Estimation of high-resolution land surface net shortwave radiation from AVIRIS data: Algorithm development and preliminary results

Hyperspectral remote sensing provides unique and abundant spectral information for quantification of the land surface shortwave radiation budget, which can be used to calibrate climate models and to estimate surface energy budget for monitoring agriculture and urban environment. However, only single broadband or multispectral data have been used in previous studies. In the present study, two methods are proposed to estimate the instantaneous land surface net shortwave radiation (NSR) with high spatial resolutions using hyperspectral remote sensing observations from the Airborne Visible Infrared Imaging Spectrometer (AVIRIS) data. Method A calculates the NSR based on separate estimation of downward radiation and surface broadband albedo, which requires ancillary information for aerosol optical depth; and Method B directly estimates the NSR from the observed radiance. Results based on radiative transfer simulations showed that the use of hyperspectral data can significantly improve NSR estimation compared with the multispectral data method. Atmospheric water vapor correction was applied to adjust the surface radiation estimation. Validation of AVIRIS NSR estimates against ground measurements from two flux networks for the period of 2006-2014 showed that the two methods were similar and had consistent accuracy in the all-sky instantaneous NSR estimation with root-mean-square-errors (RMSEs) of approximately 28-56W/m2. The pixel-based water vapor content estimation from AVIRIS data provided slightly different results than those obtained using coarse resolution remote sensing data. A simplified topographic correction algorithm was found to be able to improve the results generated from Method A; however, the degree of improvement provided by Method B was unclear, possibly because of the lack of consideration of horizontal atmospheric scattering effects from adjacent pixels. In general, hyperspectral remote sensing data have been shown to improve the NSR estimation accuracies compared with results obtained in previous studies. Additional efforts are needed to refine the NSR estimation for application to future satellite hyperspectral data