The LaRC trajectory model (LTM) (Pierce et al., 1997) was run in time forward mode for the July-August 1989 time period. The model was seeded with fire pixels as recognized by the Baum and Trepte (1999) smoke/fire/cloud mask. Meteorological data was taken from the European Centre for Medium-Range Weather Forecasts 15-year reanalysis (ERA-15). Air parcels were traced from fire regions over a period of five days globally on a 2.5° equal angle grid. Areas with large numbers of trajectories originating from boundary layer air in fire regions were considered to be likely to be influenced by smoke.
An analysis of the effects of the smoke from the Manitoba fires at Canadian surface radiometric sites was performed. Hourly shortwave surface fluxes were compared to a variety of radiative transfer model calculations, satellite data products, and trajectory model results. Numbers and ages of fire-originating trajectories passing within a two degree radius of each of the sites were collected. A theoretical "clean sky" (i.e., with no aerosols or clouds) surface shortwave flux was calculated at each surface site for each hour in July and August using the Fu-Liou radiative transfer code with TOMS ozone values and GEOS-1 meteorology.
July 23, 1989 was taken as a first test case. Smoke was heavy over the fire region of Manitoba and Ontario, and both the trajectory model and TOMS showed a tongue of smoke advected far east of the fire region, over the Canadian maritime provinces. At solar noon over each of five stations (Winnipeg, Moosonee, Big Trout Lake, Fredericton, and Outlook), the difference between the theoretical clean sky flux and the measured flux was noted. Moosonee and Big Trout Lake, in the heavy smoke area, had smoke forcings in excess of 200 Wm-2. Outlook, SK, to the west of the fire area had no smoke trajectories, and a forcing of only 9 Wm-2 . Fredericton, NB, was under a tongue of smoke according to the trajectory model and the TOMS Aerosol Index, and had a forcing of 88 Wm-2. This is taken as a proof of concept; that the tools are available to identify areas affected by forest fire smoke, and to quantify that effect.
The methods were applied to the entire July/August 1989 fire period, at twenty-four Canadian surface sites, both in and out of the fire region. Methods of identifying cloudless conditions were developed and are described. Stations which were determined by the trajectory model to be likely under the influence of smoke showed solar noon forcings approximately 50 Wm-2 higher than other stations.