Evaluation of Smoke Correction in AVIRIS Imagery

Understanding the effects of smoke on remote sensing imagery is important in environments where wildfires regularly occur and is needed for improving current atmospheric correction models. Current techniques are often able to reliably correct for scattering and absorption under typical atmospheric conditions, but fail in the presence of variable smoke. The measurements collected from the low altitude ER-2 flight equipped with the Airborne Visible / Infrared Imaging Spectrometer (AVIRIS) over Santa Barbara, California on June 17, 2016 during the 2016 NASA Student Airborne Research Program provided an opportunity to study smoke from the Sherpa wildfire with hyperspectral imagery. We atmospherically corrected the imagery obtained from the flight with the FLAASH and ATCOR atmospheric correction algorithms and evaluated our surface reflectance retrievals against reflectance data from 2011. We found that smoke intensity could be quantified using the reflectance values from the 462 nm band to categorize smoke as light, medium, and heavy. From the flight run we picked several regions of interest (ROIs) based on smoke intensity in order to study the impact of smoke on individual AVIRIS bands. Our evaluations revealed that as smoke intensity increases in a particular scene the coefficient of determination (R2) between reflectance from 2011 and 2016 decreases as a result in contrasting reflectances between 2011 and 2016. Our analysis of the atmospheric models revealed that current ATCOR and FLAASH models have the most variability in their apparent reflectance with wavelengths in the visible range followed by variability in the near-infrared. This study revealed the need of more future research on aerosols like smoke in order to provide improved atmospherically corrected imagery in the shorter wavelength bands of AVIRIS when in the presence of smoke.

Presentation Slides: Evaluation-of-Smoke-Correction-in-AVIRIS-Imagery