Assimilating atmospheric motion vector winds using a feature track correction observation operator

Abstract Atmospheric motion vector (AMV) winds have positive impacts in operational numerical weather prediction (NWP) systems. These impacts might be improved with better treatment of the following error characteristics of AMVs. First, AMVs may have wind errors due to height assignment errors. Second, AMVs may have additional wind‐speed biases in addition to those due to height assignment errors. Third, AMVs are representative of motion in a possibly thick atmospheric layer, not a single atmospheric level. Previous work proposed a variational feature track correction (FTC) method in which an observation operator is implemented that averages the NWP background winds optimally in the vertical. Here, a prototype feature track correction observation operator (FTC‐OO) is implemented in the NOAA/NCEP data assimilation (DA) system. The parameters describing the vertical averaging are determined offline based on previous DA cycles. The FTC‐OO reduces the observation minus background standard deviation by about 4%. Global observing‐system experiments (OSEs) are performed comparing the FTC‐OO with the operational observation operator. The forecast verification sample is 41 10‐day forecasts. The OSEs show that the FTC‐OO improves forecast skill, primarily for tropical geopotential height. Additional OSEs are performed that include Aeolus wind observations. The hypothesis that the Aeolus winds would enhance the impact of the FTC‐OO was not borne out in these experiments—the Aeolus observations alone have a significant positive impact, but the impact of the FTC method in the presence of the Aeolus observations is neither enhanced nor degraded compared with the impact of the FTC method alone.