Research Results from TIGGE and a Vision for a New Paradigm for Global Prediction David Parsons Chief, World Weather Research Division (WWRD)

Research Results from TIGGE

and a Vision for a New Paradigm for Global Prediction

David Parsons

Chief, World Weather Research Division (WWRD)

On behalf of the THORPEX Working Group on GIFS-TIGGE

and the Research Users of TIGGE

Special thanks to Zoltan Toth (NOAA), Richard Swinbank (Met Office), Philippe Bougeault (Meteo France),

THORPEX Intl Project Office, and TIGGE data providers

Baudouin Raoult, Manuel Fuentes - ECMWF

Steven Worley, Doug Schuster - NCAR

Bian Xiaofeng, Li Xiang - CMA

T. Nakazawa - JMA and Tang Xu -- SMB

1. What is TIGGE (THORPEX Interactive

Grand Global Esemble) and why is it important?

Pre-CAS Technical Conference

Background: the TIGGE archive• Detailed output from global ensemble forecasts to around 14 days generated

routinely at different centres around the world• Three archive centres: CMA, NCAR and ECMWF (user access began May 2006)• Ten data providers:

– ECMWF, JMA (Japan), UK Met Office (UK), CMA (China), NCEP (USA), MSC (Canada), Météo-France (France), BOM (Australia), KMA (Korea), CPTEC (Brazil) --- 1st provider began on Oct 2006 all 10 providers starting on Feb 2008

• Data exchanges using UNIDATA LDM, HTTP and FTP• In Sept 2009, 230 TB, 1.5 billion fields, growing by > 1 TB/week, 1.7 million fields/day

Why is TIGGE important? Accelerate shift of weather research and operational

communities toward ensemble prediction, which is appropriate for a non-linear, partly chaotic atmosphere

Advance ensemble research for high impact weather including enhancing collaboration between the academic and operational communities

Develop theory and practice of multi-model ensembles

Examine the feasibility of interactive ensembles responding dynamically to changing uncertainty

Develop the concept of a Global Interactive Forecasting System (GIFS)

A data base to improve individual ensemble systems including providing a challenging bench mark for the best (better) operational centres

Pre-CAS Technical Conference

User Metrics ECMWF: Over 500 users

Chinese

2. What are the results of TIGGE research?

Research results based on TIGGEMore available from http://tigge.ecmwf.int/references.html

Early research results will be shown for:Upper-air variablesSurface temperatureHeavy rainfallTropical cyclone tracks

Acknowledgments to

Young-Youn Park, KMA; Renate Hagedorn, ECMWF; Florian Pappenberger, ECMWF; Richard Swinbank et al., UK Met Office, Tetsuo Nakazawa, JMA/MRI, Takuya Komuri JMA, T. Krishnamurti FSU and Sharan Majumdar, RSMAS/U Miami

Preliminary conclusions for upper air variables Significant differences in quality between the systems

Up to 3 days differences in probabilistic forecast skill Agreement between spread and skill is the most variable aspect and

has a strong impact on probabilistic skill scores In the Tropics the spread is underestimated by almost all systems

Impact of the verification analysis Relatively little impact in the extra-Tropics (as long as the analysis

comes from one of the best systems) Large impact in the Tropics (and difficult to decide which is the best

analysis) Skill of multi-model system versus single-model systems

Only marginal improvement in the extra-Tropics Significant improvement in the Tropics (subject to significant bias

corrections)

Surface temperature Johnson and Swinbank (and subsequently Hagedorn) found from

comparisons with surface observations that the TIGGE Multi-Model (MM) forecasts of T2m outperform significantly any single model EPS

Interpreted as a proof that the variety of physics (soil, vegetation, PBL) between the models captures better the uncertainty in surface parameters

Results are sensitive to the choice of verifying analysis

Generally speaking, MM superiority comes from ECMWF, and ECMWF alone is better than any MM without ECMWF

Calibration using recent forecasts reduces the superiority of the MM, calibration using a special set of re-forecasts may offset completely the superiority of the MM and the superiority of the MM may also be challenged if uncertainty in soil moisture is added in the single systems

Incidentally: T2m from TIGGE database at Fcst time=0 is NOT an analysis of T2m temperature (it is an “intelligent” vertical interpolation) - do not use it for verification!

Verification of T2magainst observations

Multi-ModelECMWF

Met OfficeNCEP

T-2m, 250 European stations2008060100 – 2008073000 (60 cases)

0 2 4 6 8 10Lead time / days

-0.2

0.0

0.2

0.4

CRPSS

Benefits of re-forecast calibration

0 2 4 6 8 10 12 14fc-step (d)

-0.08

0

0.08

0.16

0.24

0.32

0.4

0.48

0.56

CRPSS (-)

2008060100-2008083100 (92 cases)ContinuousRankedProbabilitySkillScore

2m Temperature, Northern Extra-tropics

ECM WF-DM O

ECMWF-RFcali

TIGGE3-BC

Early Work on Heavy Rainfall

More challenging forecast and verification problem so the results are tentative and it

is difficult to generalize

Promising early results for prediction of Mei-Yu, S. China Sea Monsoon and post

typhoon heavy rainfall by Krishnamurti and colleagues

Used a subset of the models in the TIGGE archive (6 to 7 of the best models)

MM ensemble out performs the best ensemble with 1 to 2 days of lead time

added for a given level of forecast skill in the 2 to 5-day range and over 2-days of

lead time in the 10-day forecast

The study did not compare the MM to bias corrected single ensemble systems

Other work has focused on heavy rainfall and river flow in SE Europe with promising

results

David Burridge talk on general use of ensembles for flood prediction-Italian example

Tropical cyclonesTropical cyclones

Tropical cyclone tracks

Made available in near real-time for beginning in summer 2008 for the T-PARC project from Canada, China (2), ECMWF, Korea, Japan, UK, and USA (normally TIGGE has a 48-h delay)

Data is in CXML format from multiple centres hosted by Bureau of Meteorology/Australia and UCAR/USA at http://www.bom.gov.au/bmrc/projects/THORPEX/TC/index.html

Systematic investigations of performance are underway but the data set is generating significant interest in the tropical cyclone research and forecasting communities

WMO Executive Committee recommended that this real-time availability continue

JMA Ensemble Spread for TC Nargis from a Forecast Starting

Over 7 Days Before Landfall

Courtesyof T.

Nakazawa

Courtesyof T. Nakazawa

3. What is (GIFS) Global Interactive Forecast System

Real-time Operational Extension of the TIGGE Research Concept

GIFS likely emphasis on severe weather -- to advance lead-time, skill, and forecast confidence to mitigate loss of life and property and to contribute to the welfare of all WMO nations with a particular emphasis on least developed and developing countries

Will require real-time ensemble data access, product generation and distribution (common web interface using WIS concepts) --- all major efforts

Goal is have operational transition for the Severe Weather Forecasting Demonstration (SWFDP) Project --

Will require some pilot projects to develop products and then Forecast Demonstration Projects to test implementation

Global Interactive Forecast System (GIFS)

A major consideration for all Members (particularly developing nations) is the requirement to develop efficient severe weather warning systems focusing on national needs without becoming overly dependent on a large provider.

GIFS is ideally a shift of paradigm from bilateral cooperation (with a strong partner and a weak partner who needs to adjust to any change decided by the strong) to multi-lateral cooperation

GIFS has the potential to enable the provision of a variety of products of similar quality in standard format giving all countries both a sense of independence and of ownership/control of their own 'customized‘ severe weather products. 

PARADIGM FOR GIFS DEVELOPMENT

A First Step: NW Pacific Tropical Cyclones Ensemble Prediction Experiment

A five-year regional project with both a research and operational component that will feed into national efforts such as for the Shanghai MHEWS Project

Intent is to make GIFS-TIGGE typhoon track data available to Typhoon Committee members including RSMCs in near-real time beginning May 2010 via a password protected web site

Training for operational forecasters and evaluation of the utility of such data sets in a

forecast environment

Research efforts will attempt to develop and extract useful information from TIGGE ensemble data (or a subset of this data since many ensembles are not well designed for the tropics) and develop products where appropriate

Product development to feed into the operational SWFDP in Africa and then in the S. Pacific (already a request from the SWFDP to feed into RSMCs in Africa for such data)

Specific plans yet to be presented to data providers but consistent with EC and THORPEX ICSC decisions

4. Future efforts

a) A Limited Area Modeling Version of TIGGE (TIGGE-LAM)

b) Extension of the concept to monthly and seasonal

prediction?