Page 1© Crown copyright 2004 ECMWF Forecast Products Users Meeting 15th June 2006.

© Crown copyright 2004 Page 1

ECMWF Forecast Products Users Meeting

15th June 2006


New Development in Monthly Range prediction at the Met Office

Bernd Becker, Paul M. James.

Met Office monthly forecast suite

Products from the Monthly Outlook

Grosswetterlagen Analysis


Monthly Forecasting System

Coupled ocean-atmosphere integrations: a 51-member ensemble is integrated for 32 days every week.

Atmospheric component: IFS with the latest operational cycle30r1 and with a T159L62 resolution (320 * 161)

Oceanic component: HOPE (from Max Plank Institute) with a zonal resolution of 1.4 degrees and 29 vertical levels

Coupling: OASIS (CERFACS). Coupling every ocean time step (1 hour)

Perturbations:

Atmosphere: Singular vectors + stochastic physics

Ocean: SST perturbations in the initial conditions + wind stress perturbations during data assimilation.

Hindcast statistics:

5-member ensemble integrated over 32 days during the past 12 years.

Representing a 60-member ensemble.

Running every week


Q5

Q4

Q3

Q2

Q1

T

Q1 Q2 Q3 Q4 Q5

Harvesting the Ensemble(1): Rank Ordering

•Ignore shape in the baseline•Rank ordering the hind cast•Slicing into equally large chunks•Counting the forecast members in each category

Warm and dry

Cold and wet

P


Post processing

1. Data Volume reduction Derive properties of the PDF

2. Interpolation to 10 UK climate regions Down scaling

3. Calibration with historical data Bias correction

4. Interpretation of the histogram Deterministic quintile category

5. Mapping Deterministic value


Example UK temperature forecast for 10 climate districts


Holiday planner June/July 2006

Tmax

Precipitation

Week 1 Week 2 Week 3&4


-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

CSI Q PSS HSS GSS BS BSS ROC

PersitenceMPQPersPDFPDF

UK average Skill scores…

CSI: critical success indexQ: odds ratio, Yule’s Q.PSS: Peirce Skill ScoreHSS: Heidke Skill ScoreGSS: Gerrity skill ScoreBS: Brier Score BSS: Brier skill ScoreROC: Area under the ROC curve

..are derived from a 5 * 5 * 10 contingency table.Each cell records matching: Tmean, days 12 - 18• Observation / Forecast category and • the probability with that the category was predicted

Scores are calculated per category, figures in graph below are averaged over 5 categories.


Harvesting the Ensemble (2) : Grosswetterlagen (GWL)

(Paul James’ work)• A subjective classification of 29 large-scale weather types, conceived

by Baur et al. (1940s), revised by Hess and Brezowsky (50s to 70s) and maintained by the German Weather Service (to present)

• GWL patterns are characteristic synoptic circulation types, covering most of Europe and N.E. Atlantic while focused on Central

Europe• GWL events must last at least 3 days – they define “regimes”• Conceptually one of the best classification systems in existence• ( Note that GWLs are a form of clustering into a fixed number of

possible states, where the clusters have distinct synoptic meaning with consistent large-scale characteristics )

but:• Subjective, probably non-homogeneous over time• Large-scale patterns often inconsistent outside of Central Europe• Not applicable to NWP etc. unless they can be made objectiveobjective


Examples of GWL-Composites for mid-June, based on ERA40

MSLP Contours,

Precipitation Colour-

Fill Fields

2m-Temperature Anomaly Circles


Empirical Objective-GWL Classification Method

(1) Form GWL-Composites• MSLP and Geopotential Height at 500 hPa (Z500)

• ERA40, 1958-2002

• Use the official (subjective) GWL catalogue for this

• Separate composites for Winter and Summer half-years, sinusoidally-weighted, centred on mid-January / mid-July

(2) Pattern Correlations• Correlate daily MSLP / Z500 fields with each GWL base composite

• Highest correlating GWL taken as GWL for this day

• Apply subsequent temporal filtering ( logical steps ) to set most appropriate GWL regime (must last at least 3 days each)


Objective-GWLs in Ensemble Forecasts

• Run objective-GWL algorithm on each ensemble member• Yields a set of 51 catalogues of daily GWLs• Compare e.g. mean frequency of occurrence of each GWL

against hindcast and climatological observed (e.g. ERA40) frequencies

Added Value:• Indicates probable dates for changes of regime• Can form the basis for a meaningful synoptic clustering of

possible outcomes• Shows the specific influence of synoptic-scale circulation

anomalies in the forecast• Communicates the ensemble outcomes in a very effective

way to synoptic meteorologists


Histogram of GWL


Objective-GWLs in Ensembles: Verification

• Method has been running weekly on the monthly forecast since 16.03.2006

• Rigorous verification method will be needed• Quick first-order verification on most probable daily GWL

(GWL having the most ensemble members each day) has been made using following daily scores:• 2 points when GWL correct

• 1 points when a near-neighbour GWL predicted (subjectively, each GWL has about 5 near-neighbours)

• 0 points when GWL wholly incorrect ( 23 out of 29 GWLs, resp.)

• Add up points to the end of May

• Random chance should give a mean of about 3 points per day over 10 forecast weeks (ie. 0.3 pts per day per forecast)


Objective-GWLs in Ensembles: Verification

0

2

4

6

8

10

12

14

16

18

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Correct GWLs

Skill on THORPEX timescales

No obvious deterministic skill beyond about 16 days *

* But probabilistic breakdown of GWL frequencies may contain skill

Scores

Forecast Day ( T+x )


Post processing: Rank Ordering Method

Data Volume reduction before transfer to The Met Office: Calculate

1. Tercile/Quintile boundaries from the Hindcast ensemble2. Tercile/Quintile populations from the Forecast ensemble3. Maximum, Mean and Minimum from Forecast and from Hindcast4. Forecast Tercile/Quintile averages5. Average in time to week 1, 2 and 3&4.

UK Forecast: http://www.bbc.co.uk/weather/ukweather/monthly_outlook.shtml

1. Interpolation to points representing UK climate regions2. Calibration with historical UK climate region observations3. Interpretation of the Histogram, Ensemble mean or Mode in cases

with large spread, • derive deterministic forecast tercile/quintile

4. Mapping Tercile/Quintile average onto calibration PDF • to derive deterministic forecast value


Post processing: Grosswetterlagen Method

• Correlate daily MSLP / Z500 fields with each GWL base composite

• Highest correlating GWL taken as GWL for this day

• Apply subsequent temporal filtering ( logical steps ) to ascertain most appropriate GWL regime lasts at least 3 days

• Run objective-GWL algorithm on each ensemble member

• Yields a set of 51 catalogues of daily GWLs

• Compare e.g. mean frequency of occurrence of each GWL against hindcast and climatological observed (e.g. ERA40) frequencies


Holiday planner for June/July 2006


Holiday planner for June/July 2006

Wind

Wk 1

Wk 2

Wk 3&4


Future Work:

port Standardised Verification system (SVS) to R, compare with other verification packages

More streamlinedMore communicationMore efficient

Exploit daily data:Environmental Stress index (Heat stress)Monsoon onsetPeriod statistics, days above a threshold

Description of the histogram/PDF in an analytical form, derived from Mean, Standard Deviation, Skewness and Kurtosis

More complete description of the PDFLess data to carry around


Conclusion

• The monthly forecasts model runs are produced at ECMWF, products are derived at the Met Office, operationally.

•Europe is a difficult region to predict at long time range.

•The Monthly Outlook is a powerful tool to provide forecast guidance up to a month ahead in many areas.

•Grosswetterlagen analysis:•indicates probable dates for changes of regime

•can form the basis for a meaningful synoptic clustering of possible outcomes

•shows the specific influence of synoptic-scale circulation anomalies in the forecast

•communicates the ensemble outcomes in a very effective way to synoptic meteorologists

Page 1© Crown copyright 2004 ECMWF Forecast Products Users Meeting 15th June 2006.

Documents

crown copyright

histogram of gwl slide

probable daily gwl gwl

examples of gwl

objective slide

week slide

objectivegwl algorithm

present gwl patterns