Norwegian Meteorological Institute met.no GLAMEPS: GLAMEPS GLAMEPS: Grand Limited Area Model Ensemble Prediction System Overview of activities EWGLAM – Dubrovnik, October 2007 Trond Iversen with contributions from Inger-Lise Frogner, Edit Hágel, Kai Sattler, Roeland Stappers + more

Norwegian Meteorological Institute met.no is in real time to provide to all HIRLAM and ALADIN partner countries: an operational, quantitative basis for forecasting probabilities of weather events in Europe up to 60 hours in advance to the benefit of highly specified as well as general applications, including risks of high-impact weather. The GLAMEPS objective

Norwegian Meteorological Institute met.no Expectations from Short Range ensemble prediction How certain is today’s weather forecast? What are the risks of high-impact events? Forcasted risk = probability x potential damage (vulnerability) Lower predictability of “free flows” as scales decrease; i.e.: higher resolution increases the need for information about spread and the timing of spread saturation Predictability of “forced flows” is longer than “free flows”: i.e.: beneficial to separate unpredictable “free flows” from those strongly influenced by surface contrasts: e.g. topography, coast-lines, land-use, etc.

Norwegian Meteorological Institute met.no Basic Ideas in GLAMEPS An array of LAM-EPS models or model versions: –Each partner runs a unique sub-set of ensemble members –Partners who run the same model version, use different lower boundary data, or different initial and lateral boundary perturbations –Partners who run with DA, produce ensemble members based on initial and lateral boundary perturbations (one control with DA + pairs of symmetric initial perturbations) –Partners who do not run DA produce 6-20 ensemble members (pairs) Grid resolution –Now 20km, later: 10km or finer, 40 levels, identical in all model versions (should be increased to at least 60) Forecast range –60h (shorter?) - starting daily from 00UT and 12 UT A common pan-European integration domain –Or alternatively: a minimum common overlap

Norwegian Meteorological Institute met.no Aspects to consider 1.Operational aspects -In particular data storage and Real-Time distribution 2.Constructing initial and lateral boundary perturbations –Imported global eps-members enhanced w.r.t. resolution, European target, moist physics –LAM-specific perturbations (SVs, ETKF) 3.Lower boundary data perturbations –Stochastic perturbations –Switch surface schemes –Targetted Forcing Singular Vectors or Forcing Sensitivities 4.Model perturbations –Switching models (e.g. Aladin and Hirlam) –Switching physical packages (e.g. Straco, RKKF, ECMWF-physics) –Stochastic perturbations (EC: Cellular automata.) –Forcing Singular Vectors 5.EPS-calibration and probabilistic validation 6.Post-processing, graphical presentation, products 7.Further downscaling to meso- and convective scales

Norwegian Meteorological Institute met.no Quality objective To operationally produce ensemble forecasts with a spread reflecting known uncertainties in data and model; a satisfactory spread-skill relationship (calibration); and a better probabilistic skill than the operational ECMWF EPS; for the chosen forecast range of 60 hours (could be shorter); our common target domain; and weather events of our particular interest (European extremes - probabilistic skill parameters).

Norwegian Meteorological Institute met.no GLAMEPS Common Domain ALADIN Resolution: 22km 320 x 300 x 37 HIRLAM (EPS71) Resolution 0.2 deg. 306 x 260 x 40

Norwegian Meteorological Institute met.no GLAMEPS_v0: Laboratory at ECMWF To select a small set of model versions which are equally valid but significantly different, –3 different models: ALADIN, HIRLAM-STRACO, HIRLAM RK-KF To construct initial/lateral boundary perturbations –New ECMWF TEPS: define TSVs targeted to 3 domains all TSVs are orthogonal to NH SVs (EPS) and mutually TSVs: OT=24h, T159, (not yet diabatic) –Use: 30 TSVs and 50 NH SVs, Gaussian sampling to 20 members + control Probabilistic estimation (e.g. BMA), –Not started yet: awaits results Products; Quality and Value –INM package based on Magics –Predictability of the day, event risks –Reliability, BSS, ROC, Value, …

Norwegian Meteorological Institute met.no TEPS FOR EUROPE Inger-Lise Frogner GLAMEPS integration domain (HIRLAM version) Target area north (82N,15W,50N,50E) Target area central (62N,20W,33N,44E) Target area south (47N,23W,24N,32E)

Norwegian Meteorological Institute met.no First experimental setup for “TEPS for Europe” Singular vectors are computed with: –T159L62 (as opposed to T42 for operational NH SVs at ECMWF) –24h optimization time (as opposed to 48h) –Targeted in the vertical to the troposphere –Targeted SVs (TSVs) based on total energy norm –The TSVs are orthogonal NH SVs and mutually orthogonal TEPS perturbations are made from 80 SVs: –10 SVs for each of the three targets and 50 NH SVs and evolved SVs from EPS. –Includes standard stochastic physics Different amplitudes is assigned to the different sets of SVs, to give the desirable spread/skill relation –Presently under development

Norwegian Meteorological Institute met.no EXPERIMENTS 21 days in summer 2007: – , and The amplitude of NH SVs is kept as in EPS for the first experiment: The amplitude of TSVs from the three target areas for the first experiment: –Under adjustment!

Norwegian Meteorological Institute met.no COST TSVs for all three target areas: ca 450 SBUs TEPS for Europe: ca 3000 SBUs A total cost of ~3500 SBUs per run /case

Norwegian Meteorological Institute met.no Problems and obstacles Technical issues regarding running with three orthogonal sets had to be solved The high resolution of the TSVs caused unforeseen problems. The vertical diffusion scheme in the SV calculations in IFS had to be changed. The original scheme caused spurious growth.

Norwegian Meteorological Institute met.no Example of TSVs Lowest level, SV no. 2 NHSV TSV-north TSV-central TSV-south

Norwegian Meteorological Institute met.no Spread/Skill relationship MSLP, 21 summer cases 2007 ___ error of Ensemble Mean (EM), EPS ___ error of EM, Norwegian TEPS ___ error of EM, European TEPS ---- spread around EM, EPS ---- spread around EM, Norwegian TEPS ---- spread around EM, European TEPS

Norwegian Meteorological Institute met.no RMS Difference in spread between European TEPS and EPS over the 21 cases +12h +24h +36h+48h

Norwegian Meteorological Institute met.no Skill scores MSLP (example) 21 summer cases 2007 Black: EPS, 50 members Blue: European TEPS, 20 members Red: Norwegian TEPS, 20 members Event: anom < -5 hPa ROC area BSS

Norwegian Meteorological Institute met.no Further work Experimentation with the amplitudes of the SVs and TSVs will be carried out. –At the moment we are testing: Increasing the amplitude from the TSVs and at the same time reduce the amplitude from the SVs, in order to get better spread / skill in the range 0 to 60 hours as well as better scores. A winter period of 21 days will also be run. Scores for more parameters will be calculated: T850, ff10m, Z500, T2m After the tuning of TEPS is satisfactory, LAMEPS will be run with TEPS as initial and boundary conditions

Norwegian Meteorological Institute met.no HIRLAM EPS and ALADIN EPS

Norwegian Meteorological Institute met.no Main characteristics of HIRLAM EPS Kai Sattler 1. Calculates a control from HIRLAM 3d-Var (later 4D Var) + ensemble of perturbations 2. Can be used with downscaled ECMWF EPS: * 50+1 members * 12h cycle frequency * data availability: * online data * boundary data pool (intermediate storage, hindcast) 3. Old Norwegian TEPS: * 20+1 members * 24h cycle frequency * boundary data pool 4. New “TEPS for Europe” (not fully implemented yet): * 3 target areas * 12h cycle frequency * boundary data pool 5. Ensemble member specific environment settings choice of parameterization schemes 6. Selection of convection/condensation scheme for each perturbed member

Norwegian Meteorological Institute met.no ALADIN EPS Scripts for the execution of ALADIN/EPS system for the GLAMEPS domain(s) at ECMWF –(Stjepan Ivatek-Sahdan, ZAMG EPS team, Joao Ferreira) Operational ALADIN EPS (“ALADIN LAEF”) built in Austria at ECMWF machines used as starting point for the GLAMEPS-v0 laboratory So far tested: Downscaling ECMWF EPS for ALADIN. quasi-operational manner. Joao Ferreira (“Downscaling ECMWF EPS with ALADIN”)

Norwegian Meteorological Institute met.no Costs EPS_71T_15: Coarse test domain EPS_71_20: Full 0.2 deg. dom: SBU/48hfcst 22km: around 80 SBU/54h fcst HIRLAM ALADIN

Norwegian Meteorological Institute met.no Further R&D in parallel To gradually increase sources of spread, quality and reliability Include lower boundary perturbations and other types of model perturbations, e.g.: –vary model coefficients –Targeted Forcing SVs or Forcing Sensitivities, –weak 4D-Var perturbed tendencies –Stochastic physics Include alternative initial/lateral boundary perturbations –ETKF generalized breeding, –HIRLAM and ALADIN LAM SVs, Pdf-estimation, presentation, validation etc. –BMA, –Products –Validation

Norwegian Meteorological Institute met.no ALADIN SVs Efit Hágel, Richard Mladek

Norwegian Meteorological Institute met.no Case study – 27 August 2007, 00 UTC ALADIN singular vectors were computed GLAMEPS domain was used for the computations, but the target (optimization) area itself was smaller: 56N/34S/2W/40E (see on figure in green) Opt. area

Norwegian Meteorological Institute met.no Case study – 27 August 2007, 00 UTC ALADIN SVs - choices for case study: Norms: total energy norm (initial and final time) –Optimization area: 56N/34S/2W/40E –Optimization time: 12 and 24 hours –Vertical optimization: level (all levels) –Resolution: 22 and 44 km –LBC Coupling: every 3 hours (ARPEGE) Opt. area

Norwegian Meteorological Institute met.no Case study – ALADIN SVs Leading singular values for the experiments with different optimization time and resolution Lowest singular values with 44 km resolution and 12 hours optimization time Highest singular values with 22 km resolution and 24 hours optimization time

Norwegian Meteorological Institute met.no Case study – ALADIN SVs ALADIN leading singular vector at T+0h (top) and evolved singular vector at T+12h and T+24h (bottom) for wind v at model level km22 km44 km22 km 0h 12h /24h

Norwegian Meteorological Institute met.no HIRLAM SVs Jan Barkmeijer, Roel Stappers

Norwegian Meteorological Institute met.no Hirlam Case I101 (Fine) Analysis June 28 th 2006 at 3 UTC Resolution 0.2 x 0.2 degrees Optimization time 12 hours Dry total energy norm No projection operator Vertical diffusion in TL-model No diabatic processes in TL-model Leading singular value 6.6

Norwegian Meteorological Institute met.no Temperature and wind field of the leading singular vector at model level 19 (500 hPa) at initial (left) and final time (right) using the same temperature contour interval and unit wind vector. Initial Evolved Temp. Wind

Norwegian Meteorological Institute met.no Vertical energy distribution of the leading singular vector for the wind (black) and temperature field (red) at initial (left) and final time( right). Notice the difference in scaling of the horizontal axis. Initial Evolved wind Temp.

Norwegian Meteorological Institute met.no SV-spectrum Fine (0.2x0.2), coarse (0.4x0.4)

Norwegian Meteorological Institute met.no GLAMEPS_v1: distributed laboratory To set up a first phase suite tested in distributed mode (2008) based on experience from _v0 Run in hindcast mode use and use ECMWF for data exchange /storage Run ECMWF TEPS Store necessary LBC-data on accessible ECMWF disk for < 24 h Each partner downloads data and run a set of predefined LAM-EPS members Each partner to store selected results on ECMWF disk A set of probabilistic products made in batch-mode Each partner to download the entire produc suite Store on mars for future quality evaluation / validation /calibration NB: The success of GLAMEPS relies critically on dedicated partners. Most probably we need funding for supporting staff and storage at ECMWF.

Norwegian Meteorological Institute met.no ECMWF and GLAMEPS Operationally produce enhanced value intial/lateral boundary perturbations –“TEPS for Europe” as a Prediction Application Facility (PAF)? Data exchange central in RT operation –A selected set of data from TIGGE-list copied to ECMWF in RT each LAM-EPS. –At an agreed time, all partners can download the set of GLAMEPS members. Archiving –Archiving EPS and TEPS for use by GLAMEPS –Archiving GLAMEPS raw data and products Use software developed at ECMWF for –Selected probabilistic products, –Probabilistic verification and validation Calibrate and validate the entire GLAMEPS Develop and maintain –Prototype codes and scripts for downloading by partners, –Testing and quasi-operationalization in research mode, Further co-operate with ECMWF staff, scientifically and operationally.

Norwegian Meteorological Institute met.no Thank You!