NCEP/EMC HWRF Modeling System Current Status … 6h fcst Modified vortex initialization procedure: Improved mass/wind balance & storm size correction, with a focus on reducing initial

1

NCEP/EMC HWRF Modeling System

Current Status and Prioritized Stream

1.0 needs for FY11

Vijay Tallapragada

HWRF Team Lead, NCEP/EMC

HFIP Annual Review Meeting, Miami, FL

November 9, 2010

2

The HWRF Team at EMC

• Vijay Tallapragada (Team Lead)– Model development, T&E, operational support,

verification and diagnostics

• Robert Tuleya*

– Physics and Product Development

• Qingfu Liu

– Vortex initialization and Physics

• Young Kwon– Physics and model development

• Zhan Zhang– Ensembles, pre-implementation testing

• Sam Trahan– Code management, product automation, software

support, pre-implementation testing

• Eric Aligo– Microphysics, radiation, model diagnostics

• Janna O’Connor– Graphics, website maintenance, organization of

weekly meetings

• Bill Lapenta and Steve Lord

– Management and Support

• MMAB Collaborators– Hyun-Sook Kim

– Jamese Sims

– Hae-Cheol Kim

– Dan Iredell

• DA Collaborators– John Derber

– Mingjing Tong

– In-Hyuk Kwon

– Emily Liu (JCSDA/HFIP)

• GCWMB Collaborators– Hua-Lu Pan

– Dmitry Sheinin

– Fanglin Yang

• MMB Collaborators– Dusan Jovic, Tom Black

– Matt Pyle, Brad Ferrier

• Ensemble Team– Jiyayi Peng

• Land Surface– Yihua Wu

• NHC– Wallace Hogsett

3

Operational HWRF Modeling System• Current operational HWRF

– Regional-Scale Ocean-Atmosphere Coupled Modeling System

– Non-Hydrostatic system of equations formulated on a rotated latitude-longitude, Arakawa E-grid and a vertical, pressure hybrid (sigma-P) coordinate.

– Based on WRF NMM V2.0 framework with movable, vortex following high-resolution nested grid, 27 km outer domain, 9 km inner domain, 42 vertical levels

– Coupled to Princeton Ocean Model (POM) in the Atlantic (with feature based initialization of cold wake, loop current, warm and cold core eddies)

– Physical parameterizations designed for tropical environment (GFS/GFDL SAS Convection and PBL, GFDL Surface Physics, Radiation, Ferrier Microphysics)

– Advanced vortex initialization and GSI/3DVAR data assimilation consisting of Vortex relocation, Storm size and intensity correction based on tcvitals, Assimilation of satellite radiance datasets in the hurricane environment and combination of bogus (synthetic) vortex and six-hour cycling

• Upgrades implemented in 2010 to improve the forecast skill.

– New baseline version with several bug fixes

– Modified surface physics formulation and use of Gravity Wave Drag parameterization

– Addition of satellite radiance data assimilation in the hurricane environment

– Focus on reducing intensity bias and improve track forecasts

4

HWRF Atlantic Track Forecast Errors HWRF Atlantic Intensity Forecast Errors

(less skillful at all times)

Negative bias for HWRF Atlantic Intensity

Forecasts

Performance of the Operational HWRF for 2010 ATL hurricane season (till date)

Rapid growth of

intensity errors

5

HWRF E-Pac Track Forecast Errors HWRF E-Pac Intensity Forecast Errors (less

skillful at all times)

Negative bias for HWRF E-Pac Intensity

Forecasts

Performance of the Operational HWRF for 2010 EPAC hurricane season (till date)

Rapid growth of

intensity errors

6

64 kts threshold for storm size/intensity correction

(based on more than 1200 forecasts)

Systematic nature of HWRF Initialization

6

Positive bias for weaker storms, negative for stronger storms (spinup/spindown)

7

Vortex Initialization

Issues/Challenges

Broad initial vertical structure

Insufficient vortex size correction

H210 model producedFewer strong storms Than observed

HWRF 2009 productionhas too many strong storms

8

0

50

100

150

200

250

300

350

400

450

500

550

600

0 50 100 150 200 250 300 350 400

Radius (km) before the size correction

Rad

ius

(km

) aft

er th

e si

ze c

orr

ectio

n

Model radius

Current size

correction

New size

correction

Size

correction

Intensity

correction

25.1

85.0

75.0

2

1

6h fcst

Modified vortex initialization

procedure: Improved mass/wind balance &

storm size correction, with a focus on reducing

initial spinup/ spindown issues, and improve

track/intensity forecast skill in the first 12-24 hrs

Development of

the hybrid DA

system (ESRL,

NCEP, Univ

Oklahoma and

NASA) to support

the hurricane

problem

9

10

Current HWRF activities/developments• Major upgrade to the HWRF modeling system for 2011:

– Establish a community repository of the operational HWRF and upgrade HWRF infrastructure to V3.2+ (collaboration with DTC)

– Couple to HYCOM ocean model in the Atlantic

– Improve vortex initialization for more realistic storm size and intensity corrections

– Physics upgrades include testing of new GFS convection, modified PBL and modifications to treatment of horizontal diffusion

– Simplified operational system with added flexibility and multi-platform compatibility

• Subversion based code management (EMC/DTC)– HWRF-POM and HWRF-HYCOM baseline configurations for T&E

– “Trac” system for tracking changes and version control

– Synchronization of EMC and DTC repositories

• Product development– Generating simulated GOES WV and IR imagery and simulated radar reflectivity

products starting in 2010

– Additional simulated microwave products in 2011/2012

– High-frequency output and additional derived variables for diagnostics

– Enhancements to GrADS based HPLOT diagnostics and visualization software

– New enhanced HWRF website for product display and navigation

11

(H)WRF V3.2+

(H)WRF V3.2

WRF V3.1.1

WRF V3.1

WRF V3.0

WRF V2.2

WRF V2.1

WRF V205/2004

07/2009

04/2010

08/2005

12/2006

04/2008

04/2009

HWRF in the community WRF NMM Framework

HWRF operational configuration (2011)

Extensive Testing (pre implementation)

WRF Repository (hosted by DTC)

HWRF2007

HWRF2008

HWRF2009

HWRF2010

upgrades2011

•Modified vortex initialization (storm size correction and balanced vortex)•Coupling to HYCOM in the Atlantic (more realistic response to HWRF)•New GFS Deep/Shallow Convection•Modified PBL, Radiation, Microphysics•Modified horizontal diffusion

Extensive Testing (individual upgrades)

HWRF Development and Repository at EMC

02/2011

04/2011 (H)WRF V3.3

O2R

R2O

12

Benchmarking HWRFV3.2–POM

three season testing, ATLANTIC

2008-2009-2010

Red: Oper. HWRF V2.0

Cyan: HWRFV3.2 w/POM

13

Red: Oper. HWRF V2.0

Cyan: HWRFV3.2 w/POM

Benchmarking HWRFV3.2 three

season testing, East Pac 2008-

2009-2010

14

Cyan: HWRFV3.2 w/POM

Magenta: HWRFV3.2 w/HYCOM

15

Ongoing Model development & collaborative efforts

• Further advancements to the HWRF modeling system (EMC & HRD)

– third nest capability, advanced diagnostic capability (diapost), idealized simulations

• Code management and community support at DTC

– Setup HWRFV3.2+ and provide support to the community through DTC

– Development of R2O/O2R infrastructure and testing facility at DTC

• Data assimilation (EMC, HRD, CIRA)

– Real-time test of the P3 TDR data flow from aircraft to NCO/TOC/AOC and assimilation using advanced GSI.

– Ensemble data assimilation (MLEF, CIRA) and hybrid EnKF (HRD).

• HWRF Diagnostics (HFIP, EMC, NHC, FSU, CIRA, HRD, UMBC/UMD)

– Identifying forecast errors from different components of model physics and dynamics

– Hurricane model diagnostics, evaluation and verification

– Develop a common and comprehensive diagnostics framework and tools to integrate model output with available observations for verification

– Cloudtop datasets for storm scale diagnostics

• HWRF Physics (URI, GFDL, ESRL)

– Surface fluxes, sea spray and wave coupling

– Physics for high-resolution (convection, micro physics, PBL, LSM )

16

HFIP Activities at EMC

• Active participation in Hurricane Forecast Improvement Project (HFIP):

– HWRF is the “HFIP Flagship Model” for improving the forecast skill in Stream 1 approach

– Representation in all HFIP teams from EMC

– Hurricane workshop and HWRF tutorial at DTC in Feb. 2009

– Participation in HFIP diagnostics workshop at NHC in Dec. 2009

– Annual HFIP review at EMC in April 2010

– HWRF Initialization Workshop at EMC in September 2010

– HWRF Diagnostics Workshop at EMC in September 2010

– HFIP Stream 1 Diagnostics Workshop/Planning Meeting on Nov. 8th, 2010

– Second HWRF Tutorial at DTC in April 2011

• Contributions to Stream 1.5 and Stream 2 activities– Participation in FSU’s real-time multimodel ensemble (Stream 1.5)

– Real-time parallels using HWRFV3.2 at 27km/9km (EMC)

– Real-time parallels using 2010 HWRF coupled to NOAH LSM (EMC)

– Real-time parallels using 2010 HWRF with advanced data assimilation (EMC)

– High-resolution parallels using HWRFV3.2 (9km/3km) (EMC+HRD)

– HWRF ensembles based on GEFS

17

HWRF Stream 1.0 Prioritized Needs for FY11

• Comprehensive testing of HWRFV3.2 – WRF upgrades from v2.0 to v3.2 considered as a major upgrade

– Three-season testing for both Atlantic and Eastern Pacific basins

– Two baseline versions, one coupled to POM and another coupled to HyCOM

• Focus on reducing intensity errors in the first 24 hrs of forecast

– Addressing the problem of rapid growth of intensity errors in the first 24 hr forecasts

– Improved initialization procedure to have realistic representation of initial vortex

– Address the departure from gradient wind balance

– Improve storm size correction procedure using 34-kt wind radius and ROCI

– Explore options for improving vertical structure of the initial vortex

– DA experiments using inner core observations from aircraft datasets

• Focus on improving track forecast skill– Improved nest motion algorithm and fix nest feedback issues

– Evaluate and address impacts of terrain and land interaction on synoptic flow

– New GFS deep convection, shallow convection and PBL

– Improved formulations for radiation-convection-microphysics interactions

– Improved DA in the storm environment

– Systematically evaluate model biases and identify sources of errors

18

HWRF Stream 1.0 Prioritized Needs for FY11

• Focus on improving intensity forecast skill – Understand and evaluate the predictability limits of HWRF model for intensity

prediction

– Revised surface physics formulation to include affects of air-sea interaction

– Increased horizontal resolution (3 km) using triple nested HWRF (Stream 1.5)

– Improved formulations for microphysics, PBL and convection

• Comprehensive HWRF model diagnostics – Major effort is under way to conduct comprehensive model diagnostics for

initialization, large scale and vortex scale

– Define specific evaluation metrics along with suggestions for associated observed datasets

– Identification of systematic errors and model biases

– Automated model diagnostics tools to assist real-time, post-season and pre-implementation T&E of model performance

• Continuous advancements to the HWRF modeling system– Improved DA techniques to make better use of available inner core

observations

– Advancements to community WRF infrastructure to benefit operational HWRF

– Contributions from community to assist model improvement

– Dedicated efforts to develop physics suitable for high resolutions

– Enhanced interactions with other modeling groups

19

Advancing the HWRF System

.

2011 2012 2013 2014* 2015*

Resolution/

Infrastructure Triple nested HWRF

(27/9/3 km)

Increased vertical resolution, higher

model top, upgrades to WRF

infrastructure, NEMS/ESMF/NMM-B,

community R2O efforts (HFIP)

Physics Convection,

PBL,

Horizontal

diffusion

Shallow convection, Microphysics, Radiation, Surface

Physics, Coupling to Waves and Land Surface,

Physics for high-resolution

DA/ Vortex

Initialization

Storm size

correction,

gradient

balance

Inner core

(Doppler

Radar,

satellite)

Hybrid-EnKF DA, advanced vortex

relocation procedure, improved GSI

Ocean HYCOM

Coupling

Improved ocean data assimilation, physics and

resolution

Waves One-way Wave Coupling Two-way wave coupling, multi-grid surf

zone physics, effects of sea spray

Diagnostics and Product

Development

HWRF Ensembles, Coupling to Hydrological/ Surge/

Inundation models, diagnostics, product development

Ongoing Work

Potential 2011

upgrades

Ongoing

developments

*Potential

Computer upgrade

HWRFV3.2+

20

Thanks for your attention

Questions?

Real-time and pre-implementation T&E

HWRF products: http://www.emc.ncep.noaa.gov/gc_wmb/vxt/index.html

21

Additional Slides

• Key issues

22

DA Experiments with

TDR Data (super-obs)

for Hurricane Earl

23

Nest motion issues

• Moving nest algorithm– Mathematical

requirements• Stable and unique

storm center

• Being able to track the correct storm center

• Effects of topography (MSLP calculation)

– Computational requirements

• Effective and efficient

– Solution• Centroid MSLP

Montgomery et al.(2006)

C

24

Triple nested

domain

configuration

25

Coupling to HYCOM

• Systematic cooling in of ocean in

HYCOM-HWRF:

– Likely due to flux differences with

GFS.

– HYCOM set up for GFS fluxes.

– POM also cools ….

Possible solutions:

◦ Adjust HWRF fluxes to become similar to GFS fluxes

to remove spurious ocean cooling.

◦ Long term: make GFS and HWRF fluxes consistent.

26

POM vs

HYCOM

IKE example

◦ Model SST minus

daily average GOES /

AVHRR data.

◦ POM on left, too

warm under

hurricane.

◦ HYCOM on right,

too cool in general.

27

Eddy Loop CurrentT=54 hr

T=0 hr 1. Rigid subsurface

structure

2. A wave pattern in

the SST field

(noise)

3. Weak & shallow

Loop Current;

warm core eddy

velocity is larger

and deeper than

the Loop Current

4. Counter currents

exists beneath the

left and right

flanks of the Loop

Current

5. Weak Gulf Stream

signature

POM SST and Velocity Simulation

28

HYCOM SST Simulation – Field (left) and Transect

The Loop Current

warm core eddy

29

Eddy Loop CurrentT=0hr

T=36hr

HYCOM SST Simulation – Field (left) and Transect

30

Band 3 Loop Simulated Band 3 Loop Real

30

31

Biases in HWRF Synthetic WV

Imagery

Alex Colin

TD05 Danielle

31

32

A schematic diagram of the new GFDL/HWRF coupled wind-wave-current

modeling system and the air-sea interface module (ASIM) represented by the

following components: MFLUX, Sea spray model, MFBudget, and WFLUX. The

arrows indicate the prognostic variables that are passed between the model

components.