Applications of GOES-16/17 Atmospheric Motion Vectors to the Initialization and Forecast of Atmospheric Rivers in NWP Models

Atmospheric motion vectors (AMVs) are horizontal wind observations that are derived by tracking the cloud or water vapor features on successive satellite images. With the launch of Geostationary Operational Environmental Satellite-R Series (GOES-R), the Advanced Baseline Imager (ABI) was deployed onboard GOES-16 (GOES-East, which became operational on December 18, 2017) and GOES-17 (GOES-West, which became operational on February 12, 2019). These satellites have significantly improved data volume and geographic coverage for research and operations. Meanwhile, Zheng et al. (2021) identified significant data gaps within the atmospheric rivers (ARs) over the Northeast Pacific Ocean. The GOES-16/17 AMVs derived from ABI products could potentially augment the wind data in ARs. However, the uncertainty in height assignments could lead to substantial biases and errors in AMV products. Recent Atmospheric River Reconnaissance (AR Recon, Ralph et al, 2021), a field program that samples ARs and the surrounding areas to improve the forecast skill of the US West, has become operational since 2019 and provided great opportunities to validate GOES-16/17 AMVs over the Northeast Pacific Ocean. This study has three main objectives: 1) to quantify the biases and uncertainties of GOES-16/17 products around ARs over the Northeast Pacific Ocean by using dropsonde profiles collected during AR Recon; 2) to quantify the model background and analyses; and 3) to evaluate the impact of assimilating AMVs on the forecast skill of landfalling ARs through impactful case studies, using a regional WRF model and 4DEnVar system.

The GOES-16/17 AMV data used for this study are the operational AMVs available in the National Center for Environmental Prediction (NCEP) Global Forecast System (GFS). We first developed a collocation method for comparing dropsonde profiles with the nearby GOES-16/17 AMVs. A total of 1,718 dropsonde profiles collected from AR Recon 2020-21 are used for validating the AMVs. AMVs are grouped into the IR long-wave, IR short-wave, cloud-top, and deep-layer, according to the ABI bands. We computed the biases and errors of each AMV type using the collocated pairs in and near ARs over the Northeastern Pacific. Then, the differences between AMVs and GFS background and analyses were calculated. The results showed that biases in GFS background and analyses over the eastern Pacific Ocean can inherit those from AMVs, which dominate the wind observations over oceans. Next, the impacts of different types of AMVs on skill improvement were assessed through data denial experiments during two AR cases in February 2020 and January 2021. Finally, the capability for GOES-16/17 AMVs and the newly launched GOES-18, together with the AR Recon data, to alleviate data gaps in ARs will be discussed.