Status of WG1 Science Plan Issues

[email protected] of WG1 and Science Plan IssuesCOSMO General Meeting, Sibiu, 2 – 5 Sept. 2013 1

Status of WG1Science Plan Issues

Christoph SchraffDeutscher Wetterdienst, Offenbach, Germany

• KENDA

• LETKF at CNMCA : talk by Lucio Torrisi

• 3DVar at HMC (Sochi area): not (yet) as good as interpolated GME analyses; will test use of radar radial winds

• ARPA-SIM (Virginia Poli): comparison of 1D-Var+nudging and LHN

1D-Var not as good as LHNProblem: bias between moist physics of COSMO and diagnostic scheme used in 1D-Var

work will be stopped

• at DWD: •Direct nudging of radial velocity•LHN for COSMO-EU•Nudging of GPS-IWV


threshold:0.1 mm / h

FSS, 11 g.pts. (30 km)

12-UTC forecast

runs

nudging of radar radial velocity:verif. vs. radar precipitation (1 – 27 May 2012)

00-UTC forecast

runs

ETS

control (C-DE setup, with LHN)use of radial velocity added


nudging of radar radial velocity: summary

• experiment May 2012 with additional use of vr (control with LHN !) :

small positive impact on (mainly strong) precipitation otherwise neutral no show cases with strong positive impact

• next: check and improve efficiency, add few more tests, make it operational


Latent Heat Nudging for COSMO-EU

Motivation

•Reanalysis Project of HErZ would like to use LHN also on its larger domainconfiguration (COSMO-CORDEX, x = 7km)

•Does it work ?

Is there a negative interaction between LHN and the deep convection scheme ?

Experiments

•2 periods: Summer: August 2012 (incl. convective cases)

Winter: 1 Nov. – 15 Dec. 2012 (incl. snow cases)

•operational COSMO-EU setup (Cosmo V4_27, tk?min=0.4)

•operational LHN setup except: • no blacklist information used• no bright band detection• LHN within first 25 minutes of forecast (will be ~ 2.5h in operational run)


• same data as for COSMO-DE,

currently•17 German Doppler radar stations

quality controlled•2 Dutch stations•2 Belgian stations•9 France stations•3 Swiss stations•2 Czech stationssoon extended by 2 Polish stations

• every 5 minutes

• future: can be extended by OPERA composite outside COSMO-DE domain

LHN for COSMO-EU:radar data coverage


hourly precipitation at 2 Aug. 2012, 00 UTC + 4h

Radar Control (rout) LHN Experiment

LHN for COSMO-EU:case study


12 UTC 0.1 mm/h 12 UTC 2.0 mm/h

00 UTC 2.0 mm/h00 UTC 0.1 mm/h

LHN for COSMO-EU:verif. vs. radar precipitation (August 2012)


• LHN in COSMO-EU is beneficial; forecast impact:

• Summer:

• precipitation improved (esp. convective cases)

• most surface parameters slightly improved

• Winter: neutral

• Work started to extend data coverage using OPERA radar composite

LHN for COSMO-EU:Summary


Status of OPERA

• three 2-D composites are available every 15 minutes:

• current rain rate, sum over last hour, current maximum reflectivity

• Phillipe Lopez (ECMWF) found out:

• quality has increased in 2013 (with Baltrad QC) but is still not sufficient everywhere

• best quality over UK, France, Poland and Germany

• underestimation over southern Europe

• underestimation in snow cases

• lots of spurious ground echoes

• software transforming it to COSMO grids available (Stephan)


ODC composite

12 Aug 2013, 15 UTC

Status of OPERA


Status of OPERA

‘EY composite: Input for LHN at DWD,

4 different storm-dep. Z-R relationships

OPERA DC composite (C-DE domain):

1 Z-R relationship

differences particularly for low precip rates; (erroneous spaital alignment ?)

12 Aug.2013

15 UTC


Status of OPERA

• 3-D reflectivity and radial wind:

• reflectivity sent by almost all member states,

radial wind only by few(but increasing)

(Courtesy Hidde Leijnse, * indicates max. number)

• HIRLAM already gets access to volume data, but possible also for other consortia

• dissemination of volume data to member states planned for end of 2014 (to be decided: with or without quality information ?)

• discussion about meta data / quality issues are ongoing(e.g. to distinguish ‘zero echo’ – ‘no data’ (compare with raw data))


• overall strategic goals of COSMO: convective-scale EPS convective-scale ensemble-based DA

• other consortia: focus on 3D-/4D-Var for deterministic, EnKF for EPS hybrid

• Hybrid ‘4D-EnVar’ : without need of TL / adjointEnvironment Canada ! UKMO ? HIRLAM ?

• DWD: hybrid 3D-EnVar for global ICON for COSMO ? • advantages of EnVar over LETKF (for deterministic run !) :

localization on B, not in obs space better for non-local obs hybrid approach mitigates sampling error

• disadvantages of EnVar over LETKF: 4D approach limited: Use only obs times where Grib fields used costs

COSMO: continue with LETKF

COSMO Science Plan Issues on DA:General strategy


• consolidation of LETKF scheme, wherever found required,may include extensions:

– intrinsic stochastic physics (links to WG3a, WG7)

– (other types of) additive covariance inflation

– iterative (e.g. running in place like) approaches or latent heat nudging (LHN) in view of improving convection initiation

– use of ensemble members with lagged valid time to reduce phase errors

– blending techniques to combine information from the larger scales of the nesting model

COSMO Science Plan Issues on DA:Action items 2015 2017 (and beyond !)


• integrated forecasting system for nowcasting (NWC) andvery short-range-forecasting based on NWP

– towards very high resolution (1 km / sub-km)

(link to SFP project COSMO-Airport, incl. Mode-S data; sub-km)

– towards very high frequent analysis and forecast update

(link to HErZ TB1, KENDA with 5-min update using 3-d radar data)

– main issue: quality ! (need to match obs, but maintain balance)

– optional: ‘higher weights for latest observations’ (smaller obs errors, weaker data thinning) ? smaller localization scales ?

– optional: evaluation of assimilation of NWC products as pseudo-obs such as stability indices and objects (using non-conventional metrics)



• extended use of obs (particularly PBL, low-level humidity, cloud, precip)

– consolidate obs focused in KENDA PP: radar, GPS slant path, SEVIRI cloud

– improved use of screen-level obs (T2m, RH2m, uv10m)

– SEVIRI WV (water vapour) channels

– AMDAR humidity ; high-resolution aircraft Mode-S (wind) data

– microwave radiometer (+ Raman lidar) T-, q- profiles, Doppler wind lidar, ceilometer cloud base height

– (SEVIRI VIS for cloud properties, cloud radar data (CALIPSO, EarthCARE 2015)

– renewable energy: power data from wind mill farms and solar power systems satellite

• no aerosol assimilation :– for COSMO-ART : accurate description of aerosol sources – would require substantial resources



• lower BC (analysis (link to WG3b) + perturbations (link to WG7) !)

– soil moisture SM (soil temperature, etc.), e.g. by:

– include SM in control vector, add SMOS / ASCAT satellite SM

– separate 2-D column analysis: LETKF, or EPS used in Var-SMA

– snow cover and depth

– SST (perturbations e.g. by pattern generator)

• diagnosis: (explore in EnKF framework) use of FSO (Forecast Sensitivity to Observations) (→ link to HErZ project) DFS (Degrees of Freedom for Signal) diagnostics

(to assess usefulness of different observation types)

• 2017+ : adapt LETKF to ICON-LAM (review decision LETKF – EnVar)



• … see previous slides

• MTG-IRS

• MW + IR radiance over land + in cloudy areas (dep. on experience in global DA)

• parameter estimation ?

• observation error correlations ?

COSMO Science Plan Issues on DA:Action items 2018 2020


• insufficient quality of LETKF ( develop EnVar ?)

• need to maintain / increase (often temporary) human resources ! Depends on:

continuation of HErZ after 2014

new fixed position for radar DA

new temporary positions ?

EUMETSAT fellowship submitted by CNMCA: sat data on soil moisture

further increase of involvement e.g. by MCH, ARPA-SIM, CNMCA, …

… new projects / collaboration with academia

new flexible portable scripts may facilitate experimentation + collaborationwith other institutes on a technical level observation error correlations ?

COSMO Science Plan Issues on DA:risks


• consolidate / refine LETKF wherever required (enhancement techniques e.g. to correct phase errors; RIP, LHN …)

• integrated forecasting system for NWC – (V)SRNWP

• obs: radar, GPS slant, cloud; screen-level; SEVIRI WV, MTG IRS, …

• lower BC

• diagnosis: FSO , DFS

• adapt to ICON-LAM (develop EnVar ?)

Science Plan issues on data assimilation:Summary


Science Plan issues on data assimilation:Summary


developments / plansat other consortia

• AROME:

– EDA: 3DVar with perturbed obs incl. surface ana (AROME 2014, Arpege LBC) later: revisit breeding / ETKF

– multiplicative inflation of forecast perturbations: 1.15 (every 3 hrs)

– stochastic physics tendencies (SPPT): large-scale, slowly evolving

– also tried: microphysics & turbulence parameter perturbations

– perturbed surface initial conditions (SST, SM, Rsmin, LAI, Cveg, Z0oro)

– coupling with large-scale analysis: tests with Jk term in Var(weak constraint towards coupling data in Var, truncate Jk info at

k=20)

– Tests with 1-hrly cycles (3DVar)

– use of SEVIRI WV over land

– issue: model bias with strong diurnal cycle


• UKMO:

– NWP nowcasting system: Nowcasting Demonstration Project (NDP)

– 1hrly 4DVar with sub-hrly obs (…, MSG WV + AMVs, radar radial winds; hourly cloud obs (sat + surf), T2m, RH2m, uv10m) + LHN (!)

– 12-hour forecast every hour

– 40km-FSS beats STEPS (Lagrangian-extrapolation NWC) after 1.8 hrs

– forms basis for next-generation UK NWP nowcasting system

– generally: towards 4D-EnVar (without need of TL / adjoint ?)

developments / plansat other consortia

Status of WG1 Science Plan Issues

Documents