The THORPEX Interactive Global Ensemble (TIGGE)

Multi-model ensembles and

Tropical cyclone forecasting

Richard Swinbank, with thanks to

the GIFS-TIGGE working group, THORPEX IPO and Met Office & other colleagues

Ensemble TC forecasting workshop, Nanjing, 14-16 December 2011

TIGGE and ensemble forecast products

TIGGE Objectives TIGGE archive

Ensemble forecasting Probabilistic forecasting Use of multi-model ensembles

Forecast products Plans for development of Global Interactive Forecast System Developing links with CBS/SWFDP Tropical cyclone forecast products

TIGGE THORPEX Interactive Grand Global Ensemble

A major component of THORPEX: a World Weather Research Programme to accelerate the improvements in the accuracy of 1-day to 2-week high-impact weather forecasts

GEO task WE-06-03 – “TIGGE and the Development of a Global Interactive Forecast System for Weather”

Objectives: Enhance collaboration on ensemble prediction, both internationally

and between operational centres & universities. Facilitate research on ensemble prediction methods, especially

methods to combine ensembles and to correct systematic errors Enable evolution towards a prototype operational system, the “Global

Interactive Forecast System”

TIGGE data Ten of the leading global forecast centres are providing regular

ensemble predictions to support research on predictability, dynamical processes and the development of probabilistic forecasting methods.

TIGGE data is made available for research after a 48-hour delay. Near real-time access may be granted for specific projects through the THORPEX International Project Office.

Summary of TIGGE database (late 2010)

CentreEnsemblemembers

Output dataresolution

Forecastlength

Forecasts per day

Fields (out of 73)

Start date

BOM* 33 1.50º x 1.50º 10 day 2 55 3 Sep 07

CMA 15 0.56º x 0.56º 10 day 2 60 15 May 07

MSC 21 0.9º x 0.9º 16 day 2 56 3 Oct 07

CPTEC 15 0.94º x 0.94º 15 day 2 55 1 Feb 08

ECMWF 51N200 (Reduced

Gaussian)N128 after day 10

15 day 2 70 1 Oct 06

JMA 51 0.56º x 0.56º 9 day 1 61 1 Oct 06

KMA* 17 1.00º x 1.00º 10 day 2 46 28 Dec 07

Météo-France 35 1.50º x 1.50º 4.5 day 2 62 25 Oct 07

NCEP 21 1.00º x 1.00º 16 day 4 69 5 Mar 07

UKMO 24 0.83º x 0.55º 15 day 2 72 1 Oct 06

* Delivery of BoM data currently suspended; KMA resumed Aug 2011

TIGGE Archive Usage(NCAR + ECMWF)2010-11 TIGGE Archive Usage (All Portals)

1

10

100

1000

10000

100000

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug

Month

Volu

me

(GB)

0

30

60

90

120

150

Num

ber o

f Use

rs (C

ount

)

Vol Accessed (GB)

Vol Delivered (GB)

# Active Users

Information about TIGGE

Major Article in BAMS

New leaflet to publicise TIGGE to researchers

Contribution in GEO book “Crafting Geoinformation”

Tropical cyclone case study in WMO Bulletin

TIGGE website

TIGGE Research

Following the successful establishment of the TIGGE dataset, the main focus of the GIFS-TIGGE working group has shifted towards research on ensemble forecasting. Particular topics of interest include:

a posteriori calibration of ensemble forecasts (bias correction, downscaling, etc.);

combination of ensembles produced by multiple models; research on and development of probabilistic forecast

products.

TIGGE data is also invaluable as a resource for a wide range of research projects, for example on dynamical processes and predictability. Up to the end of 2010, 43 articles related to TIGGE have been published in the scientific literature

Ensemble forecasting

time Forecast uncertainty

Climatology

Initial Condition Uncertainty

X

Deterministic Forecast

Analysis

Ensemble forecasting - effect of Chaos

The atmosphere is a chaotic system: “… one flap of a seagull’s

wing may forever change the future course of the weather”, (Lorenz, 1963)

Tiny errors in how we analyse the current state of the atmosphere lead to large errors in the forecast – these are both equally valid 4-day forecasts!

Up to about 3 days ahead we can usually forecast the general pattern of the weather quite accurately

Beyond 3 days Chaos becomes a major factor

Fine details (eg rainfall) have shorter predictability

With acknowledgements to Ken Mylne for general Ensemble Forecasting slides

Lorenz Model – a simple chaotic system

Variations in predictability can be illustrated using the Lorenz (1963) model:

X aX aY

Y XZ bX Y

Z XY cZ

Simple non-linear system.

Ensemble Forecasting in the Lorenz Model

1. Predictable - deterministic OK

2. Predictable at first -

probability OK

3. Unpredictable climatology OK

Thanks to Tim Palmer, ECMWF

Ensemble Forecasting

The aim of ensemble prediction system is to represent the uncertainty in the state of the NWP model at all stages through the forecast range

Each ensemble member is designed to sample a PDF representing uncertainties in the model state. The PDF is usually represented by a control run plus many perturbed forecasts. Initial condition perturbations are designed to represent

uncertainties in the initial analysis – closely linked to data assimilation. These grow with time as a result of the chaotic nature of model dynamics.

The uncertainty in forecasts also grows as a result of model error. This effect may be represented by perturbing model physics – e.g. stochastic physics schemes.

Examples of comparison of ensemble spread & error (from 2010 MOGREPS upgrade)

Compare

Red – old

Blue − new

Note:

Ensemble spread is less than RMS errors

New system both reduces RMS error & increases spread

Under-spread in surface temperature exaggerated because of representivity errors (point observations not representative of grid-square averages)

time

Estimated forecast uncertainty

Initial Condition Uncertainty

X

Deterministic Forecast

Analysis

Under-dispersive forecast

Real forecast uncertainty

The ensemble may capture reality less often than it should.Dangerous - false sense of security!

What makes a good ensemble?

Observed

Resolution – ability to distinguish between different events.

Reliability – the forecast probabilities match the associated observed frequencies.

Calibration – apply post-processing so that the statistics match those of a reference dataset. This might include bias-correction and variance adjustment.

The observed track should lie within the ensemble.

Met Office Multi-Model Ensemble

NCEP21 Member

ECMWF51 Member

Met Office 24 Member

3 variables:

MSLP

2m Temp

500mb Height

Met Office MME results courtesy

Christine Johnson

One approach to improving the calibration is to use multi-model ensembles

ECMWF

MultimodelNCEP

UKMO

Reliability Diagrams

Multi-model gives better reliability and reduced sharpness.

Probability that 2m temperature is greater than the climatological mean at T+240. Globally averaged.

Brier

Reliability

Resolution

Probability of 2m temperature greater than climatological mean.

Brier Skill Scores

Multi-model gives improvement in reliability and resolution at all lead times.

1 day

5 day

Dissimilarity of ensembles

betweenD

between MSE

D has large values (high dissimilarity, red) if the between-model variance is large compared to the mean-square-error of the multi-model mean.

mslp temp

D+2

D+10

D+2

D+10

time

Estimated forecast uncertainty

Initial Condition Uncertainty

X

Deterministic Forecast

Analysis

Multi-model ensemble forecast

real forecast uncertainty

Use of multi-model ensembles can improve the sampling of forecast uncertainty

Multi-model ensemble forecasts of T850

Demonstrates benefit of multi-model ensemble, provided that the most skilful models are used.

Renate Hagedorn, ECMWF

Multi-model ensemble compared with reforecast calibration

Reforecast calibration gives comparable benefit to multi-model ensemble

Choice of verification data set (in this case, ERA-Interim) could have subtle but significant effect on relative benefits

Calibration could further enhance benefit of multi-model ensemble

Renate Hagedorn

Precipitation forecasts over USA

24 hour accumulations, data from 1 July 2010 to 31 October 2010.

20 members each from ECMWF, NCEP, UK Met Office, Canadian Meteorological Centre.

80-member, equally weighted, multi-model ensemble verified as well.

Verification follows Hamill and Juras (QJ, Oct 2006) to avoid over-estimating skill due to variations in climatology.

Conclusions:

ECMWF generally most skillful.

Multi-model beats all.

Tom Hamill

Summary - EPS & multi-model ensembles

Ensembles are valuable for forecasting the risks of exceeding thresholds (e.g. for high-impact weather events).

But forecasts often need calibration to correct both biases and variability (e.g. to correct under-estimates of forecast spread).

The best approach is to address the systematic errors, i.e. reduce model biases and improve the representation of model errors in the EPSs.

That is a long-term goal. In the mean time…. Use of multi-model ensembles is a pragmatic approach that

reduces calibration errors, especially where models have similar skill but different types of systematic error

Reforecasts (forecasts of past cases with current NWP models) can be used to estimate, and then correct, model biases

Towards the Global Interactive Forecast System (GIFS)

Many weather forecast situations are low probability but high risk – unlikely but potentially catastrophic. Probabilistic forecasting is a powerful tool to improve early warning of high-impact events.

The objective of the GIFS is to realise the benefits of THORPEX research by developing and evaluating probabilistic products to deliver improved forecasts of high-impact weather.

GIFS-TIGGE WG has initiated a GIFS development project to develop & evaluate products, focused on Tropical cyclones Heavy precipitation Strong winds

Tropical cyclones

As a first step, the GIFS-TIGGE working group set up a pilot project for the exchange of real-time tropical cyclone predictions using “Cyclone XML” format.

Example of combined TC track forecasts (Met Office + ECMWF)

CXML track exchange

GIFS links with SWFDP

EPS1 EPS2 EPS3

Generate Products

Regional Centre

users

National Centre

National Centre

National Centre

users usersusers

In-SituObservations

AirborneObservations

Satellite Observations

GIFS will collaborate with WMO Severe Weather Forecast Demonstration Project (SWFDP) and other FDPs and RDPs to ensure that products address

needs of operational forecasters and end users;

to provide an environment for the evaluation of prototype products.

GIFS will use global-regional-national cascade pioneered by the SWFDP. No single “GIFS centre”.

Use of web-enabled technology for generation and distribution of products.

Tropical cyclone products:Examples from Met Office

Uses Julian Heming’s code to identify and track tropical cyclones (TC), originally developed for the deterministic global Unified Model.

Tropical Cyclones (TCs) are identified where 850hPa relative vorticity (RV) maxima are greater than a threshold

For TC tracking, use search radius of 4 degrees for analysis and 5 degrees for forecast positions

Identified storms that do not match with a named storm or a TC identified at a previous time are counted as TC genesis

Met Office TC products courtesyPier Buchanan, Helen Titley and Julian Heming

MOGREPS-15 products for Ma-On, 12Z Friday 14th July 2011.

Left hand plot: 24 ensemble tracks Middle plot: strike probability – probability that the storm will be within 75 miles

within thhe next 15 days. Right hand plot: MOGREPS ensemble mean (blue), control (cyan), previous

observations (red) and deterministic track (green)

Products for named and potential storms on a basin by basin basis.

StormTracker Ma-On, 12Z Friday 14th July 2011.

Black – previous track White – deterministic model forecast Purple UKMO ensemble mean, mustard ECMWF ensemble mean and pink NCEP

GEFS ensemble mean.

TOMAS: MOGREPS-15 12Z North Atlantic forecast on Wednesday 27th October 2010.

Yasi MOGREPS Forecasts

Comparison of deterministic forecast with MOGREPS-15 ensemble mean

Comparison of MOGREPS, ECMWF and multi-model ensemble.

Tropical cyclone products from MRI/JMA http://tparc.mri-jma.go.jp/cyclone/

New tropical cyclone product:Strike probability time-series at a city

Tetsuo Nakazawa

Use of MOGREPS-15 products in SWFDDP

Warnings of heavy precipitation

Prototype product courtesy Mio Matsueda

Similar products also available for

• hot & cold temperatures,

• strong winds

Case 1: heavy rainfall by cyclone YASI (Feb. 2011)

+ 7-day forecast

TRMM dailyprecipitationexceeds the TRMM's 95th percentile.

Case 1: heavy precipitation by cyclone YASI

+ 6-day forecast

Case 1: heavy rainfall by cyclone YASI (Feb. 2011)

+ 5-day forecast

Case 1: heavy rainfall by cyclone YASI (Feb. 2011)

+ 4-day forecast

Case 1: heavy rainfall by cyclone YASI (Feb. 2011)

+ 3-day forecast

Case 1: heavy rainfall by cyclone YASI (Feb. 2011)

+ 2-day forecast

Case 1: heavy rainfall by cyclone YASI (Feb. 2011)

+ 1-day forecast

Other possible visualisation: combined warnings

Prototype product courtesy Mio Matsueda

ICDM workshop

Workshop on Dynamics and Predictability of High-Impact Weather Events and Climate Extremes

Kunming, China 6-9 August 2012 Organised by IAMAS International Commission on

Dynamical Meteorology Sponsored by WWRP, WCRP, IUGG, IAMAS,

Chinese IUGG committee, CAS, NSFC, LASG/IAP, Nanjing University & research projects.

Further information (first announcement) available from me.

Website: http://icdm2012.csp.escience.cn

Summary Since October 2006, the TIGGE archive has been

accumulating regular ensemble forecasts from leading global NWP centres.

The archive is a tremendous resource for the research community at large, particularly on EPS – including use of multi-model & other approaches to reduce systematic errors.

As the basis of the development of the future Global Interactive Forecast System, products are being developed to enhance the prediction of high-impact weather, starting with tropical cyclones.

GIFS products will be developed & evaluated in conjunction with SWFDP and other regional projects.

TIGGE website: http://tigge.ecmwf.int