HATCH: results from simulated radiances, AVIRIS and Hyperion

The atmospheric correction program High-accuracy Atmospheric Correction for Hyperspectral data (HATCH) has been developed specifically to convert radiance from imaging spectrometer sensors to reflectance on a pixel-by-pixel basis. HATCH was developed to update the previously available model, the Atmosphere Removal model (ATREM). In this paper, we test the HATCH model against model data using MODTRAN 4 as well as with Airborne Visible/Infrared Imaging Spectroradiomter and Hyperion data for which simultaneous ground reflectances were acquired. We also compare HATCH to the commercially available ACORN model. Results show that HATCH produces smoother-looking spectra than its predecessor ATREM and is less influenced by liquid water in vegetation. Comparisons with MODTRAN were made by calculating above-atmosphere radiances from a hypothetical target with a 0.5 reflectance at all wavelengths between 400 and 2500 nm and retrieving them with HATCH. The results show a maximum deviation of 10% at several wavelengths, highlighting the differences between the models. Hyperion images contain striping artifacts. We show that optimum removal is obtained by normalizing the means and standard deviations of each column after converting radiance to reflectance with an atmospheric model. HATCH produces water vapor corrections virtually unaffected by vegetation liquid water if the water vapor band at 940 nm is used in the calculation. Retrievals using the 1140-nm band are subject to errors associated with liquid water in vegetation. Retrievals of reflectance from Hyperion radiances require use of the 1140-nm band because the 940-nm band falls in the detector crossover region.