Tropical Cyclogenesis Forecasting Skill of GFS during 2002 and 2003 Hurricane Seasons

1
Tropical Cyclogenesis Forecasting Skill of GFS
during 2002 and 2003 Hurricane Seasons
Aaron Pratt Department of Meteorology The
Pennsylvania State University
2
Outline for Talk

• Background information
• Reasons for Study
• Methodology
• Spatial aspects of genesis
• In-depth analysis of GFS genesis forecasts
• Conclusions

3
Background

• Cyclogenesis Theories
• Conditional Instability of the Second Kind
  (CISK).
• Wind-Induced Surface Heat Exchange (WISHE).
• Require preexisting disturbance in both cases.
• Neither theory accounts for initial vortex
  formation.

4
Background

• 6 large-scale factors for tropical cyclogenesis
• 3 thermodynamic (SST gt 26.5 C, conditional
  instability, high mid-tropospheric moisture).
• 3 dynamic (large low-level vorticity, weak
  vertical wind shear, nonzero Coriolis parameter).
• Dynamic factors affect daily genesis likelihood
  thermodynamic factors affect seasonal genesis.
• Above conditions can occur simultaneously without
  genesis.

5
(No Transcript)
6
Background

• Western Pacific is the worlds most prolific TC
  producer many studies of cyclogenesis here.
• Monsoon trough and associated genesis.
• Atmospheric Oscillations (Madden-Julian
  Oscillation, mixed Rossby-Gravity wave).
• Eddy Angular Momentum flux convergence.
• Mesoscale interactions in convectively active
  regions.

7
Reasons For This Study Genesis challenges

• Not as much research in Atlantic tropical
  cyclogenesis as the Pacific.
• Atlantic described as most complicated basin for
  genesis (Gray 1968).
• Various scales of interaction between large-scale
  precursors and mesoscale processes.

8
Reasons for this study Modeling challenges

• Studies of genesis forecasting by operational
  models have focused on NOGAPS, UKMET, and ECMWF.
• Very few cyclogenesis studies of Global Forecast
  System (GFS), NWS primary global model.
• Also, many cyclogenesis studies use time of TD
  classification as genesis time.

9
Goals for This Study

• Desire to examine, as objectively as possible,
  skill of GFS cyclogenesis forecasts in Atlantic
  Basin.
• Compare model climatology to observed climatology
  for cyclogenesis.
• Compare genesis forecasts with GFS analyses of
  background environment
• Assessment of possible effects of Saharan Air
  Layer (SAL) on GFS genesis forecasts.

10
Jordan (1958) Caribbean soundings
Dunion Velden (2004)
11
Methodology

• Model domain size 0-30N to capture purely
  tropical cyclogenesis.
• Examined warm core, non-frontal lows only.
• Location and time noted for each genesis
  forecast.
• Comparison of genesis forecasts to observed
  tropical cyclogenesis.
• Success evaluated individually for each forecast.
  
• Allowed /- 24 hours difference between observed
  and forecasted genesis.
• Phase speed of easterly wave (8 ms-1) determined
  spatial window of success.

12
Unsuccessful (F2 black) vs. successful (S red)
forecasting criteria
13
Cyclogenesis is defined here as the formation of
any tropical low, regardless of further
intensification into TD, TS, or hurricane.
14
Case 1 Successful Forecast

• TS Bertha
• Genesis 00 UTC August 3, 2002
• Domain15-35N, 70-100W

15
TPC Surface analysis at 00 UTC Aug. 3, 2002
IR Satellite imagery at 00 UTC Aug. 2, 2002
16
96 Hour GFS Forecast from 00 UTC July 30, 2002.
Forecast valid 00 UTC August 3, 2002
48 Hour GFS Forecast from 00 UTC August 1, 2002.
Forecast valid 00 UTC August 3, 2002
17
Case 2 Failure 1 (F1) Forecast

• Hurricane Isidore
• Genesis 00 UTC September 10, 2002
• Domain 0-20N, 5-35W

18
TPC Surface analysis at 00 UTC Sept. 10, 2002
IR Satellite imagery at 00 UTC Sept. 10, 2002
19
96 Hour GFS Forecast from 00 UTC Sept. 6, 2002.
Forecast valid 00 UTC Sept 10, 2002
48 Hour GFS Forecast from 00 UTC Sept. 8, 2002.
Forecast valid 00 UTC Sept 10, 2002
20
Case 3 Failure 2 (F2) Forecasts

• September 16-17, 2002 Forecasted Genesis Event
• Domain 0-20N, 5-35W

21
TPC Surface analysis at 00 UTC Sept. 17, 2002
IR Satellite imagery at 00 UTC Sept. 17, 2002
22
60 hour forecast from 12 UTC September 14.
Forecast valid 00 UTC September 17
36 hour forecast from 12 UTC September 15.
Forecast valid 00 UTC September 17
23
Spatial Aspects of Genesis

• Part I 2002 season

24
Monthly average (1951-2001) of observed TC
occurrence
2002 GFS forecasts
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
Spatial Aspects of Genesis

• Part II 2003 Season

29
Monthly average (1951-2001) of observed TC
occurrence
2003 GFS forecasts
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
Spatial Aspects of Genesis

• Part III Compare/Contrast 2002 and 2003 season

34
(No Transcript)
35
Anomalous Vertical Shear of Zonal Wind and
Anomalous 200 hPa Vector Wind for Aug.-Oct. 2002
Anomalous Vertical Shear of Zonal Wind and
Anomalous 200 hPa Vector Wind for Aug.-Oct. 2003
36
(No Transcript)
37
In-depth analysis of GFS forecasts

• Eleven weeks examined in which only F2 forecasts
  occurred in 2002 and 2003.
• Genesis location compared to weekly-averaged
  analyses of shear and vorticity. SST also
  examined
• Location of SAL outbreaks noted as well
• Idea is to see if forecasts are in regions where
  GFS analyses are favorable/unfavorable.

38
(No Transcript)
39
In-depth analysis of GFS Genesis Forecasts Part I
40
(No Transcript)
41
(No Transcript)
42
(No Transcript)
43
00 UTC July 17 IR
00 UTC July 18 IR
00 UTC July 19 IR
00 UTC July 20 IR
00 UTC July 21 IR
00 UTC July 22 IR
00 UTC July 23 IR
July 17-23 700 hPa Relative humidity average
44
In-depth analysis of GFS Genesis Forecasts Part
II
45
(No Transcript)
46
(No Transcript)
47
(No Transcript)
48
00 UTC Sept 1 IR
00 UTC Sept 2 IR
00 UTC Sept 3 IR
00 UTC Sept 4 IR
00 UTC Sept 5 IR
00 UTC Sept 6 IR
00 UTC Sept 7 IR
Sept 1-7 700 hPa Relative humidity average
49
Conclusions

• GFS correctly models seasonal climatology
• Low number of forecasts early/late season
• Model peak close to observed seasonal peak
• Highest success rate for GFS occurs in June of
  both years
• Might be due to higher success overall by GFS in
  western Atlantic.
• Large number of F2 forecasting outcomes by GFS (
  S percentage generally decreases from June peak)

50
Conclusions

• F2 forecasts by GFS occurred mainly in regions of
  decreased background vorticity at 850 or 500 hPa.
• Excessive moisture convergence possible culprit
  in large number of genesis forecasts by GFS
• SAL influence a definite possibility on F2
  outcomes, due to its unfavorable environment
• 3 of the 11 weeks had negative SAL effects in
  regions of forecasted genesis.

51
Future Work

• More in-depth study of SAL effects on forecasted
  tropical cyclone genesis.
• Examination of averaged forecast fields and
  comparison to averaged analyzed fields and
  forecasted genesis.
• Investigation of F1 systems (why did GFS not
  anticipate their development).

52
(No Transcript)
53
Methodology

• 1339 Cyclone Phase Space (CPS) analyses of GFS
  tropical cyclogenesis forecasts in 2002 and 2003.
• CPS critical in determining forecasted core
  structure (warm vs. cold core limits inclusion
  of nontropical low forecasts in study).

54
Cyclone Phase Space - B Asymmetry paramter
B 0
B 0
B gt 0
(h1 for N. Hemisphere, -1 for S.Hemisphere)
55
Cyclone Phase Space -VT Thermal Parameter
e.g. 700hPa height
500km
ZMAX
?Z ZMAX-ZMIN isobaric height difference
within 500km radius
Proportional to geostrophic wind (Vg)
magnitude ?Z d f Vg / g (where ddistance
between height extrema, fcoriolis, ggravity)
Vertical profile of ZMAX-ZMIN is proportional to
thermal wind (VT) if d is constant
ZMIN
-VT lt 0 Cold-core, -VT gt 0 Warm-core
56
125
Mature ET Cyclones
Asymmetric/ Frontal
Transitioning, Hybrids, Sub-tropicals
Developing ET Cyclones
B
Occluded ET Cyclones
Symmetric/ non-frontal
0
Tropical Cyclones
Cold Core
Warm Core
0
300
-600
-VTL