Spatiotemporal Relations between PM<SUB>10</SUB> Concentrations and Relative Humidity in South Korea = 우리나라 PM<SUB>10</SUB> 농도와 상대습도 간의 시공간적 상관관계 분석

국문 초록 (Abstract)

우리나라의 PM10 농도는 뚜렷한 계절적 패턴을 보이고 있고, 이러한 시계열적 농도 분포는 일부 해안 지역을 제외하고 해당 지역의 습도 조건과 강한 음의 상관관계를 보였다. 주별 및 일별 자료를 이용한 상호상관분석 결과, PM10 농도는 시계열적으로 상대습도 조건에 대해 평균 1주간의 지연효과(lag effect)를 보이는 것으로 나타났다. 따라서 일정 시점의 PM10 농도는 직전 주의 습도 조건에 영향을 받고 있을 가능성이 있다. 계절효과 표준화(deseasonalization)를 통해 추출된 PM10 농도의 장기 변화 경향을 선형 회귀모형을 통해 산출하여 PM10 농도 예측을 수행하였다. PM10 농도 예측 상의 표준오차(RMSE)는 여름을 제외한 계절적 변량과 유의한 상관성을 나타냈다. 전반적으로 연구 기간(2007~2016) 중 분석된 PM10 농도 예측 오차는 PM10 농도가 전반적으로 높은 지역일수록 낮고, PM10 농도 감소 속도가 빠른 지역일수록 높게 나타났다.

다국어 초록 (Multilingual Abstract)

Time-series analyses of nationwide PM10 concentrations and relative humidity showed that they had strong seasonality and were correlated with each other in South Korea. A significant negative correlation was found between them in most locations, but their relationship was not significant at a few coastal locations. Weekly and daily analyses indicated that their cross-correlation was peaked when PM10 concentrations were 1-week lagged behind relative humidity on average. Therefore, it is likely that PM10 levels at a given time had been influenced by the moisture conditions of the previous week. Long-term(2007~2016) trend of PM10 concentrations were extracted by removing seasonal effects within the original time-series using a deseasonalization model, and it was fitted into a linear regression line for prediction of PM10 concentrations. The prediction errors of the deseasonalization model were measured as the root-mean-square error (RMSE), and they were significantly correlated with seasonal relative humidity variance except summer. Geographically, prediction accuracy of PM10 concentrations was higher in locations, where the overall PM10 levels were lower and the decreasing trends of PM10 concentrations were higher during the study period.

목차 (Table of Contents)

• Abstract
• 요약
• 1. Introduction
• 2. Materials and Methodology
• 3. Results and Discussion
• 4. Concluding Remarks
• References