5A.5
Improved Data Assimilation in the Study of Hurricane Floyd (1999)
Jenni L. Evans, BMRC, Melbourne, Vic., Australia and Penn State Univ., University Park, PA; and R. E. Hart and C. S. Velden
Hurricane Floyd (1999) was an intense hurricane that battered the east coast of the USA before transitioning to an extratropical cyclone and moving into the Canadian region. Flooding rainfall from Floyd devastated parts of North Carolina. Here we present a case study of Floyd that incorporates high-resolution model analyses and simulations as tools to understand the evolution of this storm's life cycle. A Community Hurricane Modeling System (CHUMS) is presently under development at Penn State. CHUMS is based on the Penn State/NCAR Nonhydrostatic Mesoscale Model (MM5) tailored for tropical forecasting. Initially our focus has been on the tropical Atlantic. Weak dynamical constraints in the tropics require faithful representations of divergent flow and mesoscale features in model initial conditions (ICs) to optimize the skill of a forecast. The detailed physics in MM5, down to resolution of non-hydrostatic effects at cloud scale, make it ideal for simulation of the inner core evolution of tropical cyclones. However, detailed data are needed to adequately constrain the forecast. MM5 (and hence CHUMS) has a four dimensional data assimilation (FDDA) cycle based on a Newtonian relaxation (or "nudging"; Stauffer and Seaman 1994). This continuous assimilation technique is computationally efficient and allows asynoptic data to be handled readily. Use of the FDDA initialization allows known (or assigned) error characteristics of different data types to be taken into account by varying weightings of individual data types during the assimilation. Model initial and boundary conditions are drawn from global models, which typically ingest only a small subset of available satellite winds. Thus, satellite-derived winds and other special observations (such as flight data) provide independent data sources to CHUMS, with the potential to correct inadequately resolved tropical storm structure (or other phenomena) or location. Optimal distributions and weightings of these data and the need for a vortex bogus to obtain skillful forecasts throughout the life of Floyd will be examined.
Session 5A, Adaptive Observing Systems and Data Assimilation II (Parallel with Sessions 5B and J2)
Wednesday, 24 May 2000, 10:15 AM-12:00 PM
Previous paper Next paper