(Internal) FLAASH vs. ATREM, ACORN, or ATCOR - what are the differences?

FLAASH (Fast Line-of-Sight Atmospheric Analysis of Spectral Hypercubes) is an atmospheric correction plug-in for ENVI. This Help article describes how FLAASH compares to other available atmospheric correction modules.

The difference between the FLAASH and ATREM, ACORN, ATCOR is the accuracy of the correction. All of the model based atmospheric correction programs rely on MODRAN to model radiative transfer in the atmosphere. However, FLAASH is the only one that actually includes MODTRAN in the code -- all the others use a pre-calculated database of results from which they interpolate answers based on the scene collection details. FLAASH uses the input parameters directly to set up and run MODTRAN, so its solution is more accurate (since there's no interpolation). For simple cases, the base results are going to be very similar, but for more challenging conditions (i.e., urban haze, very wet atmosphere, cloudy, high contrast targets) FLAASH will be significantly more accurate. ATREM, in particular, is not terribly accurate at the shortest and longest wavelengths, while FLAASH (or ACORN) should be significantly better. Also, because FLAASH actually runs MODTRAN instead of relying on precalculated results, you can tweak the model settings in a couple of different ways. You an increase the MODTRAN resolution (from 15 cm-1 to 1 cm-1) which should improve the correction in the longer wavelengths (especially around the 2000 nm CO2 feature); and, you can change the multiscatter model which is used to reduce the effects of aerosol haze (from a simple ISAACS model all the way up to a 16-stream DISORT model) which should improve the results in the uv-blue-green where most of the scattering occurs. Like ATREM (and ACORN) FLAASH outputs a water vapor image, but FLAASH also outputs a simple cloud mask (identifying high altitude and opaque clouds). ATREM offers no artifact/error suppression, so you're limited to post processing with things like smoothing or EFFORT. FLAASH offers a very flexible "spectral polishing" that we've found to be quite good. FLAASH also includes a correction for what we're calling the "adjacency effect", where stray light from neighboring pixels contaminates the spectra. This can be pronounced for high spectral contrast targets (like a man-made object surrounded by vegetation, where you often see a red edge that's not real). And lastly, FLAASH includes an algorithm to estimate the scene average visibility and aerosol amount.

    The second release of FLAASH will support multispectral sensors, include a wavelength calibration utility, and offer a bit more control in choosing which water absorption features to use in estimating the water vapor amount -- initially it used only the 1135 nm feature unless the sensor was set to CASI (which didn't go out that far) but it will now let the user choose the feature to use (820 nm, 940 nm, or 1135 nm). The features that are in the works for the 3rd release include a special multi-resolution version of MODTRAN (which increases the resolution at only the wavelengths where you need it so its doesn't impact processing speed so much), a special version for scenes in the littoral zone and over water, which includes a bathymetry extraction, and a correction for spectral SMILE (which affects many instruments, like HYPERION).