Verifying forecasts of smoke and air quality with satellite products The National Weather Service (NWS) offers experimental forecasts of the location and concentration of smoke using a model called Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT), developed by the NOAA Air Resources Laboratory (ARL). In 2006 NWS asked NESDIS to provide observational support for verifying the accuracy of smoke forecasts in near real time. To fulfill this request, NESDIS scientists developed a procedure based on pattern recognition to identify smoke plumes, or clouds of smoke, using fire and aerosol information from geostationary satellite data. The amount of overlap between observed and forecast smoke concentrations is used to calculate a forecast skill, as a figure of merit score.
Learn more: |
|
|
|
Comparison: A prediction of HYSPLIT smoke analysis concentration from the
HYSPLIT model (left) is
compared with a smoke observation from GOES-12 smoke
observation (right), for September 9, 2006, 16 UTC at mid-day. Observed
plumes from the GOES-12 Imager are well forecast by the HYSPLIT model in
eastern Canada. Forecast skill (figure of merit scores) for 1, 5, and 20
µg/m3 contours for this episode were 45%, 17%, and 16% respectively.
Forecast skill can only be calculated for the cloud-free regions. The
presence of clouds in the satellite field of view prevented the observation
retrieval of smoke aerosols by satellite observations over the central part
of Canada and the western United States (entire region in gray).
