other variables. This study assumed that forecast methods should be specified to enhance the accuracy of PM10 forecast and suggested ways to improve the accuracy based on estimation of increase and decreases of seasonal variables considering characteristics of influencing variables. Factors that influence PM10 concentration such as rainfalls, velocity and direction of wind, and relative humidity tend to be inconsistent and have irregular impacts. Hence the study focused on wind velocity as permanent influence factor to analyze influence changes of different variables within the scope of each wind velocity. However it was difficult to measure consistent and quantitative influences of specific variables as this varied as well. Additionally as co-variance among variables were recognized to be affected significantly, it was inevitable to avoid uncertainty. To improve accuracy of PM10 forecast, this study has established a method that increase or decrease the estimation by anticipating the possible fluctuation scope of PM10 concentration if one of the variables of weather factor fluctuates, through dividing the period into seasons and estimating sensitivity of variables that influence PM10 concentration based on the wind velocity sectors. The study also predicted that rainfall forecast will be more accurate if the impact is applied by dividing into time units. The study also pointed out that temperature and relative humidity have different impacts on PM10 concentration at day and night time, and the impacts of wind direction differ in winter and summer. In other words in the daytime, temperature rises and relative humidity falls while in the night the situation becomes the other way around. Thus PM10 concentration forecast needs to be estimated in different ways by integrating hourly PM10 concentration estimates and finalizing the forecast. The estimates that change PM10 concentration based on the wind velocity changes were anticipated by wind direction. The study found out that improving accuracy of PM10 concentration forecast face some limits as it is greatly influenced by the weather. As an estimation, this study assumed that basic research units and artificially estimated pollutant emissions, study on mechanisms of secondary particulate matter productions, observatory compliment, and enhanced forecaster's expertise are needed for better forecast. This study is not to correct errors emerged from forecasting programs during the analysis process. Rather this study can be regarded as a pre-step for the forecaster to apply analysis results to compliment forecast program gradually if the results can be combined practically. Unlike other countries, forecast history in Korea is short and some air pollution reduction measures implemented in the country influence the changing scope of pollution. Furthermore around 2/3 of the year, Korea is influenced by China's situation, making it more difficult to forecast. This study was conducted to analyze and help forecast errors from the current stage and it is my hope that this study can help those who are willing to improve forecast accuracy in the future.