Implementation of the UWCIMSS Advanced

1
Implementation of the UW-CIMSS Advanced Objective
Dvorak Technique (AODT) into TPC (NOAA USWRP
Joint Hurricane Testbed Project) Jim Kossin,
Chris Velden, and Tim Olander University of
Wisconsin Cooperative Institute for
Meteorological Satellite Studies Madison,
WI Interdepartmental Hurricane Conference,
March 2004
2
JHT Project Primary Objective Implement the
latest version of the UW-CIMSS AODT onto the NOAA
Computing Development Branch N-AWIPS architecture
for operational use at TPC.
3

• Brief Review of the UW-CIMSS Advanced ODT (AODT)
• Originally designed to imitate the Dvorak
  technique while removing operator subjectivity.
• Employs geostationary satellite IR imagery to
  estimate current TC intensity.
• Computer-based algorithm that can be executed in
  a fully automated mode, or with manual over-ride
  capability.
• Includes an algorithm designed to locate the TC
  center.
• Applicable to all stages of the TC lifecycle.
• Thorough statistical analysis indicates the AODT
  is competitive with the operational (subjective)
  Dvorak technique.
• The AODT predecessor, the ODT, has been running
  experimentally at various forecasting centers for
  a number of years.

4

• Latest version of the AODT
• Time averaging scheme reduced from 12h to 6h to
  better allow for
• identification of rapid intensity change
  events.
• Modified land interaction rules to better handle
  transient land
• interaction (e.g. TC traversing large
  islands).
• Inclusion/modification of Dvorak rules 8/9 to
  better estimate
• initial and final intensities during TC
  lifetime.
• Implementation of new automated TC center
  positioning scheme.
• Latitude-dependent MSLP bias removed.

5
Performance of the latest version of the AODT
(estimated MSLP versus recon-measured
MSLP) rmse aae bias Manually fixed
images 9.50 7.48 0.09 (mb) Auto-fixed
images 10.04 7.60 1.70 OpCenterAvg 10.65 8.09
0.22 dependent testing based on the AODT
training sample (n 1630) distribution
inherently skewed toward stronger TC
cases. rmse aae bias Auto-fixed
images 10.10 8.19 -1.26 (mb) OpCenterAvg 9.29
6.79 0.95 independent testing (n 2221) on a
more natural distribution of TC cases.
6

• Implementation onto N-AWIPS
• Restructure the AODT algorithm for congruence
  with the NOAA CDB N-AWIPS architecture. This
  requires
• Re-creation of the AODT algorithm as a library
  function.
• Creation of an Application Program Interface
  (API). This requires creation of new library
  functions enabling communication between the AODT
  and the N-AWIPS NMAP application.
• Timeline
• Delivery of API variable passing library
  functions - Feb 2004
• Delivery of API data passing library functions -
  Mar 2004
• Test run of partial AODT library function on
  NAWIPS - Apr 2004
• Comparison of McIDAS-AODT with NAWIPS-AODT - May
  2004
• Delivery of complete NAWIPS-AODT to TPC in NMAP
  upgrades
• - June 2004

7

• Summary
• A fully automated version of the UW-CIMSS AODT
  is expected to be functional at TPC in time for
  the 2004 Atlantic Hurricane season.
• The 2004 season will be used to evaluate the
  NAWIPS-AODT at TPC.
• The TPC evaluation will be used to examine and
  implement post-season modifications.
• UW-CIMSS will continue to improve the AODT and
  provide upgrades to CDC for implementation into
  the NAWIPS version. The recent expansion of the
  AODT training sample to include more weak cases
  will provide a means of improving the overall
  performance characteristics.