system is used with a recent version of NCAR's Community Atmospheric

Model to create atmospheric reanalyses for September 2006 through

August 2007. Reanalyses using conventional observations including

radiosondes, aircraft, and satellite cloud drift motion vectors are

compared to ones that also use available radio occultation soundings

from the COSMIC satellites. Differences in bias and root mean square

error in short term forecasts from the two reanalyses are used to

quantify the ability of COSMIC observations to improve both climate

analyses and short term forecasts. Special emphasis will be given to

results over tropical oceans where conventional observations are

sparse and many remote sensing observations can suffer from

contamination by clouds. Results of assimilating using a

local refractivity forward operator will be compared to results

using a nonlocal operator.