GPS radio occultation measurements are vital for climate monitoring and atmospheric temperature change detection. However, the data are irregularly distributed in space and time, which makes it inconvenient for many applications.
This innovation is a developed and implemented Bayesian interpolation technique for gridding such irregular atmospheric retrievals. This technique has been applied on over ten years of continuous GPS radio occultation (RO) measurements observed from the CHAMP and FORMOSAT-3/COSMIC spacecraft towards the generation of global monthly gridded climatologies of geopotential height and temperature in the upper troposphere and lower stratosphere (UTLS).
This software is a level 3 gridded data product. The initial focus is on geopotential height (gravity-adjusted height) and temperature records. The Bayesian interpolation approach, first introduced by Leroy (1997), has been extended and further validated (Leroy et al. 2012). Three main steps are performed for each satellite.
The first step is to combine all retrieved atmospheric temperature and geopotential height data from single ROs into sub-monthly (3- to 5-day) bins, and to sort them by the atmospheric pressure levels and output into NetCDF3.0 files. The second step is to apply the Bayesian interpolation formalism to calculate interpolation coefficients and then perform interpolation to obtain smoothed, gridded global maps of atmospheric temperature and geopotential height for each sub-monthly bin. The third step is to compute their monthly averages by averaging over all sub-monthly bins.
Bayesian interpolation uses a set of basis functions (spherical harmonics is this case) to fit to irregularly sampled data with unknown noise characteristics. What makes it Bayesian is its use of a regularizing function — much like a penalty function — that is weighted against data misfit to optimally resolve structures in the data without over-fitting the data. To reduce sampling error due to limited orbital and temporal coverage, the software performs averaging over one-month intervals. While GPS ROs do not resolve synoptic variability, they still can resolve larger-scale, longer-lived structures in the atmosphere such as the equator-to-pole temperature gradient and long-lived waves. A longer averaging period permits more soundings for improved averaging; however, it reduces the ability of sampling to resolve large-scale transient structures.
This work was done by Chi O. Ao and Olga P. Verkhoglyadova of Caltech, and Stephen Leroy of Harvard School of Engineering for NASA’s Jet Propulsion Laboratory. NPO-48993