Kalman filtering method and its application to air pollution episode forecasting. [Kalmanifilters are a class of linear minimum-error-variance sequential state estimation algorithms]
The Kalman filtering method is applied to multi-station air pollution modeling in order to obtain a useful real-time predictor of concentration levels, especially during episode situations. Special attention has been paid to avoiding certain high dimensionality problems of the Kalman filter while still retaining some of the deterministic physical information of the transport and diffusion phenomena. Moreover, a method is proposed to forecast future state values using only a probabilistic knowledge of future state-transition matrices, which is the most common situation in air pollution real-time forecasting with probabilistic meteorological input. Specifically, the method is applied to SO/sub 2/ and meteorological data (Summer 1975) supplied by the RAMS network (Regional Air Pollution Study) installed in the St. Louis Missouri area. The results of the proposed methodology are compared with those supplied by single-station predictors.
- Research Organization:
- Stanford Univ., CA (USA). Dept. of Statistics; IBM Scientific Center, Palo Alto, CA (USA)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- EY-76-S-02-2874
- OSTI ID:
- 6297010
- Report Number(s):
- COO-2874-48
- Country of Publication:
- United States
- Language:
- English
Similar Records
Relationships involving fine-particle mass, fine-particle sulfur, and ozone during episodic periods at sites in and around St. Louis, Missouri
Operational evaluation of forecast and diagnostic atmospheric models in a forest canopy environment