17.3 Error and uncertainty in ensemble predictions of tropical storms (2009

17.3

Error and uncertainty in ensemble predictions of tropical storms

Jeffrey Anderson, NCAR, Boulder, CO; and C. Snyder, H. Liu, and J. Hacker

Ensemble predictions with initial conditions generated by the

Data Assimilation Research Testbed ensemble filter data assimilation

system and the WRF prediction model are evaluated for several

tropical storms during the TPARC field campaign. The ensemble

analyses and forecasts are compared to available observations

of storm position and intensity. Both analyses and forecasts are

generally found to have too little spread to be consistent

with the observations. The use of adaptive inflation and localization

algorithms in the assimilation increases the spread but not as much

as required. An analysis of the magnitude of the adaptive inflation

is used to estimate the spatial pattern and magnitude of assimilation

system errors. In addition, the impact of assimilating COSMIC radio

occultation observations, QuikSCAT surface winds, and best track storm

position estimates on the quality of the ensemble predictions is

assessed through data denial experiments.