Analysis of spatial and temporal variations of PM 10 concentrations in the Netherlands using Kalman filtering

Abstract The spatial and temporal variations of PM10 concentrations in the Netherlands as measured by the National Air Quality Monitoring Network in the period of 1993–1994 have been analysed using descriptive statistics, principal component analysis (PCA) and Kalman filtering. Spatial differences in PM10 concentrations in the Netherlands are rather small. PM10 concentrations may be elevated by about 10–20% with respect to the yearly average, which is about 40 μg/m3, in areas with local sources such as traffic or other urban, industrial or agricultural sources. Actual PM10 concentrations vary between 20 and 50 μg/m3 throughout the year. During episodes, PM10 concentrations may increase to 4 to 5 times the annual average (>200 μg/m3). The large amount of variance explained by the first component of PCA, i.e. 85%, shows that all measuring stations observe the same pattern of daily variations which is mainly governed by large-scale weather systems.The daily variations are analysed using multiple-linear regression and Kalman filtering; the latter employed as a time-varying linear regression technique. The results of the both methods are compared and show that using wind direction, temperature and duration of precipitation as variables, ordinary linear regression explains about 25% of the variance of PM10 concentrations, while the application of the Kalman filter explains about 45% of the variance. The improvement using the Kalman filter is primarily obtained by making the explaining variables time dependent. This shows a significant effect of seasonal variation on temperature and wind direction at PM10 levels.