Environmental Time Series Analysis and Estimation with Extended Kalman Filtering

This paper addresses the problem of air pollutant profile estimation by using measurements collected from different weather stations. An algorithm is developed, based on an Extended Kalman Filter to handle missing temporal data and using the statistical Kriging method to interpolate spatial data. Combination of extended Kalman filtering with Matérn covariance function has proven to be useful in exploiting meteorological information to build reliable air quality models. We have applied the developed algorithm to estimate air pollutant profiles in the Sydney basin, which is subject to a variety of pollutant sources, including fossil-fueled electric power generation plants, high motor vehicle usage, aviation and shipping traffic. The results have shown that the proposed approach can improve accuracy of the estimation profiles.