Simulating arbitrary hyperspectral bandsets from multispectral observations via a generic Earth Observation-Land Data Assimilation System (EO-LDAS)

Chernetskiy, Maxim; Gobron, Nadine; Gómez-Dans, José; Morgan, Olivier; Disney, Mathias; Lewis, P.; Schmullius, Christiane

doi:10.1016/j.asr.2018.07.015

Cited by 2 publications

(2 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Similar approaches have been used by the Earth science community to reduce the computational burden of forward modeling of spectra, retrieval of surface conditions, and atmospheric correction (e.g., Atzberger 2004;Garcia-Cuesta et al 2009;Rivera et al 2015;Verrelst et al 2015;Gómez-Dans et al 2016;Verrelst et al 2016Verrelst et al , 2017Chernetskiy et al 2018;Yin et al 2018;Vicent et al 2018;Bue et al 2019).…”

Section: Introductionmentioning

confidence: 99%

Accurate Machine-learning Atmospheric Retrieval via a Neural-network Surrogate Model for Radiative Transfer

Himes¹,

Harrington²,

Cobb³

et al. 2022

Planet. Sci. J.

View full text Add to dashboard Cite

Atmospheric retrieval determines the properties of an atmosphere based on its measured spectrum. The low signal-to-noise ratios of exoplanet observations require a Bayesian approach to determine posterior probability distributions of each model parameter, given observed spectra. This inference is computationally expensive, as it requires many executions of a costly radiative transfer (RT) simulation for each set of sampled model parameters. Machine learning (ML) has recently been shown to provide a significant reduction in runtime for retrievals, mainly by training inverse ML models that predict parameter distributions, given observed spectra, albeit with reduced posterior accuracy. Here we present a novel approach to retrieval by training a forward ML surrogate model that predicts spectra given model parameters, providing a fast approximate RT simulation that can be used in a conventional Bayesian retrieval framework without significant loss of accuracy. We demonstrate our method on the emission spectrum of HD 189733 b and find good agreement with a traditional retrieval from the Bayesian Atmospheric Radiative Transfer (BART) code (Bhattacharyya coefficients of 0.9843–0.9972, with a mean of 0.9925, between 1D marginalized posteriors). This accuracy comes while still offering significant speed enhancements over traditional RT, albeit not as much as ML methods with lower posterior accuracy. Our method is ∼9× faster per parallel chain than BART when run on an AMD EPYC 7402P central processing unit (CPU). Neural-network computation using an NVIDIA Titan Xp graphics processing unit is 90×–180× faster per chain than BART on that CPU.

show abstract

Section: Introductionmentioning

confidence: 99%

Accurate Machine-learning Atmospheric Retrieval via a Neural-network Surrogate Model for Radiative Transfer

Himes¹,

Harrington²,

Cobb³

et al. 2022

Planet. Sci. J.

View full text Add to dashboard Cite

show abstract

“…Although the task of synthetic data generation can be accomplished with process‐based vegetation and radiative transfer models (Chernetskiy et al., 2018; Poulter et al., 2023), the computational costs associated with such global model simulations at landscape resolutions are very expensive. In this study, therefore, we seek a physics‐based but computationally efficient approach to address the task.…”

Section: Introductionmentioning

confidence: 99%

Development of the Ames Global Hyperspectral Synthetic Data Set: Surface Bidirectional Reflectance Distribution Function

et al. 2023

View full text Add to dashboard Cite

This study introduces the Ames Global Hyperspectral Synthetic Data set (AGHSD), in particular the surface bidirectional reflectance distribution function (BRDF) product, to support the NASA Surface Biology and Geology (SBG) mission development. The data set is generated based on the corresponding multispectral BRDF products from NASA's MODIS satellite sensor. Based on theories of radiative transfer in vegetation canopies, we derive a simple but robust relationship that indicates that the hyperspectral surface BRDF can be accurately approximated as a weighted sum of the soil surface reflectance, the leaf single albedo, and the canopy scattering coefficient, where the weights or coefficients are spectrally invariant and thus readily estimated from the multispectral MODIS products. We validate the algorithm with simulations by a Monte Carlo Ray Tracing model and find the results highly consistent with the theoretic derivation. Using reflectance spectra of soil and vegetation derived from existing spectral libraries, we apply the algorithm to generate the AGHSD BRDF product at 1 km and 8‐day resolutions for the year of 2019. The data set is biogeochemically and biogeophysically coherent and consistent, and serves the goal to support the SBG community in developing sciences and applications for the future global imaging spectroscopy mission.

show abstract

Simulating arbitrary hyperspectral bandsets from multispectral observations via a generic Earth Observation-Land Data Assimilation System (EO-LDAS)

Cited by 2 publications

References 45 publications

Accurate Machine-learning Atmospheric Retrieval via a Neural-network Surrogate Model for Radiative Transfer

Accurate Machine-learning Atmospheric Retrieval via a Neural-network Surrogate Model for Radiative Transfer

Development of the Ames Global Hyperspectral Synthetic Data Set: Surface Bidirectional Reflectance Distribution Function

Contact Info

Product

Resources

About