Verification of model wind structure and rainfall forecasts for 2008 Atlantic storms Tim Marchok NOAA / GFDL Collaborators: Rob Rogers (NOAA / AOML / HRD) Bob Tuleya (NCEP/EMC & Old Dominion Univ.) Mark Powell (NOAA / AOML / HRD) HFIP Hurricane Diagnostics and Verification Workshop NHC, Miami, FL 04 May 2009
33
Embed
Verification of model wind structure and rainfall forecasts for 2008 Atlantic storms
Verification of model wind structure and rainfall forecasts for 2008 Atlantic storms. HFIP Hurricane Diagnostics and Verification Workshop NHC, Miami, FL 04 May 2009. Tim Marchok NOAA / GFDL. Collaborators: Rob Rogers (NOAA / AOML / HRD) Bob Tuleya (NCEP/EMC & Old Dominion Univ.) - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Verification of model wind structure and rainfall forecasts for 2008 Atlantic storms
Tim MarchokNOAA / GFDL
Collaborators: Rob Rogers (NOAA / AOML / HRD)
Bob Tuleya (NCEP/EMC & Old Dominion Univ.)
Mark Powell (NOAA / AOML / HRD)
HFIP Hurricane Diagnostics and
Verification Workshop
NHC, Miami, FL
04 May 2009
Motivation
• Inland flooding from TC rainfall accounts for a significant percentage of deaths from U.S. landfalling TCs.
• Storm size / structure has a major impact on the amount of damage at landfall.
…both aspects have only recently begun to receive attention in terms of model evaluation
Outline
• Rainfall validation & techniques
• Application of TC QPF validation techniques to 2008 U.S. landfalling storms
• Development of model wind structure validation techniques
• Application of wind structure validation techniques to 2008 Atlantic TCs
• Rainfall patterns
• Rainfall mean & volume
• Extreme amounts
Rainfall validation: What to focus on?
The TC track: An anchor for QPF validation
GFDL
NAMGFSR-CLIPER
HWRFStage IV
Example: Hurricane Gustav 72-h total rainfall
2008 Landfalling Storms
Rainfall Patterns
20052008Equitable Threat Scores
Pattern Correlations: GFS (r = 0.78)
HWRF (r = 0.60)
GFDL (r = 0.53)
NAM (r = 0.52)
R-CLIPER (r = 0.51)
(2008)
Mean Rainfall
Rainfall volume
2008 2005
Comparison of rain volume bias by model
2008 Landfalling Storms
50% level
Difference between obs
rain value and fcst rain value at 50% level determines
index
Track-relative rain flux analyses
Rain volume: Rain flux in select bands
0–100 km
300-400 km
GFDL, HWRF GFS, NAM
Skill Indices: Pattern Matching
Pattern Matching (2005)
0.44
0.48
0.71
0.59
0 0.2 0.4 0.6 0.8 1
RCLIP
NAM
GFS
GFDL
2008 2005
• GFS performs the best in both samples
• All models have skill relative to R-CLIPER
Pattern Matching (2008)
0.40
0.48
0.78
0.52
0.50
0 0.2 0.4 0.6 0.8 1
RCLIP
NAM
GFS
HWRF
GFDL
Skill Indices: Mean / Volume
Volume (2005)
0.72
0.75
0.68
0.72
0 0.2 0.4 0.6 0.8 1
RCLIP
NAM
GFS
GFDL
2008 2005
Mean / Volume (2008)
0.36
0.80
0.79
0.32
0.70
0 0.2 0.4 0.6 0.8 1
RCLIP
NAM
GFS
HWRF
GFDL
Skill Indices: Extreme Amounts
Extreme Amounts (2005)
0.62
0.76
0.77
0.57
0 0.2 0.4 0.6 0.8 1
RCLIP
NAM
GFS
GFDL
2008 2005
• GFDL overforecasts the large amounts, HWRF underforecasts them, resulting in comparable indices