Superobbing vs. Thinning Atmospheric Motion Vectors: An Evaluation of Methods Used at FNMOC and ECMWF

Observation impacts based on the adjoint of the modeling and data assimilation systems show that the U.S. Navy's global Numerical Weather Prediction (NWP) system has greater beneficial impacts from Atmospheric Motion Vectors (AMVs) derived from geostationary satellite imagery than most other operational NWP centers. Recent work by our group and by Gelaro and Merkova at NASA Goddard/GMAO suggests that the increased beneficial impact is primarily a result of assimilating a larger number of AMVs, since the Navy system uses AMVs from both the operational providers (NESDIS, EUMETSAT, and JMA) and from the University of Wisconsin/CIMSS and the Air Force Weather Agency (AFWA). However, there has also been a suggestion that the NRL superobbing procedure has some advantage over thinning in terms of producing observation impacts that are more spatially consistent.

To investigate this question in more detail, a pair of experiments using the U.S. Navy's NWP system is presented that compares the NRL superobbing procedure with the ECMWF thinning procedure. The NRL superobbing procedure collects AMVs from a particular satellite and channel in a “prism” that is 2° latitude by 2° longitude at the equator and averages the u and v components to form superobs only if the individual AMVs to be averaged are sufficiently similar. If they are not, an attempt is made to form a superob after rejecting one or two outliers; failing that, the prism is divided into quarters horizontally, and an attempt is made to form superobs within each quarter. The ECMWF thinning procedure simply selects the AMV with the greatest Quality Indicator (QI) value. The experiments are run using data after January 2012 so that hourly winds are available from all satellites and providers. Verification is done in comparison to both self-analyses and ECMWF analyses in addition to an evaluation of observation impact.