Estimating the Impact of Assimilating AEOLUS Winds and Mesoscale Atmospheric Motion Vectors on Tropical Cyclone Forecasts: A Case Study with Hurricane Dorian

Tropical cyclone (TC) forecasts can be improved when models are initialized with an accurate representation of the wind field. To this end, forecasts of the winds from a previous cycle are combined with observations in and around the storm to provide the best initial conditions for the current cycle. Wind observations come from several sources, including scatterometers, aircraft reconnaissance missions, satellite measurements, and other sources. In this presentation, we evaluate the improvement of TC forecasts when observations from two of these sources are used: lidar wind profiles measured from the AEOLUS satellite and high-resolution mesoscale atmospheric motion vectors (AMVs) derived from satellite observations. These measurements provide information about different aspects of the wind field and are available under different conditions. For AEOLUS winds to be available within the model domain, the satellite’s orbit needed to pass near the storm. When present, AEOLUS profiles provided measurements within the storm and in areas of the storm environment that otherwise had limited or no observations. Unlike AEOLUS winds, AMVs can be retrieved through the tracking of cloud and water vapor features in satellite images from geostationary satellites, making them consistently available. Through data assimilation, the forecasted wind field from a previous model cycle is updated by the observed wind field of the storm and its environment to provide the initial conditions for a new forecast.

In this case study, we use HWRF to investigate if tropical cyclone forecasts can be improved by assimilating AEOLUS and high-resolution mesoscale AMV winds. To test this, we evaluate 63 forecast cycles from Hurricane Dorian during the 2019 Atlantic hurricane season. In a series of Observing System Experiments (OSEs), we systematically include observations from these sources to test their impact on the forecasts. In particular, analysis and prediction skill is assessed using common metrics, such as hurricane track, intensity, and radii of 34-, 50-, and 64-knot winds. Our control experiment consists of assimilating all non-restricted conventional and satellite observations available in operations, excluding those from AEOLUS. We note that global scale AMVs are included in this set of observations. In a second experiment, we add AEOLUS winds to the suite of observations being assimilated. In a third experiment, we expand the use of AMVs by including high-resolution, mesoscale AMV observations. We include these observations both with and without the global scale AMVs present. Our final experiment includes both the AEOLUS and mesoscale AMVs. Similar to the previous experiment, we test using both sources of AMVs together and the mesoscale AMVs alone.