6.2 Track, Intensity, & Structure Changes at

Landfall – Forecasting Challenges

Working Group:

Philippe Caroff (France) Joe Courtney (Australia)

David Grant (Australia) Koji Kato (Japan)

M. Mohapatra (India) C.H. Qian (China)

Third International Workshop on Tropical Cyclone Landfall Process

Jeju, Republic of Korea 8-10 December 2014

S.T. Chan

Hong Kong Observatory

Jim Davidson

Formerly Bureau of Meteorology

Australia

AGENDA

• Track changes • Midget TCs (“All-or-nothing scenario”) • Rapid changes in TC motion • Timing of landfall • Biases in NWP forecasts

• Intensity changes • Uncertainty in TC intensity estimates • Rapid intensification & weakening • Land effects

• Structure changes • Limitation in TC intensity & structure forecasting • Lack of reliable analysis • TCs evolved from monsoon depressions/gyres • Eyewall replacement cycles • Shift of convection /wind maxima during landfall phase

• Summary and discussion

FORECASTING CHALLENGES

• Focus on practical aspects of forecasting illustrated with recent TC cases

• When landfall actually occurs, decision time already passed (warning, disaster preparation)

• Hours or even days prior to landfall the most critical period to determine landfall point, final intensity & structure at landfall

• Inevitably some overlap with discussion last week during IWTC-VIII

Track changes

• Midget TCs (“All-or-nothing scenario”)

• Rapid changes in TC motion

• Timing of landfall

• Biases in NWP forecasts

Forecast errors in Atlantic Basin reduced by 50+% over

past 10-20 years (Cangialosi & Franklin, 2014)

Error at T+24h generally < 100 km

TC track problem solved?

Performance of Track Forecasting

< 100 km RSMC Tokyo

100 km may only correspond to spread of severe

weather of storm!

Cyclone Tracy devastated Darwin in 1974 – the most

compact hurricane on record in Australia with gale wind

radius 48 km

A 100-km shift in 24 h can make a difference between

“close shave” and catastrophic direct hit!

Can we do any way better if Tracy was to threaten

Darwin today?

Midget TCs (“All-or-nothing”)

Still a challenging topic in particular for TCs making

sudden turn followed by acceleration

Errors in general higher by 20% for T+72 h for such TCs in

NIO basin

Rapid changes in TC motion

Recent case: TC Viyaru

in May 2013

Recurved over BoB &

accelerated to 40-50

km/h 12 hours prior to

landfall

Accurate landfall point

from models

Landfall point error by consensus forecast 63 km only

Yet landfall time error 10+ hours!

Impact of Viyaru grossly over-predicted by IMD

Large error in timing of landfall particularly undesirable

Rapid changes in TC motion

Forecast

Forecast based on

00 UTC on 14 May 2013

Forecast based on

00 UTC on 15 May 2013

Lead time to landfall: 56 h Lead time to landfall: 32 h

JMA GSM 137 / +10 63 / +12

NCEP GFS 289 / -4 169 / +4

ECMWF IFS 259 / +4 274 / +12

Multi-model

Consensus 63 / +10 63 / +13

Official 89 / +10 94 / +13

(Landfall point error/Landfall time error)

Timing of landfall

TC Yasi in Feb 2011

Aust. Cat. 5

(200+ km/h) at

landfall

City of Cardwell

hardest hit with

storm surge 5.33 m

Importance of accurate timing of landfall highlighted

Timing of landfall (Cont’d)

Had Yasi hit Cardwell a

few hours earlier at

astronomical high tide

Water level would have

been ~2 m higher!

• Systematic biases in track forecasts by models observed from time to time

• T. Nanmadol in 2011

• Systematic right biases by models, possibly due to under-estimation of strength of sub-tropical ridge

Biases in NWP forecast

ECMWF

T639

JMA

GFS

Similar for Kai-tak in Aug 2012

T+120h error 904 km vs. 455 km for 2012 average

Still remember Fengshen in 2008?

Biases in NWP forecast (Cont’d)

Intensity changes

• Uncertainty in TC intensity estimates

• Rapid intensification & weakening

• Land effects

NHC admitted that smaller errors in medium-term forecasts largely

attributed to lack of rapidly intensifying hurricanes; IMD also noted

no notable skill improvement from 2005-2011

Current skill level of intensity

forecasting

No net improvement for T+24h

Notable decrease in errors for T+72h & 96h

Unlike TC tracking, MSW of TC almost never measured

operationally in the lack of in-situ measurements

Intensity analysis remains an ongoing problem

Particularly problematic when same TC tracked by

agencies, e.g. in WN Pacific

Significant differences in intensity among various best-

track archives exist

Different wind averaging periods only accounted for part

of differences, operational analysis practices also

different among agenices

Uncertainty in TC intensity estimates

Recent example: SuperT Rammasun in Jul 2014

Claimed by CMA to be possibly strongest typhoon to

make landfall over China, with 1-hour drop of ~70 hPa to

899.2 hPa recorded during Rammasun’s passage over

Hainan

Uncertainty in TC intensity estimates

(Cont’d)

MSW (kt) as reported by different agencies at 06 UTC,

18 Jul:

Such a great difference even during landfall of TC!

Great confusion to members of public

Highly questionable whether accurate & robust intensity

analysis/forecasting method can ever be developed &

validated!

With more remote sensing methods emerging, estimates

from different agencies may diverge even more!

Uncertainty in TC intensity estimates

(Cont’d)

RSMC Tokyo JTWC HKO CMA

90 135 115 120

Forecasters still at the mercy of abrupt & unexpected

rapid change in intensity prior to landfall

Notable case dealt with by RSMC La Reunion last TC

season – TC Helen in Mar 2014

Extreme conditions predicted for Magdagascar by

forecasters

Rapid intensification/weakening

24 hours

55 kt @ 11 UTC, Mar 29 125 kt @ 11 UTC, Mar 30

Helen made landfall 28 hours later, but as a tropical

storm!

Drop of 85 kt in 24 hours to 40 kt, an unofficial record

for weakening rate over sea according to IBTrACS

Inner core completely dismantled in 5 hours

Impact of storm greatly over-estimated!

Rapid intensification/weakening

(Cont’d)

5 hours

Similar example from NIO –

TC Lehar in Nov 2013

Lehar weakened over the

sea from 75 kt to 25 kt in 18

hours – rainfall, wind and

storm surge over-estimated

Rapid intensification/weakening

(Cont’d)

Strong limitations in intensity prediction –> warning

strategy highly constrained by current state of science

Many people at stake, forecasters tend to overwarn

Also cases with TC decaying rather slowly after making

landfall

Especially so over NIO during the post-monsoon season

with still abundant moisture over land

Example: TC Phailin in Oct 2013 – Statistically-based

decay model predicted much faster decay than actual

Land effects

Decay of PHAILIN after landfall by IMD DECAY model

based on intensity at landfall time

115

59

34

25 22 21

115

75

4035

30 30

0

20

40

60

80

100

120

140

0 6 12 18 24 30 36

Lead Time (h)

De

ca

y In

ten

sit

y (

kt)

Predicted

Observed

Case of TC Laurence in Dec 2009

Storm forecast to skirt along NW coast of Australia and

intensify slightly to Cat. 2 storm

Yet Laurence did not directly feel the effects of land &

it intensitied into Cat. 5 before making landfall

Land effects smaller for larger storms?

Land effects (Cont’d)

Predicted Observed

Structure changes

• Limitation in TC intensity forecasting & lack of reliable analysis

• TCs evolved from monsoon depressions/gyres

• Eyewall replacement cycles

• Shift of convection /wind maxima during landfall phase

As important as intensity for assimilation to models and

inputs to impact assessment tools, e.g. storm surge

models

Commonly used parameters include RMW, wind radii for

various threshold (e.g. 34, 50 & 64 kt) in each quadrant,

radius & pressure of OMCI

While some/all of parameters being analysed by centres,

not many equipped to issue structure change forecast to

public

Structure of TCs

Intensity the first fundamental parameter to define

structure of TC, which could only be as accurate as

intensity of TC

Similar to intensity problem, detailed structure of TCs

over most basins not yet known in the lack of in-situ

measurements, e.g. routine aircraft surveillance

Even with surveillance flights, NHC currently does not

release verification results of forecast wind radii

considering lack of ground truths to facilitate reliable

verification

Limitation in TC intensity forecasting

Monsoon gyres – large circulation with strongest winds in

the outer circulation & light winds & lack of deep

convection near the centre

Common in western Pacific & may develop into TCs

Not a typical textbook TC & Dvorak analysis usually gives

way too low intensity estimates

Recent example: STS Nakri in July 2014

TCs evolved from monsoon

depressions/gyres

Nakri first named a TD by RSMC Tokyo on 28 Jul

Upgraded to STS on 31 Jul when passing Ryukyu Islands

55 kt wind reported at airport on Amami Island

JTWC subsequently issued warning at 06 UTC, 2 Aug

No reliable method to assess intensity/structure of such

systems for assessing their impacts and warning strategy

TCs evolved from monsoon

depressions/gyres (Cont’d)

13 UTC, 31 Jul 2014 00 UTC, 1 Aug 2014

• ERCs of TCs lead to rapid deviations of MSW and broadening of surface wind field

• Dvorak estimates may indicate contradictory trends

Eyewall replacement cycles (ERCs)

• Severe TC Ita underwent ERC off the coast of northern Queensland & weakened close to landfall

• Warnings issued by Brisbane TCWC over-estimated the intensity & attracted critics from public

• TC structure change at landfall due to frictional & topographic effects lead to distortion of wind radii -> shifting of convection & associated wind maxima during TC landfall

• A forecasting challenge as similarly-tracked TCs could produce totally different rainfall & wind damage patterns

• For TCs making landfall over eastern China, torrential rain usually occurs on the right flank of TC, but sometimes seen on the left

• Better understanding of TC-land interaction under different circumstances still required

Shifting of convection/wind maxima

during landfall phase

Summary & discussion

• RI/RW prior to landfall among the biggest challenges & reliable objective guidance generally lacking in many basins ->Effective guidance indicating at least chance of RI/RW or range of possible intensities much needed for risk assessment & warning strategy formulation

• No matter how accurate track guidance is, still need to cope with forecasts/warnings bearing different degrees of uncertainty -> Such uncertainties should be properly communicated

Summary & discussion

• Yet, communicating forecast uncertainty should NOT be overdone – sometimes counterproductive to disaster preparation efforts

• To get strongest response from public, just stick to the main message though at the risk of overwarning (re. “All-or-nothing scenario”)

Summary & discussion (Cont’d)

• Way of better communicating track uncertainty since developed

• Use of ECMWF EPS forecast by RSMC La Reunion beginning 2011-2012 TC season

• EPS forecast, when available, could better define uncertainty than climatologically-based cones of uncertainty

• BoM also practicing similar technique with reference to model consensus & prevailing synoptic pattern apart from EPS ensemble

Summary & discussion (Cont’d)

• Assessment of forecast uncertainty should extend to timing of landfall, intensity & structure of TC -> allow the public, government & private sectors to better assess the impact of TCs

• One step towards this goal by NHC:

Summary & discussion (Cont’d)

RECOMMENDATIONS

1. Full investigation to cause of rapid changes in motion of

TCs; reasons for persistent biases developed in NWP to

be documented and removed if possible

2. Closer communication & coordination among warning

agencies needed to narrow operational differences in TC

intensity estimates

3. Research priority given to development of operational

guidance tools on RI/RW, or at least range of possible

intensities achievably by storms during landfall phase

4. Further research efforts put on intensity/structure

change due to land/topographic/lake effects

RECOMMENDATIONS (CONT’D)

5. Development of reliable methods for effective analysis

of TC intensity & structure; international/regional

efforts to collect in-situ TC observations highly

encouraged

6. Better understanding on evolution of structure of MGs,

especially point of TC genesis required

7. Further research on TC intensity change associated with

ERCs encouraged

8. Met. Services to improve on techniques to communicate

forecast/warning uncertainties. Apart from TC positions,

emphasis be also put on timing of landfall, intensity &

structure of TC

End