HR: 0800h
AN: A21C-0635    [Abstracts]
TI: Cycling Variational Assimilation of Remotely Sensed Observations for Prediction of Hurricane Katrina
AU: * Chen, S
EM: shachen@ucdavis.edu
AF: UC Davis, One Shields Avenue, Davis, CA 95616,
AU: Lim, E
EM: lim@ucar.edu>
AF: MMM/NCAR, 3450 Mitchell Lane, Boulder, CO 80301,
AU: Lee, W
EM: wenchau@ucar.edu
AF: MMM/NCAR, 3450 Mitchell Lane, Boulder, CO 80301,
AU: Davis, C
EM: cdavis@ucar.edu
AF: MMM/NCAR, 3450 Mitchell Lane, Boulder, CO 80301,
AU: Bell, M
EM: mbell@ucar.edu
AF: MMM/NCAR, 3450 Mitchell Lane, Boulder, CO 80301,
AU: Xio, Q
EM: hsiao@ucar.edu
AF: MMM/NCAR, 3450 Mitchell Lane, Boulder, CO 80301,
AU: Lin, H
EM: hclin@ucar.edu
AF: MMM/NCAR, 3450 Mitchell Lane, Boulder, CO 80301,
AU: Holland, G
EM: gholland@ucar.edu
AF: MMM/NCAR, 3450 Mitchell Lane, Boulder, CO 80301,
AB: The impact of assimilating observations from conventional instruments, Radar, and satellites on Hurricane Katrina (2005) was assessed using high-resolution model simulations. Four nested domains were configured with grid spacings of 54 km, 18 km, 4.5 km, and 1.5 km for domains 1 to 4, respectively. Three sets of experiments, EXP1-3, with various initial times and data cycling periods were conducted. The data cycling periods were 1800 UTC August 25 to 0000 UTC August 26 (6 h), 0000 UTC to 0600 UTC August 26 (6 h), and 1800 UTC August 25 to 0600 UTC August 26 (12 h) for EXP1-3, respectively. The first two sets (i.e. EXP1 and EXP2) were to evaluate the impact of assimilating different observations on Katrina simulations, while the last one (EXP3) was to examine the influence of different assimilation periods for radar data. Thirty-six hour model integrations were then performed after the data cycling on all four domains. The results from EXP1 shows that the assimilation of radar and conventional data had a positive impact on simulated storm¡¦s intensities during the first 24 hours, while the influence of satellite data was also positive, though less significant, and the influence was able to extend the whole 36-h simulation since the coverage of satellite data was much larger than that of radar. However, none of EXP1 experiments with the assimilation of observations had reduced track errors, which increased with time and reached about 200-250 km after a 36-h integration. The forecasts from EXP2 show great improvement in simulated storm¡¦s intensity and track after the use of observations, in particular for the first 24 h of the forecasts. The track error was close to 50 km during the whole simulation period. The results from EXP3, which started the data cycling at 1800 UTC August 25 as in EXP1, confirmed that the observations from 0000 UTC to 0600 UTC August 26 used in this study played a key role in improving simulated tracks. Moreover, the results further concluded that the impact on the simulated track was contributed from GTS data and/or satellite data. Doppler radar data assimilation mainly contributed to the improvement of the hurricane intensity forecast. Finally, all three sets of experiments indicated that the assimilation of radar data in an interval of every three hours for a 6-h time period is an optimal setting.
DE: 0312 Air/sea constituent fluxes (3339, 4504)
SC: Atmospheric Sciences [A]
MN: 2007 Fall Meeting