![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
XCO2 Retrieval Errors from a PCA-based Approach to Fast Radiative Transfer |
VerfasserIn |
Peter Somkuti, Hartmut Boesch, Vijay Natraj, Pushkar Kopparla |
Konferenz |
EGU General Assembly 2017
|
Medientyp |
Artikel
|
Sprache |
en
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 19 (2017) |
Datensatznummer |
250143013
|
Publikation (Nr.) |
EGU/EGU2017-6702.pdf |
|
|
|
Zusammenfassung |
Multiple-scattering radiative transfer (RT) calculations are an integral part of forward models
used to infer greenhouse gas concentrations in the shortwave-infrared spectral range
from satellite missions such as GOSAT or OCO-2. Such calculations are, however,
computationally expensive and, combined with the recent growth in data volume, necessitate
the use of acceleration methods in order to make retrievals feasible on an operational
level.
The principle component analysis (PCA)-based approach to fast radiative transfer
introduced by Natraj et al. 2005 is a spectral binning method, in which the many line-by-line
monochromatic calculations are replaced by a small set of representative ones. From the PCA
performed on the optical layer properties for a scene-dependent atmosphere, the results
of the representative calculations are mapped onto all spectral points in the given
band.
Since this RT scheme is an approximation, the computed top-of-atmosphere radiances
exhibit errors compared to the “full” line-by-line calculation. These errors ultimately
propagate into the final retrieved greenhouse gas concentrations, and their magnitude depends
on scene-dependent parameters such as aerosol loadings or viewing geometry. An advantage
of this method is the ability to choose the degree of accuracy by increasing or decreasing
the number of empirical orthogonal functions used for the reconstruction of the
radiances.
We have performed a large set of global simulations based on real GOSAT scenes and
assess the retrieval errors induced by the fast RT approximation through linear error analysis.
We find that across a wide range of geophysical parameters, the errors are for the
most part smaller than ± 0.2 ppm and ± 0.06 ppm (out of roughly 400 ppm) for
ocean and land scenes respectively. A fast RT scheme that produces low errors is
important, since regional biases in XCO2 even in the low sub-ppm range can cause
significant changes in carbon fluxes obtained from inversions (Chevallier et al.
2007). |
|
|
|
|
|