EnKF Data Assimilation of dual-Doppler radar data from hurricane Guillermo

In this work we present the assimilation of dual-Doppler radar observations for rapidly intensifying hurricane Guillermo (1997) using the Ensemble Kalman Filter (EnKF). A unique aspect of Guillermo was that during the period of radar observations strong convective bursts, attributable to wind shear, formed primarily within the eastern eyewall. To reproduce this observed structure within a hurricane model, background wind shear of some magnitude must be specified; as well as turbulence parameters appropriately adjusted so that the impact of the shear on the simulated hurricane vortex can be realized. To help understand this complex nonlinear interaction between shear and modeled turbulence, an ensemble of simulations have been conducted during which these key parameters were varied over a certain range and then sampled via a Latin Hypercube approach. Further, given the computational resources of Jaguar, a DOE parallel machine residing at Oakridge National Lab that is comprised of 220,000 processors, relatively large ranges of these parameters could be sampled with the impact of ensemble size on parameter estimation, via EnKF, being a key piece of the work to be presented, i.e., results from ensemble sizes consisting of at least 500 members will be shown. Thus, given these large computational resources, it is now possible to begin to examine whether or not hurricane models can actually reproduce the structure of a given hurricane, in addition to intensity and track. Another important aspect of this work is that the EnKF method was matrix-free and employed a unique preconditioner. The matrix-free EnKF approach enabled the large amount of dual-Doppler radar data to be readily incorporated during the assimilation stage: this development was also essential for an accurate estimation of the various turbulence parameters and environmental wind shear. Details concerning the matrix-free implementation of the EnKF method along with results illustrating the ability of the EnKF approach to determine the appropriate combination of parameters necessary to reproduce the observed structure of Guillermo will be presented during the talk.