Thompson, David: Jet Propulsion Laboratory / Caltech (Project Lead)
Project Funding:
2016 - 2019
NRA: 2016 NASA: Citizen Science for Earth Systems Program
Funded by NASA
Abstract:
Remote Visible / ShortWave InfraRed (VSWIR) spectroscopy, typified by NASA’s Next-Generation Airborne Visible/Infrared Imaging Spectrometer (AVIRIS-NG), is a powerful tool to map the composition, health, and biodiversity of Earth’s terrestrial and aquatic ecosystems. These studies must first estimate surface reflectance, removing atmospheric effects like absorption and scattering by water vapor and aerosols. Since atmospheric state varies spatially and temporally, and is insufficiently constrained by climatological models, it is important to estimate constituents directly from the VSWIR data. However, water vapor and aerosol estimation is a significant ongoing challenge for existing atmospheric correction codes. Conventional VSWIR atmospheric correction methods evolved from multi-band approaches, and do not fully utilize the rich spectroscopic data available. We hypothesize that a fully spectroscopic retrieval with a resolved (line-by-line) Radiative Transfer (RT) calculation can improve accuracy and ultimately reduce biases in global ecosystem investigations. The 2015-16 AVIRIS-NG India campaign offers an ideal opportunity to test this hypothesis. Tropical and subtropical environments, including many terrestrial and aquatic biomes of India, host a large fraction of Earth’s primary productivity and biodiversity. However, these regions are highly underrepresented in VSWIR atmospheric research. This is a serious gap, because many tropical environments show extremes in water vapor content, aerosol loadings, and absorbing aerosols from combustion byproducts, which challenge existing atmospheric correction methods. The India dataset is an extreme case, spanning tropical and mid-latitude climates, over 5 km of elevation changes, and a wide range of biomes. In this sense, the India dataset is a rare microcosm of the global challenge and a unique resource for refining atmospheric correction. The proposed project will use the recent AVIRIS-NG India dataset to significantly improve the state of the art for atmospheric correction of tropical atmospheres, reducing regional biases in global ecosystem studies. To accomplish this, we will exploit spectroscopic techniques that are already pervasive in atmospheric remote sounding disciplines. Specifically, we will use Optimal Estimation (OE) retrieval theory and a fast, fully resolved (line-by-line) RT solver. It will use the complete spectral measurement to recover a physical model incorporating aerosol and H2O, the interactions between scattering and absorption, and the coupling between surface reflectance and atmosphere. The OE approach addresses known shortcomings of conventional methods, and adopted widely, would raise VSWIR atmospheric correction to the standards of rigor and interpretability of atmospheric sounding missions.We will validate results using existing in-situ reference spectra, reflectance measurements from assigned partners in India, and objective spectral quality metrics for scenes without any ground reference data. We will evaluate the new models for both speed and accuracy and apply the best performer on the entire AVIRIS-NG India dataset for NASA and ISRO investigators. This will yield insight into the aerosol conditions in the India campaign, an important benchmark dataset. Upon completion, all advances would be folded into the standard AVIRIS-NG analysis and applied universally to benefit future campaigns. Independently, OE algorithms enable Degrees of Freedom (DOF) analyses to quantify the true information content of VSWIR spectroscopy for improving retrieval efficiency, a benchmark in our understanding of these sensors and an important milestone on the path toward consistent global studies.
Publications:
Thompson, D. R., Natraj, V., Green, R. O., Helmlinger, M. C., Gao, B., Eastwood, M. L. 2018. Optimal estimation for imaging spectrometer atmospheric correction. Remote Sensing of Environment. 216, 355-373. DOI: 10.1016/j.rse.2018.07.003
More details may be found in the following project profile(s):
