Abstract
AbstractIn this work, we extend the principal component analysis (PCA)‐based approach to accelerate radiative transfer (RT) calculations by accounting for the spectral variation of aerosol properties. Using linear error analysis, the errors induced by this fast RT method are quantified for a large number of simulated Greenhouse Gases Observing Satellite (GOSAT) measurements (N≈ 30,000). The computational speedup of the approach is typically 2 orders of magnitude compared to a line‐by‐line discrete ordinates calculation with 16 streams, while the radiance residuals do not exceed 0.01% for the most part compared to the same baseline calculations. We find that the errors due to the PCA‐based approach tend to be less than ±0.06 ppm for both land and ocean scenes when two or more empirical orthogonal functions are used. One advantage of this method is that it maintains the high accuracy over a large range of aerosol optical depths. This technique shows great potential to be used in operational retrievals for GOSAT and other remote sensing missions.
Highlights
Carbon dioxide (CO2) is the most important anthropogenic greenhouse gas, and its increase from the beginning of the industrial era has contributed 2.2 ± 1.1 W/m2 to the global radiative forcing (Myhre et al, 2013)
The radiance residuals are below that of those reported in alternative methods when more than 1 empirical orthogonal functions (EOFs) is used, at the cost of more binned calculations when compared to low-streams interpolation (LSI) or linear-k
To understand the estimated errors in terms of their magnitude, the XCO2 errors due to the principal component analysis (PCA) approach can be compared to other errors that make up the total error budget of a retrieval
Summary
Carbon dioxide (CO2) is the most important anthropogenic greenhouse gas, and its increase from the beginning of the industrial era (middle to late eighteenth century) has contributed 2.2 ± 1.1 W/m2 to the global radiative forcing (Myhre et al, 2013). For both LSI and linear-k, the only tuneable parameter, apart from the number of quadrature streams employed for the high-accuracy calculations (which is a parameter of the employed RT models, rather than the acceleration scheme), is the set of bin boundaries (usually approximately logarithmically spaced in the gas optical depth dimension) which has to be chosen by the user for the specific spectral ranges While this is true for the PCA-based approach as well, it offers the opportunity for the user to choose the number of empirical orthogonal functions (EOFs) used in the reconstruction of the radiances.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
