initialization, prediction and diagnosis of the rapid intensification of tropical cyclones using

The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology

Initialization, Prediction and Diagnosis ofthe Rapid Intensification of Tropical Cyclones using

the Australian Community Climate and Earth System Simulator, ACCESS

Michael Reeder, Noel Davidson, Jeff Kepert, Craig Bishop, Peter Steinle and Kevin Tory

with Yi Xiao, Harry Weber, Yimin Ma, Hongyan Zhu, Xingbao Wang, Mai Nguyen,

Lawrie Rikus, Richard Dare, Ying Jun ChenAnd

Roger Smith and Michael Montgomery (Honorary Members)

Special thanks to WEP and ESM Programs, and UKMO

Weather and Environmental Prediction and

Environmental System Modelling Groups

CAWCR, Centre for Australian Weather and Climate Research

A Partnership between CSIRO and the Bureau of Meteorology

Acknowledgments: Kamal Puri, Gary Dietachmayer

TC behaviour and forecast issues:

• Track,

• Genesis,

• Intensification/RI/Decay,

• Structure Change (size, etc),

• ET

• Landfall!!!

Points of Origin

Points of Final Decay

Points with Min. CP

Tropical Cyclone Characteristics in the Australian Region

(Dare and Davidson, 2004, MWR)


Scope of Talk

• Operational ACCESS-TC

• ACCESS-TC System Configuration VS, 4DVAR Initialization, Verification (track, intensity, structure)

• Related, Diagnostic Projects (ShudderShudder, Points of collaboration) (Testing new params, new data sources, even mechanisms/processes)

• Future Plans

ACCESS-TC Verification:NMOC Real-time Forecasts 2011 WNP Region, 10 TCs:

Number of ForecastsMean Track Error, Mean ABS Central Pressure Error, (B-corrected), A-TC and Persistence

Not All Good News …

Track Forecasts from available operational systems for Heidi and Iggy (A-TC, EC, UK, JMA, GFS, NGP, GFDN,……

Deficiencies in (a) LSE, and/or (b) Vortex Structure ???

ACCESS-TC vs ECMWF for TC IGGY from base times 20120126/12Z and 20120127/00Z

Left Panels: Observed and forecast tracks and central pressures from ACCESS-TC

Centre Panels: 72-hour forecasts of MSLP from ACCESS-TC;

Right Panels: 72-hour forecasts of MSLP from ECMWF


ACCESS-TC for Operations and Research

1. Resolution:

0.110X50L, re-locatable grid, with TC near centre of domain, option for higher-resolution forecasts.

2. Vortex Specification: (a) Structure based on observed location, central pressure and size (tuned and validated using ~6000

dropsonde observations from the Atlantic)(b) Only synthetic MSLP obs used in the 4DVAR to (a) relocate the storm to observed location, (b)

define the inner-core circulation, and (c) impose steering flow asymmetries consistent with the past motion.

3. Initialization using 4DVAR Assimilation: 5 cycles of 4DVAR over 24 hours. Uses all standard obs data, plus synthetic MSLP obs (no upper air

synthetic obs).

4DVAR then: (a) Defines the horizontal structure of the inner-core at the observed location, (CP, VMAX, RMW, R34) (b) Builds the vertical structure from MSLP obs, (c) Constructs the secondary circulation, (d) Creates a balanced TC circulation at the observed location, with correct (?) structure and intensity.(e) Creates a structure which is responsive to environmental wind shear without imposing constraints

on the vertical-stacking or tilt of the circulation. (important for vortex dynamics and cloud asymmetries)

4. Forecast Model: UKMO Unified Model from ACCESS.


Verification of large scale forecasts

MSLP RMSE: Global forecasts over the Australian Region

Improved Prediction of the LSE of storms,

compared to previous Australian Global System

The Centre for Australian Weather and Climate ResearchA partnership between CSIRO and the Bureau of Meteorology

OBS Network

Without Vortex Specification : Initial Position/Intensity Errors for TC Anthony were ~ 230km and 5hPa

With Vortex Specification: Initial Position/Intensity Errors reduced to 40km and 0hPaVS: blue MSLP obs in upper left panel: dense enough to define Vmax at RMW, extensive enough to merge with LSE.


For Anthony at Landfall:

Obsvd and fcast track and intensity without and with VS

500 hPa Initial Condition without and with VS (synthetic MSLP obs only)

Note construction of 3-D structure

4DVAR defines depth and tilt, important for evolution of vortex

Vortex Specification (Weber, 2011)

Figure 2: Tangential wind v(r) in m s-1 as a function of radius in km of Hurricane Fran on September 29, 1996 (top) and Hurricane Floyd on September 19, 1999 (bottom). Thick lines represent the average v(r) of all flight passes and the AVSM output v(r) (smoother curve). The thin lines define an envelope given by the minimum and maximum v(r) of all flight passes at each radial grid point. The input parameters of AVSM are operational estimates of roci and vm – c in (e), (f).

Validation of Vortex Structure: Use EXBT data sets for the NA and NP to validate TC structures obtained from the Vortex Specification (RMW, R34). (CLOK: Charlie Lok)


Validation of Vortex Structure. I: Cloud Fields (Rikus, 20XX)

Actual and Synthetic Cloud Imagery

Yasi at t = 0 and t = 46 hours from base time, 12Z, 20110131

• 4DVAR initializes the ascent and moisture fields.

• Model maintains cloud fields during the forecast.


Validation of Vortex Structure.

II: Cloud Bands and Convective Asymmetries

85GHz Imagery (left panels) and ACCESS-TC 500 hPa vertical motion field at t = 6 (initialized with 4DVAR) and t = 55 hours for Yasi from base time 00Z, 20110131

• Note regions of observed active inner rainbands and eyewall convection, and corresponding forecast regions of strong and weak ascent.

• Based on use of synthetic MSLP obs and 4DVAR, structures are consistent from even the early hours of the forecast.

• Rainfall in TCs (Ying Jun Chen)

Preliminary Validation of Vortex Structure. III: Intensity and Windfields(Verification of R34)

(Y. Ma)

Critical for Storm Surge and Rainfall

For Yasi from base time 00Z, 20110131:

Time series of forecast

(a) Central Pressure,

(b) Maximum Wind,

(c) Radius of Maximum Wind,

(d) Radius of 64, 50 and 34 knot winds.

Symbols indicate estimated values, where available

Encouraging preliminary verification

***** What defines size and the RMW? <<<<<


Illustrative Example: TC YASI

Forecast and Observed

Tracks and Intensities from

ACCESS-TC at 4km resolution

Tropical Cyclone Projects

Verification and Post-analysis of Operational ACCESS-TC

Enhancements to the Boundary Layer Parameterization for ACCESS-TC.

Secondary Eyewall Formation and Eyewall Replacement Cycles in Tropical Cyclone Simulations

Genesis Applications: OWZ Diagnostics and ACCESS-TC

Downstream Development during the Extratropical Transition of Tropical Cyclones: Observational Evidence and Influence on Storm Structure.

Sensitivity of Prediction of Intensity and Vortex Structure to Initial Vortex Structure

Rainfall in TCs

Inner-core Structure Change during Rapid Intensification

Amplifying Planetary Rossby Waves and Extreme Rain Events in Current and Future Climates


ACCESS-TC – Improving air-sea exchange parameterisation, plus inclusion of sea spray processesYimin Ma and Colleagues

- Realistic physical representation of air-sea exchange in high wind conditions in TCs.

- Validate with CBLAST Data• Small changes in track forecast • Large improvements in intensity forecast • Small changes in outer structure

Structure prediction for YASI (2011). Base time 20110131/00Z

Statistics for track and intensity biases

(ME, ME ABS: Mean Error, Mean Absolute Error)


BL Parameterisation in TCs: Jeff Kepert

• Previous work has shown a substantial sensitivity to choice of parameterisation (Braun and Tao 2000, Smith and Thomsen 2010).

• Why are they different?• Which scheme is the most suitable?

• Method:• Use diagnostic 3-d model of TC BL (Kepert and Wang 2001) to make

interpretation easier – all simulations are the same above the BL.

• Implement four simplified parameterisations in the model, representative of those used in TC modelling.

• Compare and work out why.


Comparing the schemes

Scheme Near-surface log layer?

Inflow? Supergrad-ient flow?

Recommend-ed for use?

MM5 Bulk and Blackadar

No Strongest (exceeds obs)

Strongest (exceeds obs)

NO!

Louis Yes Moderate Moderate Yes

Higher-order (TKE scheme)

Yes Moderate Moderate Yes

(but expensive)

K-profile

(MRF, YSU)

Yes Weaker, depending on BL depth

Weaker, depending on BL depth

Yes, provided BL depth is ok (check!!).

Secondary Eyewall Formation and Eyewall Replacement Cycles in Tropical Cyclone Simulations Xingbao Wang and Colleagues

Tangential WindRadar Reflectivity(dBZ)

Idealized Initial Vortex in an Acquiescent Environment

Nested down to Resolution of 2/3 km.

East-west Diameter-time, Hovmoller diagram of Radar Reflectivity and Tangential Wind

Radar return from 63 to 86 h

Note SEF and start of ERC in bottom 8 panels

Hypothesis: As a dynamic response to an UnBalanced Force in the boundary layer (sum of pressures gradient, centrifugal, Coriolis, friction forces, etc…), a secondary maximum convergence zone (SMCZ) in the radial flow is generated in the boundary layer at a radius of about double the RMW. In the moist, conditionally-unstable atmosphere, the vertical updraft induced by the SMCZ triggers moist convection, which results in the Secondary Eyewall Formation.

OWZ Diagnostics for Genesis

Vertically-aligned, moist regions with curvature vorticity in low shear

Kevin Tory and Colleagues

Application of ACCESS-TC to Genesis Forecasting:

72, 60, 48, 36 hour forecasts verifying at 00UTC, 20111226: ~ Genesis time for Grant.

Downstream Development during the Extratropical Transition of Tropical Cyclones:

Lili Liu, Noel Davidson and Hongyan Zhu

Time-longitude series of Stream Function Anomaly (deviation from zonal mean) at 45N on 250hPa level for (a) Hurricane Michael, (b) Hurricane Wilma, (c) Hurricane Maria, and (d) Hurricane Rita (non-ET). The arrow is the propagation direction of trough/ridge wave train and the black dot is the position of the hurricane around ET time.

ET is often associated with Downstream Development Events

Capture and ET of Hurricane Maria

Fig.10: Dry, No-Initial-Vortex Simulation of MSLP for Maria. (a) Base time 00UTC 3 Sep 2005. (b) , (c), (d) are 24-, 48- and 72- hour MSLP simulations.

The black dot is the position of Hurricane Maria at the valid simulation time.

DD Dynamics can establish the large scale environment and capturing trough for ET


Structure and Structure Change critical for Rainfall and Storm Surge => Need for Mesoscale DA and Correct Initial Vortex Structure.

LARGE NATURAL VARIABILITY IN Tropical Cyclone STRUCTURE:(VMAX, CP, RMW, R34, ROCI)What determines the variability?What determines the RMW?Is storm structure important for the evolution of the storm?Correct prediction of CP and Vmax (intensity) does not imply correct prediction of structure.

*** Visualize the differences in rainfall and storm surge associated with different structures.

LARGE VARIABILITY IN Tropical Cyclone STRUCTURE (Ma and Davidson, 2012)

Initialized and 48-hour forecasts of the radial profiles of tangential wind from the 7 synthetic structures at t = 0 (top panels) and t = 48 hours (bottom panels) for Bonnie, Ivan and Katrina from base times 12UTC 23 August 1998, 00UTC, 12 September 2005 and 00UTC, 27 August 2005. Plotted symbols indicate estimated values of Vmax, R64, R50 and R34. Units are m/s for wind and kms for radius.

example

US14111; 7.45 151.83; atoll site

0

100

200

300

400

500

600

2001

0802

2002

0228

2002

0628

2002

0701

2002

0708

2002

1207

2003

0413

2003

0521

2003

0814

2003

1122

2004

0404

2004

0626

2004

1019

2004

1212

2005

0312

2006

1216

date

rain

rat

e (m

m/2

4h

r)

atoll rain TRMM 3B42

correl=0.906

US14585; 9.57 138.12; coastal with orographic influence

0

20

40

60

80

100

120

140

160

1998

1015

1998

1209

2000

0819

2001

0702

2002

0628

2002

1122

2003

0717

2003

0715

2003

1126

2004

0409

2004

0519

2004

0614

2004

0819

2004

1214

2005

0602

2005

0905

2006

0630

2006

0703

2006

0930

2006

1127

2007

0517

2007

0710

2008

0924

2009

1016

date

rain

rat

e (m

m/2

4hr

)

pacrain coastal with orographic influence TRMM 3b42

correl=0.749

US14111Gauge:749mmTRMM: 563mm

US14419Gauge: 256mmTRMM: 423mm

US14590Gauge:223mmTRMM: 204mm

Rainfall in TCs : Ying Jun Chen, Kevin Walsh, Beth Ebert, Noel Davidson

Analyse (TRMM, Atoll rain, rain gauge obs),

Verify,

Predict

TC Rainfall

23

25

27

29

31

33

270 272 274 276 278880

900

920

940

960

980

1000

0 6 12 18 24 30 36 42 48 54 60 66 72

FCST hour

Pm

in (

hp

a)

-2

-1

0

1

2

3

4

5

6

0 6 12 18 24 30 36 42 48 54 60 66

dVm SDM

Mai NGUYEN: Rapid Intensification: Inner-Core Processes:

Internal Structure Change during RI (Vacillation Cycles, QJRMS, 2011)

Left panel: Analysis of 24-hour rainfall accumulations for 29 January 1990. Centre and right panels: 500 hPa wind valid 24 and 29 January 1990, respectively. X marks the approximate location of Tropical Cyclone Tina at the analysis times, as it moves to the southeast and is captured and transitions into a midlatitude system.

Figure 1: (a) Number of extreme rain events by month, and (b) percentage of events occurring within each latitude zone, by month.

Amplifying Planetary Rossby Waves and Extreme Rain Events

in Current and Future Climates

ACCESS-TC: Future Plans

• Upgrades to APS1(more satellite data, higher resolution, improved physics, ….) • NWP and basic research applications from special experimental data sets: TPARC/TCS08, PREDICT: Genesis and Rapid Intensification (NOPP/ONR)

• Specification, Prediction and Validation of TC Structure (CP, Vmax, RMW, R34, ROCI): Critical for prediction of track, intensity, structure, storm surge and rainfall • Experiments with High Resolution Initialization and Prediction; Experiments with Ensemble Prediction; Experiments with Revised and New Physics; Diagnostics for TC boundary layer and moist processes • Enhancements with 4DVAR (inner and outer loops) (NOPP/ONR) Impact of extra observation types

• Rainfall in TCs (Ying Jun Chen, Walsh, Ebert, Davidson)

• Influence of Amplifying Rossby Waves on TC structure and intensity (NOPP/ONR)

• Inner-core Dynamics (eg, What defines RMW and R34? Mai Nguyen) (NOPP/ONR)

• Challenge: Initialize CAT 3 - 5 storms without the use of reconnaissance data or vortex specification?

• Happy to collaborate on and/or provide data for (i) testing assimilation of new obs data, (ii) testing new parameterisations, (iii) assessing mechanisms.

initialization, prediction and diagnosis of the rapid intensification of tropical cyclones using

Documents