Analyzing structural uncertainty in rOPS and GPAC/ROPP processing: the chain from bending angle to dry-air atmospheric profiles
Presenter:
Veronika Proschek
Wegener Center for Climate and Global Change (WEGC) and Institute for Geophysics, Astrophysics and Meteorology/Institute of Physics, University of Graz, Graz, Austria
(1) Wegener Center for Climate and Global Change (WEGC) and Institute for Geophysics, Astrophysics and Meteorology/Institute of Physics, University of Graz, Graz, Austria (2) Danish Meteorological Institute (DMI), Copenhagen, Denmark
Poster
In a joint project of WEGC and the ROM SAF we investigated structural uncertainties within the Level 2a (L2a) processing chain of the radio occultation (RO) retrieval algorithms of two different processing systems, namely the GNSS Processing and Archive Center (GPAC) implementation of the Radio Occultation Processing Package (ROPP) at DMI used for the generation of the first ROM SAF Climate Data Record and the Reference Occultation Processing System (rOPS) used for R&D processing by WEGC, with focus on validation and climate studies. We understand L2a structural uncertainty in this context as the part of the uncertainty emerging in retrieved profiles that derives from different plausible algorithmic choices and numerical implementations in the L2a retrieval steps of rOPS and GPAC/ROPP when we supply both processing systems with identical input data.
The L2a chain of both systems contains three retrieval steps with potential for such structural uncertainty. These are the high-altitude initialization of bending angles by statistical optimization, the refractivity retrieval by an Abelian integral, and the dry-air pressure and temperature retrieval by a hydrostatic integral followed by a local equation-of-state conversion. The statistical optimization algorithms differ significantly in algorithm choices and the source of background profiles, with rOPS using ECMWF/ERA5 short-range forecasts and GPAC/ROPP a climatological profile database (BAROCLIM). The Abelian and hydrostatic integrals for the rOPS employ a baseband approach for minimizing bias effects, the ROPP uses careful discretization. The equation-of-state conversion to temperature may differ only in choices of constants since both systems use an ideal-gas formulation. We tightly coupled rOPS and ROPP in such a way that each of the three L2a retrieval steps can be feeded with identical or differing inputs, starting with bending angles received from either ROPP or rOPS L1b processing, or statistically optimized bending angles (entry second step), or refractivities (entry third step).
We present the structural differences between rOPS and GPAC/ROPP for each of the L2a steps which we analyzed based on a thoroughly selected very diverse RO dataset of test months (covering different resolutions of ECMWF background profiles, low and high solar activity conditions, special atmospheric conditions like severe sudden stratospheric warming events, and using multiple RO mission data from CHAMP, GRACE-A, Metop-A/B, and COSMIC). We find that the main differences are induced by different statistical optimization processes in the first L2a step, from the different background profiles and weighting approaches at the high altitudes from about 60 km to 120 km. The second and third step implementations, while different in numerical algorithm choices, show essentially negligible structural difference, pointing to robust implementations in both systems. We also show results from complementary validations from cross-evaluating the L2a retrieval performance based on independent profiles from MIPAS and SABER.
Poster in PDF:
