The ASTER TIR Atmospheric Correction Task


The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) is one of several instruments on Terra, the first of a fleet of satellites that make up the Earth Observing System. Originally scheduled for launch in June of 1998, a long string of delays and problems were finally overcome, and Terra launched from Vandenberg AFB, a few seconds before 10:57 AM PST on Saturday, December 18, 1999. As a science team member, I was there to watch the launch. We were far enough away that it did not sound any louder than I expected, but it was brighter than I expected. Looking at the rocket flame was just like looking at the sun, really too bright to look at. This should not be surprising in retrospect, but this was the first (and so far only) rocket launch I have seen up close.

The ASTER instrument has 3 down-looking radiometers, a visual and near infrared radiometer (VNIR), which covers the wavelength range 0.52 - 0.86 microns; the short wave infrared radiometer (SWIR), 1.6 - 2.4 microns, and the thermal infrared radiometer (TIR), 8.1 - 11.3 microns. I worked with the TIR instrument science team. The goal of the TIR is to retrieve the land surface temperature and emissivity. Water is relatively easy to do because of its homogenous nature; it's pretty much the same everywhere. Not so for land, which can be bare rock, sandy desert, forest, prairie, and so on. Temperature measurements can tell us about energy flow from the ground to the air, or vice-versa. Emissivity measurements can tell us about mineralogical composition. The latter is especially poorly known on a global scale, and ASTER will be the first attempt to create reliable satellite based mineral maps on a global scale. The relatively small 90 meter pixel size for, wavelengths this long, allows for detailed examination of relatively small areas, and coupled with the temperature data, makes ASTER TIR an excellent instrument for locating and tracking thermal pollution, diagnosing the effect of land use and land cover changes, and for tracking volcano gas plumes (which can be nearly invisible in the visual wavelength range).

In order for all this to work right, the effect of the atmosphere on the thermal infrared observations must be removed. Although we call this atmospheric correction, a better phrase would be atmospheric compensation; the atmosphere is not wrong, and therefore in need of correction, it is simply there, and we want to compensate for that fact in the final data product. We use the MODTRAN atmospheric transmission code package, originally developed at the USAF Phillips Laboratory, as the kernel for our own atmospheric correction package; MODTRAN does all of the radiative transfer calculations. The problem is how to decide on the input. We needed software that could run in an automated, data pipeline environment. After significant research, we decided to use a data product produced by the National Climate Data Center (NCDC), which consists of vertical chemical & physical profiles of the atmosphere, distributed globally on a 1x1 degree grid, and updated every 6 hours. We take the 4 profiles surrounding the grid area that includes the observation location, and do a bi-linear interpolation between pairs to interpolate that profile at the observation location, including tests to avoid accidentally creating unphysical profiles (which turns out, as far as I know, to be a problem we anticipated, but that never arose in practice). We validated the automated system by traveling to various locations around the world with suitable target areas, and comparing radiance derived from in-situ data to radiance derived as the standard data product. Our automated system came within 1% of the in-situ results.




Page updated, and URL's checked: 2 February 2015


Back to Tim Thompson's Home Page