Comparison of GOES‐17 Atmospheric Motion Vectors With AR Recon Dropsonde Data and Assessment of Wind Fields in the Global Forecast System During Atmospheric River Events

Abstract Atmospheric motion vectors (AMVs) represent horizontal wind derived by tracking cloud or water vapor features on successive satellite images. The launch of the Geostationary Operational Environmental Satellite‐R Series (GOES‐R), including GOES‐16 (GOES‐East) and GOES‐17 (GOES‐West), has significantly enhanced AMV data volume and geographic coverage over the contiguous United States (U.S.) and adjacent oceans. AMVs from GOES‐16/17 products can augment wind data in data‐sparse oceanic areas such as those frequented by atmospheric rivers (ARs). However, AMVs exhibit biases and uncertainties, especially due to height assignment issues, and there are fewer conventional data (e.g., radiosondes) to assess GOES‐17 AMVs over oceans. The AR Reconnaissance (AR Recon) samples ARs to improve forecast skill over the U.S. West and provides a unique opportunity to compare GOES‐17 AMVs. This study quantifies biases and uncertainties in GOES‐17 AMVs in the northeast Pacific using dropsondes from AR Recon, and assesses Global Forecast System (GFS) model wind analyses and background fields during AR events. Results for four representative AR cases show that GOES‐R AMVs improved wind data distribution compared to cases prior to GOES‐R becoming operational, particularly in the upper and lower troposphere. A comparison with dropsondes reveals a small vector wind speed bias of −0.7 m s −1 . The uncertainty for AMVs is estimated at 5–6 m s −1 . Comparison of collocated GFS model background wind fields shows small biases. Data assimilation reduces root‐mean‐squared differences, but the small biases in operational AMVs need further attention as they are a predominant wind data source in the GFS over oceanic regions.