Development of the ECMWF probabilistic system

Transcription

1 Development of the ECMWF probabilistic system Roberto Buizza, Martin Leutbecher and Tim Palmer The following contributions are acknowledged: Jean Bidlot, Manuel Fuentes, Mats Hamrud, Alfred Hofstadler, Graham Holt, Martin Miller, Mark Rodwell, Adrian Simmons, Frederic Vitart and Nils Wedi for their help in the development of VAREPS Glenn Shutts (The Met Office, UK) for his contribution to the development of CASB Tiziana Paccagnella (SMR-ARPA, Italy) for her input on LEPS Jutta Thielen (JRC, Italy) for her input on EFAS Luigi Cavaleri (ISMAR-CNR, Italy) for his input on severe weather prediction Bill Bourke (BMRC Australia), Dehui Chen (CMA China), Antonio Marcos Mendonca (CPTEC Brazil), Mike Sestak (FNMOC US), Masayuki Kyouda (JMA Japan), Hee- Dong Yoo (KMA Korea), Chantal Cote (MSC Canada), Gerard Pellerin (MSC Canada), Peter Houtekamer (MSC Canada) and Zoltan Toth (NCEP US) for their contribution on TIGGE Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 1

2 The four topics discussed in this contribution Verification Development The future Applications Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 2

3 The four key messages of this talk Verification. The ECMWF Ensemble Prediction System (EPS) has been continuously improving, with probabilistic products gaining ~2 d/de in predictability. Development. Recent changes (definition of tropical areas, Gaussian sampling, use of more SVs) have improved EPS forecasts. Future changes in the ensemble configuration (VAREPS), and in the simulation of initial and model uncertainties are expected to improve the accuracy of EPS forecasts. VAREPS will extend the EPS to 15 days, and will be the first step towards a seamless probabilistic prediction system. Applications. The establishment of stronger links between global and limited-area ensemble systems, and between meteorological and hydrological forecasting systems have been helping the development of valuable forecasting tools. The future. Collaborations between ECMWF, its users and the whole international community will help addressing open issues in probabilistic forecasting and designing future probabilistic analysis and forecasting systems. Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 3

4 Outline 1. Verification: performance of the ECMWF EPS from May 1994 to date 2. Development: Changes in the simulation of initial and model uncertainties The forthcoming VAriable Resolution EPS (VAREPS) 3. Applications: Link with limited-area ensemble systems Severe weather: acqua alta prediction in Venice in 1966 Hydrological ensemble prediction 4. The future: Open questions TIGGE Towards a fully probabilistic analysis and prediction system Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 4

5 1. The ECMWF Ensemble Prediction System The Ensemble Prediction System (EPS) consists of day forecasts run at resolution T L 255L40 (~80km, 40 levels) [1,5,7,8,13,11,15]. NH SH TR The EPS is run twice a-day, at 00 and 12 UTC. Initial uncertainties are simulated by perturbing the unperturbed analyses with a combination of T42L40 singular vectors, computed to optimize total energy growth over a 48h time interval (OTI). Definition of the perturbed ICs Products Model uncertainties are simulated by adding stochastic perturbations to the tendencies due to parameterized physical processes. Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 5

6 Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) The ECMWF Ensemble Prediction System Each ensemble member evolution is given by the time integration of perturbed model equations starting from perturbed initial conditions The model tendency perturbation is defined at each grid point by where r(x) is a random number. = + + = T t j j j j j dt t e P t e P t e A T e 0 )], ( ), ( ), ( [ ) ( δ ) ( ) ( ) ( 0 d de d e d e j j + = = + + = area N k k k j k k j j SV d d SV d SV d de 1,, )] 2 2, (,0) ( [ ) ( β α ),, ( ), ( ),, ( p P r p P j j j φ λ φ λ φ λ δ =

7 1. Since May 94 the EPS has changed twelve times Since Dec 1992, 42 model cycles (which included changes in the ECMWF model and DA system) were implemented, and the EPS configuration was modified 12 times. Date Description Singular Vectors's characteristics Forecast characteristics HRES VRES OTI Target area EVO SVs sampl HRES VRES Tend # Mod Imp Dec 1992 Oper Impl T21 L19 36h globe NO simm T63 L19 10d 33 NO Feb 1993 SV LPO " " " NHx " " " " " " " Aug 1994 SV OTI " " 48h " " " " " " " " Mar 1995 SV hor resol T42 " " " " " " " " " " Mar 1996 NH+SH SV " " " (NH+SH)x " " " " " " " Dec 1996 resol/mem " L31 " " " " TL159 L31 " 51 " Mar 1998 EVO SV " " " " YES " TL159 L31 " " " Oct 1998 Stoch Ph " " " " " " " " " " YES Oct 1999 ver resol " L40 " " " " " L40 " " " Nov 2000 FC hor resol " " " " " " TL255 " " " " Jan 2002 TC SVs " " " (NH+SH)x+TC " " " " " " " Sep 2004 sampling T42 L40 48h (NH+SH)x+TC YES Gauss TL255 L40 10d 51 YES Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 7

8 1. The EPS performance has been continuously increasing EPS upgrades have been leading to an increased accuracy in the prediction of the large-scale flow, as shown, e.g., by the time evolution of the RPPS of the 500 hpa geopotential height over the NH. RPSS RPSS - NH Z500 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Date d+3 d+5 d+7 Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 8

9 1. The EPS performance has been continuously increasing Improvements over Europe have been slightly smaller, as can be detected by comparing the time evolution of the RPPS for 500 hpa geopotential height predictions over Europe (left) and NH (previous slide). RPSS RPSS - EU Z500 Jan-94 Jan-95 Jan-96 Jan-97 Jan-98 Jan-99 Jan-00 Jan-01 Jan-02 Jan-03 Jan-04 Jan-05 Date d+3 d+5 d+7 Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 9

10 1. Summary: predictability gains for NH Z500 Considering 500 hpa geopotential height predictions over the NH at d+5 and d+7, diagnostic studies have indicated that: the EPS control has improved by more than 1d/de the EPS ens-mean has improved by ~1.5d/de Days Predictability gains (linear trend estimates) - NH Z500 d+5 d+7 the EPS probabilistic products have improved by more than 2d/de CON ACC EM ACC CON TS[f>c] CON ROCA[f>C] EPS ROCA[f>c EPS RPSS EPS BSS[f>c] EPS BSS[f>(c+s)] ESP BSS[f<(c-s)] Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 10

11 1. Summary: predictability gains for Europe Z500 Results are slightly less positive for Europe, for which diagnostic studies have indicated that: the EPS control has improved by ~0.75d/de the EPS ens-mean has improved by ~1d/de the EPS probabilistic products have improved by ~1.5d/de Days Predictability gains (linear trend estimates) - EU Z500 CON ACC d+5 d+7 EM ACC CON TS[f>c] CON ROCA[f>C] EPS ROCA[f>c EPS RPSS EPS BSS[f>c] EPS BSS[f>(c+s)] ESP BSS[f<(c-s)] Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 11

12 1. Comparison of the ECMWF, MSC and NCEP EPSs (JJA02) Recent studies [2,9] have shown that, accordingly to many accuracy measures, the ECMWF EPS can be considered the most accurate singlemodel ensemble system. This is shown, e.g., by the comparison of the EV* of 10-member ensembles based on the ECMWF, MSC (Meteorological Service of Canada) and NCEP (National Centers for Environmental Predictions) EPSs [9] (Z500 over NH). * EV, the potential economic value, is the reduction of the mean expenses with respect to the reduction that can be achieved by using a perfect forecast [4,16]. (Source: Buizza et al [9]) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 12

13 1. Comparison of the ECMWF, MSC and NCEP EPSs (JJA02) The ECMWF leading performance [9], estimated to be equivalent to a gain of ~1 day of predictability, has been linked to: A better analysis A better model A better estimation of the PDF of forecast states. This latest point can be seen, e.g., by comparing the ensemble spread and the ensemble-mean forecast error of 10-member ensembles based on the NCEP, MSC and ECMWF EPSs (Z500 over NH). (Source: Buizza et al [9]) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 13

14 1. ECMWF, MSC and NCEP EPSs lack medium-range spread This comparison indicated that in the ECMWF-EPS: initially, the EPS perturbations are growing too quickly compared to error growth in the medium-range, the EPS is under-dispersive <error> <spread> These two problems are more evident in the NCEP and the MSC EPSs due to the slower growth of their initial perturbations, and to a combination of lower resolution and different simulation of model uncertainties. Day 1 Day 5 Day 10 Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 14

15 Outline 1. Verification: performance of the ECMWF EPS from May 1994 to date 2. Development: Changes in the simulation of initial and model uncertainties The forthcoming VAriable Resolution EPS (VAREPS) 3. Applications: Link with limited-area ensemble systems Severe weather: acqua alta prediction in Venice in 1966 Hydrological ensemble prediction 4. The future: Open questions TIGGE Towards a fully probabilistic analysis and prediction system Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 15

16 2. Initial uncertainties: definition of the TC-SVs target areas The EPS before September 2004 had one weakness linked to the definition of the target areas used to compute the tropical singular vectors [1,15]: TR-SVs were computed inside areas with northern boundary with λ 25 N: this was causing an artificial ensemble-spread reduction when tropical cyclones were crossing 25 N TR-SVs were computed only if WMO cl-2 TC were detected between 25 S-25 N Too few tropical areas (up to 4) were considered Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 16

17 2. TR-SVs target areas: the Sep 04 change On 28 Sep 04, a change was introduced to address this issue. In the new system: Target areas are computed considering TCtrack predictions Areas are allowed to extend north of 30ºN Up to 6 areas can now be targeted Tropical depression (WMO cl 1) detected between 40 S- 40 N are targeted SVs are computed using a new ortho-normalization procedure Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 17

18 2. TR-SVs target areas: impact of the Sep 04 change Results based on 44 cases (from 3 Aug to 15 Sep 2004) indicate that the implemented changes in the computation of the tropical areas has a positive impact on the reliability diagram of strike probability. Reliability diagram for strike probabilities Old CY28R2 EPS New CY28R3 EPS Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 18

19 2. The Sep 04 and the forthcoming changes in SVs sampling Before September 2004, the distributions of the α j,k and β j,k coefficients was defined by the rotation algorithm and was not Gaussian: de j N SV ( d) = [ α j, k SVk ( d,0) + β j, k SVk ( d 2, + 2d)] area k = 1 Since Sept 2004, the distributions of the α j,k and β j,k coefficients that define the EPS initial perturbations have been set to be Gaussian [11]. It is worth pointing out that since September 2004, the 50 EPS initial perturbations have not been any more symmetric, but symmetry will be re-introduced in the next model cycle (under testing). Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 19

20 2. Gaussian sampling with 50 instead of 25 singular vectors The adoption of Gaussian sampling makes it simpler to set the number of SVs, N SV, independently from the ensemble size, N ENS. The next model cycle will also see an increase in N SV from 25 to 50. N SV has been increased to sample more dynamically important phase space directions. Sensitivity experiments (based on 27 cases) have indicated that the N SV increase is increasing the skill of probabilistic scores. Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 20

21 2. Initial uncertainties Ongoing research Why shall the definition of the EPS initial perturbations be changed? In the current EPS: SVs are computed at T42L40 resolution over a 48h time optimization interval Extra-tropical SVs are still computed with a tangent dry physics [3] Tropical SVs are computed with a tangent moist physics [1,12,15], but with the state vector still defined in terms of [V,D,T,ln(sp)] only (i.e. without humidity) To better capture perturbations growth, especially in cases of intense, small-scale cyclonic developments, it is thought that a tangent moist physics should be used. Recent results [10] have indicated that when moist processes are considered, a T63 truncation would be better than a T42, and a 24h OTI is more suitable than the 48h OTI used for dry SVs. The use of 24h, T L 95 SVs computed with the new moist tangent physics is under testing. Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 21

22 2. Impact of moist processes on T63L31-24h SVs 27 Dec 99 00Z: French storm Martin. The top panels [10] show a weighted geographical distribution of the first 10 T63L31-24h dry SVs at initial and final time (ci x50 at final time). The bottom panels show the weighted distribution of the first 10 T63L31-24h full-physics SVs, superimposed on the basic state total column water content. In the moist experiment, SVs evolve along the upstream side of the tongue of moisture into the storm region. (Source: Coutinho et al [10]) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 22

23 2. Impact of moist physics on T63L31-24h SVs 2 Aug 97 00Z; Storm over Ireland. The two top panels [10] show a weighted geographical distribution of the first 10 T63L31-24h dry SVs targeted to grow in [30-90N; 30W- 40E] at initial and final time ; ci x50 at final time). The two bottom panels show the weighted distribution of the first 10 T63L31-24h full-physics targeted SVs, superimposed on the basic state total column water content. In the moist experiment, SVs evolve along the tongue of moisture into the storm region. (Source: Coutinho et al [10]) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 23

24 2. Model imperfections Should the approach be changed? In the current EPS: Model imperfections are simulated using stochastic physics, a simple scheme designed to simulate the random errors in parameterized forcing that are coherent among the different parameterization schemes (moist-processes, turbulence, ). Coherence with respect to parameterization schemes has been achieved by applying the stochastic forcing on total tendencies. Space and time coherence has been obtained by imposing space-time correlation on the random numbers. The scheme has been shown [14] to have a positive impact on the EPS, especially on the accuracy of probabilistic precipitation prediction. But diagnostics and recent studies [17] have indicated that the scheme has some weaknesses: In the lower levels, it seems to generate too large spread and too intense rainfall In the upper levels its impact on the ensemble spread is rather limited (~5%) Random numbers have a very crude spatial and temporal correlations It is controlled by parameters that have been tuned in a rather ad-hoc manner Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 24

25 2. Cellular Automaton Stochastic Backscatter Scheme The new Cellular Automaton Stochastic Backscatter Scheme [17] (CASBS): CASBS is based on the physical argument that kinetic energy sources that counteract energy drain occurring in the near-grid scale can improve the performance of numerical models. Kinetic energy is backscattered by introducing vorticity perturbations into the flow with a magnitude proportional to the square root of the total dissipation rate. The spatial form of vorticity perturbations is derived from an exotic pattern generator (cellular automaton) that crudely represents the spatial/temporal correlations of the atmospheric meso-scale T L 159L40 EPS experiments for 10 cases have indicated that: CASBS reduces the excessive heavy rainfall events It is more effective at generating model spread It generates a better meso-scale energy spectrum Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 25

26 2. CASBS positive impact on heavy precipitation events Experiments based on T L 159L40 EPS forecasts for 10 cases indicate that: The operational stochastic physics scheme (dashed blue) generates too many cases of heavy precipitation CASBS (dash green) performs more in agreement with observed statistics (black solid) (Source: Shutts [17]) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 26

27 2. CASBS positive impact on EPS spread Experiments based on T L 159L40 EPS forecasts for 10 cases indicate that: CASBS (red solid) induces more divergence among the ensemble members than the operational scheme (blue dashed) CASBS ensemblespread around the control is closer to the average error of the control forecast (black chain-dashed) New CASBS scheme Operational EPS Initial perturbation only Control forecast Error (Source: Shutts [17]) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 27

28 2. VAREPS Rationale: T L 399 resolution up to 15 days is unaffordable, and the benefits of extending the EPS to day 15 outweighs the disadvantages of loosing resolution Predictability of small scales is lost relatively earlier in the forecast range. Therefore, while forecasts benefit from a resolution increase in the early forecast range, they do not suffer so much from a resolution reduction in the long range. VAREPS planned configuration: D0-7: T L 399L40, dt=1800s D7-15: T L 255L40, dt=2700s VAriable Resolution EPS T0 T1 T2 Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 28

29 2. Expected impact of EPS upgrade: π(tp 10mm/12h) Results based on 30 cases, 51-member ensembles (CY28R3) indicate a positive impact of EPS upgrade. Considering ROCA as accuracy measure for probabilistic predictions of TP in excess of 10mm/12h over NH, results indicate a gain of ~12h during the first 7 forecast days. ROCA(X) ROCA(TP>10mm/12h) - NH (30 cases) T255 T399 VAREPS_D forecast day Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 29

30 2. Expected impact of EPS upgrade: π(z500 cli) Results based on 30 cases, 51-member ensembles (CY28R3) indicate a positive impact of EPS upgrade. Considering ROCA as accuracy measure for probabilistic predictions of positive Z500 anomalies over NH, results indicate a small gain up to forecast day 11 (~6h at forecast day 7). ROCA(X) ROCA(Z500>CLI) - NH (30 cases) T255 T399 VAREPS_D forecast day Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 30

31 Outline 1. Verification: performance of the ECMWF EPS from May 1994 to date 2. Development: Changes in the simulation of initial and model uncertainties The forthcoming VAriable Resolution EPS (VAREPS) 3. Applications: Link with limited-area ensemble systems Severe weather: acqua alta prediction in Venice in 1966 Hydrological ensemble prediction 4. The future: Open questions TIGGE Towards a fully probabilistic analysis and prediction system Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 31

32 3. Link with limited-area ensemble systems Over Europe, there are 4 operational Limited-area EPSs (SRNWP-PEPS, COSMO- LEPS, NORLAMEPS and PEACE) that produce daily 81 forecasts with horizontal resolution ranging from 7 to 120 km, and with forecast length ranging from 30 to 120 hours. 8 further centres (Met Office, INM, DMI, HMS, MeteoSwiss, SAR, PIED-SE) are developing and testing LEPSs. Studies have shown that compared to global EPSs, limited-area EPSs are better able to predict small-scale, local phenomena. This figure shows the t+96h forecast of the probability of total precipitation exceeding 20mm/d given by the EPS (left) and the COSMO-LEPS system for 29 Aug 2003 (Ticino flood). Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 32

33 3. Severe weather: acqua alta in Venice (1966) In 1966, November 4, Venice suffered the worst flood ever recorded. On the same day the Arno river flooded Firenze with disastrous consequences for art. At the time there was little anticipation of the severity of the storm. A research project established by ISMAR-CNR (Venezia, Italy) and including ECMWF and ISAC-CNR (Bologna, Italy) is addressing the following crucial question: should this happen today, what would we be able to do? Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 33

34 3. Hydrological applications: EFAS and HEPEX The European Flood Alert System (EFAS) and the Hydrological Ensemble Prediction Experiment (HEPEX) are two examples of initiatives promoting and testing ensemble methods in hydrology. HEPEX 2 nd WS: July 05, NOAA ( Observations Data assimilation Global M ~al modelling Local M ~al modelling Local H ~al modelling Quality control Weather data Hydrological data Observation perturbation methods 3/4D-Var Data Assimilation Systems Ensemble Data Assimilation Global perturbation methods Model design Model resolution Ensemble Size Nested Limited Area Models Local perturbation methods Postprocessing methods Combination methods Rainfall, runoff models Verification methods Users value Verification data Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 34

35 3. EFAS: the European Flood Alert System EFAS is a forecasting tool designed to give early-warnings for European rivers with catchments in excess of 2000 km. A pre-operational prototype is under testing at the Joint Research Center (JRC, Ispra). The system uses meteorological inputs from DWD (forecasts up to 7 days), ECMWF (high-resolution and ensemble forecasts up to 10 days) and aims to provide single and probabilistic predictions. This figure shows the prediction of the risk of flooding from 28 Oct 2004 for the subsequent 10-days computed using the ECMWF and DWD high-resolution forecasts (left) and the EPS (right). Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 35

36 Outline 1. Verification: performance of the ECMWF EPS from May 1994 to date 2. Development: Changes in the simulation of initial and model uncertainties The forthcoming VAriable Resolution EPS (VAREPS) 3. Applications: Link with limited-area ensemble systems Severe weather: acqua alta prediction in Venice in 1966 Hydrological ensemble prediction 4. The future: Open questions TIGGE Towards a fully probabilistic analysis and prediction system Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 36

37 4. Open questions in ensemble prediction 2004 and 2005 have seen the completion of studies ([2], [9]) that analyzed the performance of 4 of the leading operational global ensemble prediction systems (the BMRC-, ECMWF-, NCEP- and MSC-EPSs). During this period, 8 workshops were held to discuss ensemble-related issues: The 1 st HEPEX WS (ECMWF, UK, March 2004) The WS on Ensemble Methods (The Met Office, UK, Oct 2004) The 2 nd WS of the NAEFS (NCEP, US, Nov 2004) The 2 nd WS on EFAS (JRC, Italy, Nov 2004) The 1 st TIGGE WS (ECMWF, UK, March 2005) The SRNWP workshop on (ARPA-SMR Bologna, Italy, Apr 2005) The WS on representation of sub-grid-scale processes (ECMWF, UK, June 2005) The 2 nd HEPEX WS (NOAA, US, July 2005) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 37

38 4. Open questions in ensemble prediction This is a brief summary of relevant issues identified in these workshops onto which research is focussing: In the short-range, methodology and design affects ensemble characteristics What is the importance of the initial perturbation method? What is the importance of the method used to simulate model uncertainty? In the medium-range, model and data-assimilation qualities matter The performance of EPSs strongly depends on the quality of the data assimilation system used to create the unperturbed initial conditions, and the numerical model used to generate the forecasts [8] A sample-all approach should be followed A successful ensemble prediction system should simulate the effect of both initial and model related uncertainties on forecast errors [8] Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 38

39 4. Open questions ensemble prediction A multi-model multi-analysis system may be necessary In the ECMWF, MSC and NCEP EPSs, the spread is still insufficient to systematically capture reality, suggesting that none of them is capable alone to simulate all sources of forecast uncertainties [8] Increasing ensemble size beyond 50 matters less than increasing resolution Today, 351 members are run daily with resolution from T L 119 to T L 255. By sharing production costs, ~50 members could be run at T L 399 (~60km) Would such an approach lead to a more skilful ensemble system? Communication with (end-) users While the use of ensemble products by forecasters has been increasing during the past few years, its use by the end-users is still rather limited. Which is the best way to communicate uncertainty to end-users? Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 39

40 4. TIGGE TIGGE (the THORPEX Interactive Grand Global Ensemble) is a framework for international collaboration in development and testing of ensemble prediction systems. TIGGE could lead to: An enhanced international collaboration between operational centres and universities A deeper understanding of the contribution of observation, initial and model uncertainties to forecast error, and the design of more valuable ensemble systems The developments of new methods of combining ensembles from different sources and of correcting for systematic errors (biases, spread over-/underestimation) Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 40

41 4. TIGGE will benefit from existing ensemble systems Globally, there are 9 operational Global Ensemble Prediction Systems (BMRC, CMA, CPTEC, ECMWF, FNMOC, JMA, KMA, MSC and NCEP) that produce daily 351 forecasts with horizontal resolution ranging from T62 to TL255 (~80km), and with forecast length ranging from 8 to 16 days. 3 further centres (MetOffice, NCMRWF, SAWS) are developing and testing global ensemble systems. Over Europe, there are 4 operational Limited-area EPSs (SRNWP-PEPS, COSMO- LEPS, NORLAMEPS and PEACE) that produce daily 81 forecasts with horizontal resolution ranging from 7 to 25 km, and with forecast length ranging from 30 to 120 hours. 8 further centres (NOR, MetOffice, INM, DMI, HMS, MeteoSwiss, SAR, PIED- SE) are developing and testing LEPSs. Over North-America, there is 1 operational Limited-area EPSs (NCEP-SREF) that produces daily 30 forecasts with horizontal resolution of 32 km, and a 63-hour forecast length. Another centre (MSC) is testing a LEPS. Over Australia, BMRC is testing a 16-member, 0.5 degree resolution, 72-hour LEPS. Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 41

42 4. Ex: π Z500 (00,120h): BMRC, CPTEC, ECMWF, FNMOC, NCEP Europe: 120h forecast probability of T850<0 degrees. What is the PR(T850<0) in Firenze? BMRC gives 0%, the others more than 20% probability*. BMRC CPTEC ECMWF * This is just one case: probability forecasts should be verified on a large dataset. FNMOC NCEP EC-AN Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 42

43 4. Ex: π Z500 (12,120h): BMRC, ECMWF, JMA, KMA, NCEP Europe: 120h forecast probability of T850<0 degrees. What is the PR(T850<0) in Tunisia? BMRC gives a zero probability.* BMRC ECMWF JMA * This is just one case: probability forecasts should be verified on a large dataset. KMA NCEP EC-AN Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 43

44 4. TIGGE could lead to a MUMMA-GEPS TIGGE could lead to a Multi-Model, Multi-Analysis Global Ensemble Prediction System (MUMMA-GEPS), with N production centers (yellow stars) and few data-hubs (red) connected by high-speed, high-capacity communication lines. Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 44

45 4. EDA: towards a probabilistic analysis & forecast system? Ensemble Data Assimilation [6] may be used in the future to generate the EPS initial perturbations. A future EPS configuration could include: N-member EDA N*M member EDA-SV EPS, T L 399(d0:7)+T L 255(d7:15) ICs from each perturbed members and/or the EDA ensemble-mean EDA perturbed members EDA ensemble-mean High-resolution forecast Low resolution forecast Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 45

46 Acknowledgements The success of the ECMWF EPS is the result of the continuous work of ECMWF staff, consultants and visitors who had continuously improved the ECMWF model, analysis, diagnostic and technical systems, and of very successful collaborations with its member states and other international institutions. The work of all contributors, and in particular of former ECMWF staff (Jan Barkmeijer, Franco Molteni, Robert Mureau, Anders Persson, Otto Pessonen, Thomas Petroliagis, David Richardson, Stefano Tibaldi), visitors and consultants (Bill Bourke, Mariane Coutinho, Martin Ehrendorfer, Ron Gelaro, Isla Gilmour, Dennis Hartmann, Andrea Montani, Steve Mullen, Kamal Puri, Carolyn Reynolds, Joe Tribbia) who worked directly on the ECMWF Ensemble Prediction System is acknowledged (I hope that the list of names is complete: please forgive me if this is not the case). Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 46

47 References [1] Barkmeijer, J., Buizza, R., Palmer, T. N., Puri, K., & Mahfouf, J.-F., 2001: Tropical singular vectors computed with linearized diabatic physics. Q. J. R. Meteorol. Soc., 127, [2] Bourke, W., Buizza, R., & Naughton, M., 2004: Performance of the ECMWF and the BoM Ensemble Systems in the Southern Hemisphere. Mon. Wea. Rev., 132, [3] Buizza, R., 1994: Sensitivity of Optimal Unstable Structures. Q. J. R. Meteorol. Soc., 120, [4] Buizza, R., 2001: Accuracy and economic value of categorical and probabilistic forecasts of of discrete events. Mon. Wea. Rev., 129, [5] Buizza, R., & Palmer, T. N., 1995: The singular vector structure of the atmospheric general general circulation. J. Atmos. Sci., 52, [6] Buizza, R., & Palmer, T. N., 1999: Ensemble Data Assimilation. Proceedings of the AMS 13 th 13 th Conference on Numerical Weather Prediction, Sep 1999, published by AMS, [7] Buizza, R., Miller, M., & Palmer, T. N., 1999: Stochastic representation of model uncertainties uncertainties in the ECMWF Ensemble Prediction System. Q. J. R. Meteorol. Soc., 125, [8] Buizza, R., Richardson, D. S., & Palmer, T. N., 2003: Benefits of increased resolution in the the ECMWF ensemble system and comparison with poor-man's ensembles. Q. J. R. Meteorol. Meteorol. Soc.,129, Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 47

48 References (cont.) [9] Buizza, R., Houtekamer, P. L., Toth, Z., Pellerin, G., Wei, M., & Zhu, Y., 2005: A comparison of comparison of the ECMWF, MSC and NCEP Global Ensemble Prediction Systems. Mon. Wea. Wea. Rev., 133, [10] Coutinho, M. M., Hoskins, B. J., & Buizza, R., 2004: The influence of physical processes on on extra-tropical singular vectors. J. Atmos. Sci., 61, [11] Ehrendorfer, M., & Beck, A., 2003: Singular vector-based multivariate sampling in ensemble ensemble prediction ECMWF Technical Memorandum n. 416 (available from ECMWF). [12] Mahfouf, J.-F., 1999: Influence of physical processes on the tangent linear approximation. approximation. Tellus, 51A, [13] Molteni, F., Buizza, R., Palmer, T. N., & Petroliagis, T., 1996: The new ECMWF ensemble ensemble prediction system: methodology and validation. Q. J. R. Meteorol. Soc., 122, [14] Mullen, S., & Buizza, R., 2001: Quantitative precipitation forecasts over the United States by States by the ECMWF Ensemble Prediction System. Mon. Wea. Rev.,129, [15] Puri, K., Barkmeijer, J., & Palmer, T. N., 2001: Ensemble prediction of tropical cyclones using using targeted diabatic singular vectors. Q. J. R. Meteorol. Soc., 127, [16] Richardson, D. S., 2000: Skill and relative economic value of the ECMWF Ensemble Prediction System. Q. J. R. Meteorol. Soc., 127, [17] Shutts, G., 2004: A stochastic kinetic energy backscatter algorithm for use in ensemble prediction systems. ECMWF Technical Memorandum n. 449 (available from ECMWF). Buizza et al: Development of the ECMWF probabilistic system (User Meeting, June 2005) - 48