Study On Meteorological Conditions, Particulate Matter And PM2.5 Over PM10 Ratio

Abstract

BACKGROUND Some studies have found that in time-series analysis, the effect size of particulate matter (PM) on mortality may change over time and season. One reason could be a change of toxicity of PM. To verify temporality of its toxicity, we used PM2.5 fraction in PM10, as a surrogate of its toxicity. We analyzed temporality of the fraction and association with meteorological conditions which are main confounders in time-series analysis. METHODS Particulate matter and meteorological conditions were measured on Jongro district by National Institute of Environmental Research and Korea Meteorological Administration, respectively. Those two stations are 3.3Km away. Meteorological factors were rain, wind-speed, temperature, relative humidity and diurnal temperature range (DTR). Using a generalized additive model, association between PM and meteorological conditions were adjusted. Time-trend for unmeasured confounders were adjusted. To identify relevance of single monitoring station for meteorology in time-series analysis, we did the same analysis for PM data from a station in Gangdong district which are 15.4 km away from the station in Jongro district. RESULTS PM10 and PM2.5 has decreased over time. But the fraction of PM2.5 has increased consistently. As the level of rain and wind-speed increased, both PM10 and PM2.5 mass decreased. Temperature and wind-speed showed negative association with them but daily temperature range (DTR) showed positive. For the fraction of PM2.5 in PM10 we found that temperature was positively associated while rain and DTR were not significant. Similar results were found for the farthest PM station. DISCUSSION This study showed toxicity of PM10 may increase over time. Also, its toxicity may depend on meteorological conditions. This implies that the effect size of PM estimated from time-series studies can vary. In addition, our result implies that a use of single meteorological station would be valid in time-series analysis.