FE 1 Koppert et al., DWD4COPS COPS – DWD Contributions Hans-Joachim Koppert, Michael Baldauf, Michael Denhard, Werner Wergen Deutscher Wetterdienst
FE 1 Koppert et al., DWD4COPS Overview nCOSMO-K Goals Case Studies & Verification Status nNinJo nPEPS Background Implementation Micro PEPS nData-Assimilation Issues
FE 1 Koppert et al., DWD4COPS COSMO-K (formerly known as LMK Lokal-Modell-Kürzestfrist) Goals Development of a model-based NWP system for very short range (‘Kürzestfrist’) forecasts (18 h) of severe weather events on the meso- scale, especially those related to deep moist convection (super- and multi-cell thunderstorms, squall-lines, MCCs, rainbands,...) interactions with fine-scale topography (severe downslope winds, Föhn-storms, flash floodings, fog,...) GME (global) x = 40 km * 40 GP t = 133 sec. T = 7 days COSMO-E (Europe) x = 7 km 665 * 657 * 40 GP t = 40 sec. T= 78 h COSMO-K (regional) x = 2.8 km 421 * 461 * 50 GP t = 25 sec. T = 18 h 8 forecasts / day
FE 1 Koppert et al., DWD4COPS Convectively enhanced frontal precipitation, , 18 UTC Obs.: up to 20 mm/12 h
FE 1 Koppert et al., DWD4COPS LAF-ensemble, 1h-precip.-sum, target time: , 18 UTC h h h h h h
FE 1 Koppert et al., DWD4COPS Convectively enhanced frontal precipitation , h (from 0 UTC & 12 UTC-run)
FE 1 Koppert et al., DWD4COPS Problem: missed convection initiation in LMK other examples: July 2006
FE 1 Koppert et al., DWD4COPS Sept Oct Synop-Verification RMSE of wind speed |v| 10m LMK LME 0 UTC-runs 12 UTC-runs U. Damrath
FE 1 Koppert et al., DWD4COPS Synop-Verification of pre-operational LMK Gusts and Precipitation, Oct. 2006, 12 UTC-runs LMK LME ETS TSS Gusts Precipit ation Precipitation: July ‘05: TSS generally higher Sept. ‘06: LMK higher TSS due to LHN, Oct. + Nov. ‘06: LMK mostly higher TSS (FBI ~ equal) Dec. ‘06: LMK smaller TSS (no LHN?) Gusts: ETS generally higher (but sometimes also higher FBI)
FE 1 Koppert et al., DWD4COPS nCOSMO-K in pre-operational use since n18 h- (21 h-) forecasts are simulated every 3 h (LAF- ensemble) nexplicit simulation of deep moist convection with its life cycle generates good predicitions of precipitation in the case of synoptic forced events (e.g. lines of thunderstorms) ndynamical effects better represented due to higher resolution strong downslope winds lee waves (e.g. improved glider forecasts) nradar observations of the DWD-radar network have an essential influence on the initial state (improved precipitation forecast for the first ~ 4..5 h) Status and Summary COSMO-K
FE 1 Koppert et al., DWD4COPS NinJo nNinJo is an international Workstation project Partners are: MeteoSuisse, DMI, MSC nFocus on Supporting Process Weather Forecasting nBase Functionality Data decoding Data storage 2D-display of all data types needed for operations Smooth 3D-extension currently worked on, prototype exists Interactive and batch ( this summer ) processing Chart-based display with zooming and panning Diagram-based display: time series, cross sections, tephigramms Data display in different layers Animation and automatic updates nMeteorological Functionality Interactive chart generation ( fronts etc. ) “On Screen” analyses Weather monitoring and warning generation nOld workstation system is currently phased out
FE 1 Koppert et al., DWD4COPS NinJo Supported Data Types nSurface and upper air observations Synop, Ship, Metar, Temp …. nGrid GME, COSMO, ECMWF, HIRLAM, GEM, GFS, nSatellite Geostationary satellite Polar orbiters nRadar SCIT Storm cell and identification nLightning Different networks nGeo data Vector Raster
FE 1 Koppert et al., DWD4COPS NinJo The Application nThe main window nMultiple scenes nLayers nBasic operation bar Zoom, pan, measure, reproject, print … nLayer bar nLayer specific tool bar nLayer specific menu bar nAnimation bar
FE 1 Koppert et al., DWD4COPS NinJo - Diagrams nA COSMO-E (LME) sounding With several derived parameters ( CAPE.. ) Available interactively for every point on the map and every model that’s in the database nA COSMO-K (LMK) Cross-Section Works with model and p-surfaces 2D-cross sections ( wind, temperature, clouds, …. ) 1D- cross section ( hourly rainrates, T2m… )
FE 1 Koppert et al., DWD4COPS NinJo-Status Karlsruhe and Hohenheim nNinJo servers and clients available Software installed by Consultant (paid by DWD) Single server installation Currently runs on data provided by DWD and routed through FU Berlin Supply through DWDSAT also possible 2MBit satellite data stream Observations ( surface, Upper air, lightning, radar … ) and model data Subsampled GME and COSMO-E, no COSMO-K Prepared for additional data e.g. COSMO-K Additional data needed ( e.g. Konrad ) has to reported –FTP based supply has to be set-up well in advance –Band-widths issues ? nStandard operational DWD-installation, based on NinJo 1.22 nStill low-cost alternative JavaMAP
FE 1 Koppert et al., DWD4COPS European regional multi-model ensemble SRNWP-PEPS Combines the most sophisticated operational limited area models in Europe the ensemble size depends on location ensemble size
FE 1 Koppert et al., DWD4COPS SRNWP-PEPS …. used to generate warnings of extreme events Output variables (surface fields only) Total precipitation Total snow Maximum 10 m wind speed Maximum 10 m wind gust speed 2 m temperature relative humidity 2m global radiation at surface Products: Ensemble mean Probabilities of exceeding thresholds +18h and +30h 6 AM and 6 PM
FE 1 Koppert et al., DWD4COPS Participating Models
FE 1 Koppert et al., DWD4COPS [%] ensemble mean observations probabilities RR >50mm [mm] 24h precipitation run: UTC, available: :05 UTC valid: , 6 UTC
FE 1 Koppert et al., DWD4COPS Output variables: tigge+ list Models: modelhor. res.institution COSMO-CH22,2MeteoSwiss ITA-LM2,8CNMCA COSMO-LAMI2,8ARPA-SIM COSMO-K2,8DWD MOLOCH2,2ISAC-CNR / ARPAL-CFMI AROME2,5Météo-France GEM-LAM 2,5Environment Canada Time schedule: March dry-run : Thu Wed April: set up of MICRO-PEPS April dry-run : Tue Mon ensemble size MICRO-PEPS
FE 1 Koppert et al., DWD4COPS Scenarios for Data Assimilation Real time Experiment data are available for operational runs Impact studies in delayed mode by excluding data from experiment Pro: Potentially better operational forecasts, direct feedback from monitoring Con: More difficult to monitor and need for extra delayed mode assimilations because of incomplete coverage Delayed mode Operational runs only use standard observations Impact studies in delayed mode by including data from experiment Pro: More controlled set-up and easier monitoring, complete data set Con: No impact on operational forecasts Common requirements Detailed list of experimental stations for blacklist and possibly whitelist. Distribution of data in agreed formats and in known ways