Plans for ShortRange Ensemble Forecast at INM

1
Plans for Short-Range Ensemble Forecast at INM

• José A. García-Moya
• SMNT INM
• Workshop on Short Range Ensemble Forecast
• Madrid, October, 3-4 2002.

2
Goals of the workshop

• Methodology and guidelines to build our Ensemble
  System, but after yesterday presentations
• Guidelines to chose a System name
• Downscaling ECMWF ? EPS
• LEPS (Tibaldi, ARPA)
• LAMEPS (Mylne, UKMO)
• SREF (Du, NCEP)
• PEACE (Nicolau, MeteoFrance)
• LAMEPS (Theis, Univ Bonn-DWD)
• SREPS (García-Moya, INM)
• SREPULAM (Orfila, INM)

3
Introduction

• Surface parameters are the most important ones
  for weather forecast.
• Forecast of extreme events (convective precip,
  gales,) is probabilistic.
• Short Range Ensemble prediction can help to
  forecast these events.
• Forecast risk (Palmer, ECMWF Seminar) is the goal
  for both Medium- and, also, Short-Range
  Prediction.

4
Errors of the short range forecasts

• Due to model formulation.
• Due to simplifications in parameterisation
  schemes.
• Due to uncertainty in the initial state.
• Special for LAMs, due to errors in lateral
  boundary conditions.

5
SREPS I

• Multi-model approach.
• Stochastic physics.
• Multi-boundaries
• From few global deterministic models.
• From global model EPS (ECMWF).

6
SREPS II

• Different assimilation techniques
• OI
• Variational (3D or 4D)
• Perturbed analysis
• Singular vectors (ECMWF).
• Breeding (NCEP).
• Scaled Lagged Average Forecast (SAMEX).

7
What do we need?

• Enough computer power.
• Research
• Following workshop recommendations.
• Technical difficulties.
• Large storage system.
• Database software (MARS).
• Postprocessing and graphics software.
• Enough staff for maintenance and monitoring.
• Verification software.

8
INM New computer

• Cray Inc. SV2.
• 40 vector PEs in June 2003.
• 32 additional PEs in December 2004.
• Around 360 IFS sustained Gfs in December 2004.

9
Characteristics

• Integration area Euro-Atlantic.
• Resolution, around 25 Km and 31 vertical levels.
• 72 hours forecast.
• 4 / 8 times a day (depending on computer power).

10
Area
11
Multi-model

• Hirlam
• HRM, former limited area model from DWD.
• MM5.
• ETA model or any other available numerical model.

12
Boundary Conditions

• From different global deterministic models
• ECMWF.
• Hirlam.
• AVN (NCEP).
• GME, DWD global model.

13
Multi-analysis

• Assimilation techniques
• Hirlam OI
• Hirlam 3DVAR
• Perturbations (research needed)
• SLAF (AN ? K ( AN - FCHH ), 0 ? k ? 1 )
• Singular vectors
• Breeding

14
Ensemble

• 4 models.
• 4 boundary conditions.
• 2 analysis.
• 4 last ensembles (HH, HH-6, HH-12, HH-18).
• 32 member ensemble every 6 hours
• Time-lagged Super-Ensemble of 128 members every 6
  hours.

15
Verification

• Objective verification against observations
• Show different performance of members.
• Try a quality index for each member (research
  needed).
• Probabilistic forecast verification
• ROC curves.
• Economic value of the forecast.
• Rank histograms (Talagrand diagrams).

16
Assumption on probability

• Basic research needed
• Same probability for each member of the ensemble.
• Probability function of the quality index of the
  member (DEMETER result on multimodel ensemble).

17
Implementation plan

• Computer depending
• Offline daily running June 2004.
• Operations June 2005.

18
Further research

• Singular vectors vs. breeding.
• Stochastic physics.

19
(No Transcript)
20
Few control runs Assim cycle

• Hirlam 3DVAR ECMWF boundaries (6 hours old).
• HRM, OI analysis GME boundaries (12 hours old).
• ETA, OI/3DVAR analysis AVN (NCEP) boundaries
  (12 hours old).
• IFS/INM 3DVAR analysis.
• UKMO Unified model analysis UKMO boundaries (6
  hours old).