203
The THORPEX observation impact Inter-comparison experiment

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 24 January 2011
The THORPEX observation impact Inter-comparison experiment
Washington State Convention Center
Ronald Gelaro, NASA/GSFC, Greenbelt, MD; and R. Langland, S. Pellerin, and R. Todling

The first stage of an experiment to directly compare the impacts of observations in different forecast systems has been completed as part of The Observing System Research and Predictability Experiment (THORPEX) initiative to quantify the value of observations provided by the current global observing network in terms of numerical weather prediction. An adjoint-based approach was used to compare the impact of observations on 24-hr forecasts in three systems: the Goddard Earth Observing System (GEOS-5) of the NASA Global Modeling and Assimilation Office, the Navy Operational Global Atmospheric Prediction System (NOGAPS) of the Naval Research Laboratory, and the Global Deterministic Prediction System (GDPS) of Environment Canada. With the adjoint technique, the impacts of all observations are computed simultaneously from a single execution of the system, allowing results to be easily aggregated according to data type, location, satellite sounding channel, or other attribute. The technique is highly economical as compared with running multiple data denial or observing system experiments (OSEs), but its accuracy is generally limited to forecast ranges of 1-3 days.

Despite differences in the assimilation algorithms and forecast models, the impacts of the major observation types are similar in each forecast system in a global sense. However, regional details and other aspects of the results can differ substantially. Large forecast error reductions are provided by satellite radiances, geostationary satellite winds, radiosondes and commercial aircraft. Other observation types provide smaller impacts individually, but their combined impact is significant. In all systems, only a small majority (less than 55%) of the total number of observations assimilated actually improves the forecast, and most of the overall improvement comes from a large number of observations that have relatively small individual impacts. Both results point to the advantage of increasing the number of observations assimilated as opposed to seeking a more limited set that produces only the largest impacts, and to the potential importance of having some level of redundancy between observing systems. Accounting for this behavior may also be important when considering strategies for deploying adaptive (or ``targeted") components of the observing system.