TIGGE at ECMWF. David Richardson, Head, Meteorological Operations Section Slide 1. Slide 1

Transcription

1 TIGGE at ECMWF David Richardson, Head, Meteorological Operations Section Slide 1 Slide 1

2 ECMWF TIGGE archive The TIGGE database now contains five years of global EPS data Holds more than 520 terabytes (2.6 billion fields). There are around 1300 registered users of the TIGGE data portal Slide 2 Slide 2

3 ECMWF TIGGE archive 80 Number of active users Series Slide 3 Slide 3

4 ECMWF TIGGE archive Retrieved Data volumes (GBytes) Delivered Slide 4 Slide 4

5 ECMWF TIGGE archive - changes BoM, KMA stopped providing EPS to TIGGE archive KMA now testing sending higher resolution EPS New data: JMA 00 UTC control step 0 (as analysis) CMA analysis (06, 18)?? problem with database for 1 week needed to ask centres to resend. All gaps filled. Thanks to all who did that Access new batch access to TIGGE archive (python, perl) next generation data portal under development Planned GEOWOW developments: timeseries for small number of fields TIGGE-LAM, netcdf Slide 5 Slide 5

6 GEOWOW (GEOSS interoperability for Weather, Ocean and Water) is an EU-funded FP7 project that will begin in September GEOWOW will propose and validate a multi-disciplinary, distributed architectural model federating Earth Observation and other Earth Science data holdings and put this model forward as the European contribution to the Global Earth Observation System of Systems (GEOSS) Slide 6Common Infrastructure. Slide 6

7 GEOWOW The GEO Capacity Building Strategy focuses on three elements: human, institutional and infrastructure. The Weather component of the GEOWOW project will address all three by improving the access to THORPEX Interactive Grand Global Ensemble (TIGGE) data and developing and demonstrating products using this data in collaboration with users in developing countries, including providing education and training. GEOWOW will significantly enhance the accessibility of the TIGGE archive at ECMWF for the wider user community, in particular the ability to efficiently access long time series of forecast data at user-specified locations. Slide 7 Slide 7

8 GEOWOW 3 years: September 2011 August 2014 Co-ordinated by ESA Total funding from EU: 7 M euros For weather: 1.1 M euros ECMWF Met Office Météo-France Karlsruhe Institute of Technology Slide 8 Slide 8

9 The operational forecast system High resolution deterministic forecast: twice per day 16 km 91-level, to 10 days ahead Ensemble forecast (EPS): twice daily 51 members, 32/65 km 62-level, to 15 days ahead extended to 32 days once a week (Monthly forecast) Ocean waves: twice daily Global: 10 days ahead at 28 km Limited Area Wave (LAW): 5 days ahead at 10 km Ensemble: 15 days ahead at 55 km Seasonal forecast: once a month 41-members, 125 km 62 levels, to 7 months ahead a sub-set ensemble of 11 members is run for 13 months every quarter Slide 9 Slide 9

10 The operational ECMWF EPS The operational version of the EPS includes 51 forecasts with resolution: T L 639L62 (~32km, 62 levels) from day 0 to 10, T L 319L62 (~64km, 62 levels) from day 10 to 15 (32 at 00UTC on Thursdays). Initial uncertainties are simulated by adding to the unperturbed analyses a combination of T42L62 singular vectors, computed to optimize total energy growth over a 48h time interval (OTI), and perturbations generated using the new ECMWF Ensemble Data Assimilation (EDA) system. Model uncertainties are simulated by adding stochastic perturbations to the tendencies due to parameterized physical processes (SPPT scheme) and using a stochastic backscatter (SPBS) scheme. The EPS is run twice a day, at 00 and 12 UTC; the 00 UTC run is fully coupled to the HOPE ocean model after day 10. Slide 10 NH SH TR Definition of the perturbed ICs Products

11 Summary of EPS changes Initial perturbations: 2009 : Combination of Singular Vectors (SVs) optimised in two 48-hour windows: [t0, t0+48h] and [t0-48h, t0] 2011 : Combination of EDA perturbations (diff. between 10 EDA analyses) and SVs optimised in the [t0, t0+48h] window with reduced (50%) amplitude Simulation of model uncertainty: 2009 : SPPT with 1 spatial scale, white-noise time variab. with 6-hour time scale 2011 : SPPT with 3 spatial and time scales, red-noise variability in time (Markov chain), plus stochastic backscatter (SPBS) Slide 11

12 Simulation of initial uncertainty using EDA Since 22 June 2010, differences between 6h perturbed forecasts from the ECMWF Ensemble Data Assimilation system have been combined with singular vectors to generate the EPS initial perturbations The ten T L 399L91 EDA perturbed analyses are generated by randomly perturbing the observations and the SST, and by running the forecast model with the SPPT stochastic scheme. The random perturbations are defined by sampling a normal distribution with the observation error standard deviation U850 background error standard deviation Randomization method (ope, left) EDA (cy36r4, right) Since 18 May 2011, the EDA has been used to specify the background errors of the day in the highresolution 4D-Var Slide 12

13 Simulation of model uncertainty: multiscale SPPT Since Oct 1998, the EPS has included a stochastic scheme designed to simulate random model errors due to parameterized physical processes (SPPT). 500 km 6 h 1000 km 3 d 2000 km 30 d Since Nov 2010, the scheme includes a multi-scale pattern generator to account for parameterization errors on multiple spatial and temporal scales. Slide 13 (from M Leutbecher)

14 Simulation of model uncertainty: SPBS Since Nov 2010, a stochastic backscatter scheme (SPBS) is also used in the EPS: Rationale: a fraction of the dissipated energy is backscattered upscale and acts as streamfunction forcing for the resolved-scale flow (Shutts & Palmer 2004, Shutts 2005, Berner et al 2009) Streamfunction forcing is given by: Streamfunction forcing Backscatter ratio Total dissipation rate Pattern generator Recent improvements/updates include: Revised convective dissipation calculation Revised dissipation rate smoothing Slide 14 Changed pressure dependency of vertical correlations Option to force only part of the spectrum [reduces computational cost and avoids problems detected with small scale forcing]

15 Recent operational changes June 2010 (cycle 36r2) includes Changes to EPS initial-time perturbations 9 November 2010 (cycle 36r4) includes New cloud scheme New surface analysis schemes are introduced for snow and soil moisture EPS: revised model uncertainty; retuned initial perturbation amplitudes 18 May 2011 (cycle 37r2) includes Data assimilation changes Technical change: model-level data in GRIB-2 format Slide 15 Slide 15

16 EPS probability skill, RPSS, T850, N hem ECMWF EPS verification temperature 850hPa Ranked probability skill score \HG 0.8\(12mMA = 12 months moving average) N Hem Extratrop (lat 20.0 to 90.0, lon to 180.0) 00UTC,12UTC,beginning 0.8 T+72 12hMA T+168 T+120 T+72 T hMA T hMA Slide Monthly score and 12-month running mean (bold) of Ranked Probability Skill Score for EPS forecasts of T850 at day 3 (blue), 5 (red) and 7 (black) for N hem Slide 16

17 EPS probability skill, RPSS, T850, Europe ECMWF EPS verification temperature 850hPa Ranked probability skill score \HG 0.8\(12mMA = 12 months moving average) Europe (lat 35.0 to 75.0, lon to 42.5) 00UTC,12UTC,beginning 0.9 T+72 12hMA T+168 T+120 T+72 T hMA T hMA Slide Monthly score and 12-month running mean (bold) of Ranked Probability Skill Score for EPS forecasts of T850 at day 3 (blue), 5 (red) and 7 (black) for Europe Slide 17

18 EPS skill T850 NH 850hPa temperature Continuous ranked probability skill score N Hem Extratrop (lat 20.0 to 90.0, lon to 180.0) Date: UTC to UTC CMC UKMO NCEP ECMWF oper_an ti enfo prod Mean method: fair Population: 180,180,180,180,180,180,180,180,180,180,179,179,179,179, Slide Forecast Day Slide 18

19 m/s EPS spread/skill 850 wind tropics Ensemble Mean 850hPa wind speed Tropics (lat to 20.0, lon to 180.0) Date: UTC to UTC oper_an od enfo 0001 Mean method: standard 3 spread djf2010 rmse djf2010 spread djf2011 rmse djf2011 spread djf2009 rmse djf Slide Forecast Day Slide 19

20 EPS skill 850 wind tropics 850hPa wind speed Continuous ranked probability skill score Tropics (lat to 20.0, lon to 180.0) Date: UTC to UTC oper_an od enfo 0001 Mean method: standard ,12utc ,12utc ,12utc ,12utc ,12utc Slide Forecast Day Slide 20

21 EPS probabilistic precipitation skill CRPSS for 24-h Precipitation, ExTrop, ECMWF, h Precip, to , ExTrop Lead time (days) CRPSS Year 0.00 ECMWF ( 0.178) JMA ( 0.066) UKMO ( 0.037) NCEP (-0.008) Lead time (days) Slide 21 Slide 21

22 Product Development New parameters from deterministic forecast Height of lowest cloud base Height of 0 C level New parameters from EPS L, M, H cloud cover Slide 22 Slide 22

23 New EPS clustering New clustering operational since 16 November Graphical products on the ECMWF website at: Cluster fields available via dissemination The disseminated products are based on the day 5-7 time range (time steps 120, 132, 144, 156 and 168) Old clusters will be switched off later this year Slide 23 Slide 23

24 New EPS clustering Tuesday 7 June UTC ECMWF EPS Cluster scenario hpa Geopotential Reference step t Domain 75/340/30/40 Cont. in cluster=2 Det. in cluster=2 forecast t+120 VT:Sunday 12 June UTC Cluster: 1(of 3), population: 22, repres. member: 21 forecast t+144 VT:Monday 13 June UTC Cluster: 1(of 3), population: 22, repres. member: 21 forecast t+168 VT:Tuesday 14 June UTC Cluster: 1(of 3), population: 22, repres. member: forecast t+120 VT:Sunday 12 June UTC Cluster: 2(of 3), population: 17, repres. member: 0 forecast t+144 VT:Monday 13 June UTC Cluster: 2(of 3), population: 17, repres. member: 0 forecast t+168 VT:Tuesday 14 June UTC Cluster: 2(of 3), population: 17, repres. member: forecast t+120 VT:Sunday 12 June UTC Cluster: 3(of 3), population: 12, repres. member: 4 forecast t+144 VT:Monday 13 June UTC Cluster: 3(of 3), population: 12, repres. member: 4 forecast t+168 VT:Tuesday 14 June UTC Cluster: 3(of 3), population: 12, repres. member: Slide Slide 24

25 Regime transitions metres n. of clusters DEC 2010 JAN 2011 FEB MAR APR MAY EPS forecast of large-scale weather regimes. Daily time series for November 2010 to April 2011 of the number of EPS clusters and observed climatological regimes (coloured Slide 25 circles). The climatological regime associated with each cluster is indicated by the colour of the bar: blue - positive NAO pattern, green - negative NAO, red blocking, violet - Atlantic ridge. The EPS clusters are computed over forecast days 5-7. Slide 25

26 Climatological regimes Slide 26 Slide 26

27 Tropical cyclones Tracking of tropical cyclones developing during forecast (medium-range and monthly products on web) Operational tracks deterministic up to 1 hour earlier Under test: replacement of operational tracker (will include tracks beyond 5 days) Slide 27 Slide 27

28 Tropical cyclones Typhoon Songda May UTC Probability that 04W will pass within 120km radius during the next 120 hours tracks: black=oper, green=ctrl, blue=eps numbers: observed positions at t+..h 120 E 140 E N 20 N N 10 N E 140 E 5 Operational tracker Slide 28 Slide 28

29 Tropical cyclones Date ECMWF Probability that 04W will pass within 120 km radius during the next 240 hours tracks: solid=oper; dot=ctrl > 90 % Probability (%) of Tropical Cyclone Intensity falling in each category TD[up to 16] TS [17-32] HR1[33-42] HR2 [43-48] HR3 [> 48 m/s] 50 N 40 N 30 N 20 N 100 E 120 E 140 E 160 E 50 N 40 N 30 N 20 N Fri 20 Sat 21 Sun 22 Mon 23 Tue 24 Wed 25 Thu 26 Fri 27 Sat 28 Sun 29 Mon 30 May m Wind Speed (m/s) 10 N 0 N 100 E 120 E 140 E 160 E List of ensemble members numbers forecast Tropical Cyclone Intensity category in colours: TD[up to 16] TS[17-32] HR1 [33-42] HR2 [43-48] HR3 [ >48 m/s] +024 h : hr ct h : hr ct h : hr ct h : hr ct h : hr ct h : hr ct h : hr ct h : hr ct h : hr ct h : hr ct N 0 N Fri 20 Sat 21 Sun 22 Mon 23 Tue 24 Wed 25 Thu 26 Fri 27 Sat 28 Sun 29 Mon 30 May 2011 Mean Sea Level Pressure in Tropical Cyclone Centre (hpa) Slide 29 Fri 20 Sat 21 Sun 22 Mon 23 Tue 24 Wed 25 Thu 26 Fri 27 Sat 28 Sun 29 Mon 30 May 2011 Slide 29

30 0-24h 24-48h 48-72h 72-96h h h h Track plumes for the TC are plotted for the EPS members, T1279 and Control. Each colour represents 24h forecast interval (EPS) Slide 30 Slide 30

31 Product Development Extra-tropical cyclone products (web products) Surface weather parameters (new web plots) Extension of EFI Slide 31 Slide 31

32 eccharts new interactive web plots Slide 32 Slide 32