This is part of a larger mesoscale simulation study, for generating satellite-viewed radiances and brightness temperatures for simulated cases. In the fire simulation, the fires were embedded in a summertime severe weather case, such as occurred on 8 May 2003. (The other cases are a lake effect snow case and a hurricane case, the results of which are covered elsewhere.)
For the wildfire case, fires were simulated at 400 m spatial and 5 minute temporal resolution in a model grid over Kansas with real terrain. The fires were set in a 9 x 15 grid, with 9 rows of increasing fire temperature from 400 K (bottom) to 1200 K (top) in 100 K increments. The 15 columns on varying fire size start with a 1 x 1 pixel fire (on the left), to 2 x 1 pixels, to 2 x 2 pixels, and ending with 8 x 1 pixels, and finally 8 x 8 pixels (on the right).
Data and images of those fires were then generated at both GOES-R ABI and GOES-12 Imager spatial and temporal resolutions in order to test the improvement in fire detection that will be possible with ABI over current GOES. The simulations included three GOES-R ABI bands (3.9, 10.35, and 11.2 um) and two GOES-12 Imager bands (3.9 and 10.7 um), in order to compare current GOES to future GOES-R. The fire cases included three different scenarios: 1) constant fires with no clouds; 2) constant fires with clouds; and 3) varying/flickering fires with no clouds.
The fire scene was simulated for a total of 6 hours, starting at 1800 UTC and ending at 2400 UTC. The GOES-R ABI simulations included 72 time steps at a 5 minute increment, and the GOES-12 simulations included 24 time steps at a 15 minute increment. Both of these are the default time intervals for the respective satellite imagery.
There is also a large spatial resolution difference between the simulations for GOES-R ABI and GOES-12. The anticipated point spread function (PSF) for GOES-R ABI was used to turn the 400 m fires into ABI imagery at approximately 2.4 x 3.2 km resolution at the Kansas location (2 x 2 km at sub satellite point) assuming GOES-R to be located at the 75°W location, same as GOES-12. For GOES-12 the known PSF was used to generate imagery at approximately 4.8 x 6.4 km resolution at the Kansas location (4 x 4 km resolution at the sub-satellite point).
Loops of the simulated GOES-R ABI imagery are displayed below for each of the three spectral bands and for each of the three different fire cases, for a total of 9 simulations. Simulations for GOES-12 were for the constant fire, no-cloud case only and for only two spectral bands, as a baseline for comparison between current and future GOES.
After the imagery is analyzed by the fire detection software, the results will be compared to the actual fire sizes and temperatures. The expected result will be a large variation in detected fires even for the same fire temperature and size. That will likely be due to the sampling of the sub-pixel (400 m) fires by the larger spatial resolution of both GOES-R ABI and GOES-12 pixels. These simulations will allow bounds to be set on the ability to quantitatively assess fires from GOES. Results will appear here as they are obtained.GOES-12 Central America 24 Apr 04