Real-time mesoscale ensemble data assimilation for Atlantic TC

It has been hypothesized that improved data assimilation in the vicinity of tropical cyclones could improve both track and intensity forecasts. Ensemble data assimilation algorithms, such as the ensemble Kalman filter (EnKF), offer an alternative to variational or vortex bogusing methods because observation information is incorporated into the model state vector via flow-dependent error statistics, which should take into account the unique dynamics of a TC. To date, these techniques have shown promise in individual case studies; however, this approach has not been tested over longer periods of time.

For the 2009 Atlantic Hurricane Season, the convective-resolving TC forecasts at the National Center for Atmospheric Research (NCAR) were initialized from one member of a cycling mesoscale EnKF system coupled to the advanced research version (ARW) of WRF. The EnKF system used here assimilates observations each six hours from conventional data, including synoptic dropsondes, and TC advisory position and intensity data. This talk will describe the system setup and performance of the assimilation system over the entire season. Although observations had difficulty correcting the position of some storms (i.e., Erika and Grace), most analysis times were characterized analysis position errors of less than 50 km. In contrast, the intensity errors were comparable to both HWRF and GFDL. Although no TC-specific wind information was assimilated, the analysis and prior forecast 34 knot wind radii errors are less than both HWRF and GFDL.