m. baldauf , k. helmert, b. hassler, k. stephan, s. klink,
DESCRIPTION
M. Baldauf , K. Helmert, B. Hassler, K. Stephan, S. Klink, C. Schraff, A. Seifert, J. Förstner, T. Reinhardt, P. Prohl, C.-J. Lenz, U. Damrath Deutscher Wetterdienst, Offenbach, Germany F. Theunert Amt für Wehrgeophysik, Traben-Trarbach, Germany. - PowerPoint PPT PresentationTRANSCRIPT
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 1 -
LMK (Lokal-Modell-Kürzestfrist)
3rd COPS and GOP WorkshopUniversity of Hohenheim, Stuttgart
10./11.04.2006
M. Baldauf, K. Helmert, B. Hassler, K. Stephan, S. Klink,
C. Schraff, A. Seifert, J. Förstner, T. Reinhardt, P. Prohl,
C.-J. Lenz, U. Damrath Deutscher Wetterdienst, Offenbach, Germany
F. TheunertAmt für Wehrgeophysik, Traben-Trarbach, Germany
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 2 -
Outline
• LMK - Project
• Case study examples of precipitation; verifications
• Developments
• Physical Parameterisations
• Dynamics, Numerics
• Latent Heat Nudging
• Radar quality product
• Time table
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 3 -
LMK - (Lokal-Modell-Kürzestfrist) Subproject P2 of the ‘Aktionsprogramm 2003’ (DWD)
Goals
• Development of a model-based NWP system for very short range (‘Kürzestfrist’) forecasts (2-18 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, ...)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 4 -
• grid length: x = 2.8 km
• direct simulation of the coarser parts of deep convection
• interactions with fine scale topography
• timestep t=30 sec.
• 421 x 461 x 50 grid points~ 1200 * 1300 * 22 km³lowest layer in 10 m above ground
• forecast duration: 18 h
• center of the domain 10° E, 50° N
• boundary values from LM / LME(x = 7 km)
LMK-Configuration
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 5 -
LMK 2.8 km & explicit convection (M9)•Numerical schemes•Physical parameterizations•Lateral and upper boundaries•Case studies and intercomparison•LMK Testsuites
Latent Heat Nudging (M8)•Thermodynamic feedback and interactions of LHN•Adaptation of LHN to prognostic precipitation
Radar (M7)•5-min DX-radar composit, European composit •pattern recognition of false echoes --> DXQ - quality product
Verification (M10)•Traditional verifikation (Syn/Temp)•Use of synthetic satellite- and synthetic radar tools•New methods (pattern recognition, upscaling, ...)
LMK - Subprojects (‚Measurements‘)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 6 -
Deep Convection none!-> no distinction between convective and stratiform precipitation!
Shallow ConvectionTiedtke (1989) shallow conv., only for Hcloud < 2 km
Physical Parameterisations (I)
Soil-Vegetation-Model
LM LMK
TERRA, 7 levels, additional freezing/melting
Tiedtke (1989)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 7 -
Monthly Mean of diurnal cycle of area averaged precipitation, July 2004
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 8 -
12.07.2004
02.07.2004
07.07.2004
18.07.200417.07.2004
...... radar
___ LMK, 00 UTC
___ LM, 00 UTC
precipitation: diurnal cycle of area average
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 9 -
Monthly precipitation mean of ‚July 2004‘
ground obs. radar obs. LM
LMK w/o DA LMK + DA LMK + DA
Mean: 103 mm Mean: 80 mm Mean: 95 mm
Mean: 67 mm Mean: 78 mm Mean: 72 mm
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 10 -
12.07.2004
02.07.2004
07.07.2004
18.07.200417.07.2004
...... radar
___ LMK, 00 UTC
___ LM, 00 UTC
precipitation: diurnal cycle of area average
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 11 -
ground obs. radar obs. LM
LMK w/o DA LMK + DA LMK + DA
24h precipitation sum ‚12.07.2004, 06...30 UTC‘
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 12 -
12.07.2004
02.07.2004
07.07.2004
18.07.200417.07.2004
...... radar
___ LMK, 00 UTC
___ LM, 00 UTC
precipitation: diurnal cycle of area average
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 13 -
ground obs. radar obs. LM
LMK w/o DA LMK + DA LMK + DA
24h precipitation sum ‚17.07.2004, 06...30 UTC‘
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 14 -
12.07.2004
02.07.2004
07.07.2004
18.07.200417.07.2004
...... radar
___ LMK, 00 UTC
___ LM, 00 UTC
precipitation: diurnal cycle of area average
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 15 -
Example: 18.07.2004, 00 UTC + 10 h
LM LMKRadar
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 16 -
Example: 18.07.2004, 00 UTC + 12 h
LM LMKRadar
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 17 -
Example: 18.07.2004, 00 UTC + 18 h
LM LMKRadar
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 18 -
True skill statistics (TSS)
for precipitation rates
July 2004, 12-UTC-runs
LMK: averaging to the 9 neighbouring grid points
=TS 2.2=TS 2.2b=TS 1.7=TS 1.6=TS 1.5
> 0.1 mm/h > 2 mm/h
> 10 mm/h
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 19 -
LM and LMK: ETS for gusts > 20 m/s, January 2004, 00-UTC-runs
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 20 -
Bias (mean error) of 2m-temperature, July 2004, 00-h-runs
=TS 1.5=TS 1.6=TS 1.7=TS 2.2=TS 2.2b
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 21 -
Conclusions, Verification results:
• Precipitation:
• LMK delivers partly better verification results than LM/LME
• Diurnal cycle (especially time of maximum) is often more realistic
• problem up to now: general underestimation of the precipitation amount in convective events solution proposals: improvement in physics-dynamics-coupling, cloud physics, shallow convection, subscale cloud coverage
• Gusts: better quality of LMK-forecasts compared to LM/LME because of finer numerical (vertical and horizontal) and no convection param.
• 2m-temperature:stronger bias than in LM, because no soil moisture analysis (SMA) is used up to now; can LHN help?
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 22 -
LMK 2.8 km & explicit convection (M9)•Numerical schemes•Physical parameterizations•Lateral and upper boundaries•Case studies and intercomparison•LMK Testsuites
Latent Heat Nudging (M8)•Thermodynamic feedback and interactions of LHN•Adaptation of LHN to prognostic precipitation
Radar (M7)•5-min DX-radar composit, European composit •pattern recognition of false echoes --> DXQ - quality product
Verification (M10)•Traditional verifikation (Syn/Temp)•Use of synthetic satellite- and synthetic radar tools•New methods (pattern recognition, upscaling, ...)
LMK - Subprojects (‚Measurements‘)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 23 -
Deep Convection
Shallow Convection Tiedtke (1989) shallow conv., only for Hcloud < 2 km
Physical Parameterisations (I)
Soil-Vegetation-Model
LM LMK
TERRA, 7 levels, additional freezing/melting of snow
Tiedtke (1989) none!-> no distinction between convective and stratiform precipitation!
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 24 -
Shallow convection
based on Tiedtke-schemeN-S-crosssection
rh(with shallow convection) Diff.: rh(with sh. conv) - rh(without sh. conv.)
Dr. F. Theunert (AGeoBW)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 25 -
Radiation
Physical Parameterisations (II)
Cloud Microphysics 5-class-scheme 6-class (Graupel)-scheme
6-class/2 moments-scheme(Seifert, Beheng, 2000)
(for research/benchmark purposes)
Turbulence• 1-dim.• 1 eq. model (progn. TKE)
2-flux-scheme (Ritter, Geleyn,1992)
• 3-dim., full coordinate transforms (Herzog et al., 2002; Baldauf, 2005)
• 1 eq. model (progn. TKE) • moist turbulence (=condensation alters
buoyancy-prod. of TKE)
LM LMK
upscaling, higher frequency
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 26 -
Underestimation of precipitation in LMK
• systematic underestimation of ~20 % in many cases• even enhanced in strong convective situations
Possible solution approaches:• enhancement of subscale coverage in the boundary layer
cloud diagnostic scheme• change of entrainment-/detrainment-coefficients in the
parameterisation of shallow convection• Tuning of cloud physics: reduced evaporation of rain below cloud base
Side effects:• all changes influence cloud coverage and 2m-temperature
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 27 -
LMK 3.17 + changes in moisture turbulence (enhanced subscale coverage)
+ changes in shallow convection (enhanced entrainment/detrainment)
Sensitivity study for the problem of underestimation of precipitation
case study ‚26.08.2004‘
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 28 -
Radiation
Physical Parameterisations (II)
Cloud Microphysics 5-class-scheme 6-class (Graupel)-scheme
6-class/2 moments-scheme(Seifert, Beheng, 2000)
(for research/benchmark purposes)
Turbulence• 1-dim.• 1 eq. model (progn. TKE)
2-flux-scheme (Ritter, Geleyn,1992)
• 3-dim., full coordinate transforms (Herzog et al., 2002; Baldauf, 2005, 2006)
• 1 eq. model (progn. TKE) • moist turbulence (=condensation alters
buoyancy-prod. of TKE)
LM LMK
upscaling, higher frequency
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 29 -
Metric terms of 3D-turbulence
scalar flux divergence:
scalar fluxes:
analogous:‚vectorial‘ diffusion of u, v, w
Baldauf (2005), COSMO-Newsl. Nr. 5
earth curvature
terrain following coordinates
LES-3D-turbulence model from ‘Litfass-LM’ Herzog et al. (2003) COSMO Techn. rep. 4
vertical
horizontal(cartesian)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 30 -
total precip. in 18 h
Real case study: LMK (2.8 km resolution) ‚12.08.2004, 12 UTC-run‘
‚3D-turb., with metric‘ - ‚1D‘
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 31 -
Advectionu, v, w, T, p‘
Numerics
Grid
horizontal: centered diff. 2. ordervertical: implicit 2. order
Bott-2, conservation form orSemi-Lagrange, tricubic/trilin.
horizontal: upwind 3., 5. ordercentered diff. 4., 6. order
vertical: implicit 2. orderimplicit 3. order
Time integration 3-timelevels: Leapfrog
horizontal: Arakawa-Cvertical: Lorenz
2-timelevels: Runge-Kutta 2. order, 3. order, 3. order TVD
LM LMK
Advectionqv, qc, qi, qr, ..., TKE
qv, qc: centered diff.2. orderqi: Lin, Rood qr, qs: Semi-Lagrange (trilin.)
Smoothing
Divergence damping
4. order diffusion
Asselin-filtering
4. order diffusion (?)
filtering of orographyfiltering of orographystronger filtering of the Alps
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 32 -
Current work in Dynamics:
• diabatic terms in pressure equation
• influence of deep/shallow atmosphere
Current work in Numerics:
• unrealistic cold temperatures in narrow valleys (‚cold pool problem‘)
• is due to pressure gradient in terrain following coordinates --> solution by
• Dynamic lower boundary condition (Gassmann, 2004)
• slope dependent orographic filtering
• modified physics-dynamics-Coupling
• improved 3. order vertical advection for dynamic variables (u,v,w,T,p‘)
• radiative upper boundary condition
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 33 -
LMK 2.8 km & explicit convection (M9)•Numerical schemes•Physical parameterizations•Lateral and upper boundaries•Case studies and intercomparison•LMK Testsuites
Latent Heat Nudging (M8)•Thermodynamic feedback and interactions of LHN•Adaptation of LHN to prognostic precipitation
Radar (M7)•5-min DX-radar composit, European composit •pattern recognition of false echoes --> DXQ - quality product
Verification (M10)•Traditional verifikation (Syn/Temp)•Use of synthetic satellite- and synthetic radar tools•New methods (pattern recognition, upscaling, ...)
LMK - Subprojects (‚Measurements‘)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 34 -
ground obs.:Synop, ship, buoy
upper air obs.:AMDAR,Radiosonde (TEMP)PILOT
Latent Heat Nudging
Data assimilation
T2m Soil moisture
LM LMK
SMA
Nudging
SMA (?)(not used up to now)
Radarprecipitation scanresolution: 1 km * 1°
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 35 -
Basics of Latent Heat Nudging
Differences (or ratios) between (radar) measured and simulated precipitation rates are interpreted as a lack/surplus of latent heat along the trajectory of a condensed particle.
Basic assumption of LHN: this relation is valid in a vertical model columnvertical structure of latent heating <--> temperature increments (optional: moisture increments, e.g. by conservation of relative humidity)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 36 -
scores for hourly precipitation : with latent heat nudging / without latent heat nudging
Testsuite ‚Juli 2004‘ (7.-16.7.04)
>2.0 mm>0.1 mm threshold values
FBI
ETS
12-UTC forecasts
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 37 -
scores for hourly precipitation : with latent heat nudging / without latent heat nudging
Testsuite ‚Juli 2004‘ (7.-16.7.04)
>2.0 mm>0.1 mm threshold values
FBI
ETS
18-UTC forecasts
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 38 -
>2.0 mm
FBI
ETS
>0.1 mm threshold values
scores for hourly precipitation : with latent heat nudging / without latent heat nudging
Testsuite ‚Juli 2004‘ (7.-16.7.04)
0-UTC forecasts
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 39 -
Conclusions; Latent Heat Nudging:
• An assimilation of spatially and temporally highly resolved data is needed to
trigger convection on the LMK scale
• Latent Heat Nudging of radar reflectivities can be used for this task
• Forecasts especially for 12h-, 18h-runs can be improved.
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 40 -
LMK 2.8 km & explicit convection (M9)•Numerical schemes•Physical parameterizations•Lateral and upper boundaries•Case studies and intercomparison•LMK Testsuites
Latent Heat Nudging (M8)•Thermodynamic feedback and interactions of LHN•Adaptation of LHN to prognostic precipitation
Radar (M7)•5-min DX-radar composit, European composit •pattern recognition of false echoes --> DXQ - quality product
Verification (M10)•Traditional verifikation (Syn/Temp)•Use of synthetic satellite- and synthetic radar tools•New methods (pattern recognition, upscaling, ...)
LMK - Subprojects (‚Measurements‘)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 41 -
Examples (I)
Negativ spokesHannover, 17.01.2006 16:00 UTC
free pixels or pixel groups
Rostock, 12.02.2006 23:00 UTC
Anomalous propagationBerlin, 6.9.2005 12:55 UTC
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 42 -
strong positive sectorsreson: technical radar problems
detection is essential !
detection by:- sharp edges- histograms
‚Klops‘
reasons: warm bubbles?
aerosols?
detection by:
- histograms
Examples (II)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 43 -
Spoke recognition
Berlin 21.06.2005 22:10 UTC
after spoke recognition‘Pos. 3’ of quality productOriginal
LHN
Examples (III)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 44 -
Quality product (DXQ)
same data format as DX-data (planned: BUFR-format)
• Header:
– signature ('DXQ'), date/time, radar location, ....
– coding of radar errors, concerning the whole radar picture:– hardware errors (Wartung, e.g technical problems)– ‘Klops’– radome-Damping
• binary part: pixelwise coding of 8 radar errors (1. bit: recognized, 2. bit: corrected)
1. exceed of threshold in signal processor (RVP) (LOG/SQI/WSP/CCOR/speckle)
2. cluster (def. by 9 neighbouring pixels)
3. spokes - positive- and negativespokes
4. vertical reflectivity profile
5. bright band
6. distance dependent damping
7. ......
8. .......
April 2006: DXQ-composit (QY) and adequate DX-composit (RY) operational
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 45 -
Project LMK: time table of the pre-operational test phase
• July 2003 Start of project LMK• End 2003 First test suites with 2.8 km resolution• End 2004 First test suites with data assimilation
• March 2006: Prototype-version of LMK-system with LHN for ‘summer 2005’ (test suites 3.x).
• April 2006: DXQ-Composit operational• Mai 2006: Start of pre-operational test phase.
Fine-Tuning and evaluation of all components
• June 2006 ‘Introduction group’ (TI14)‘Evaluation group’ (FE 15, WV, AG Evaluierung)
• Dez. 2006: End of AP2003• Spring 2007: Start of operational use after decision of the KG-NWV
and board of directors
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 46 -
LMK-forecast runs (18h-forecasts, started every 3h) -> LAF-ensembleavailable at: ..., 1, 4, 7, 10, ... , 22, ... UTC
Lagged Average Forecast (LAF)-Ensemble (Prinzip)
LMK-data assimilation cycle (rapid update cycle) • continuous assimilation -> nudging• short cut off ( ~ ½ h)• use all data, especially radar data• adiitionally: split into ‘main run analysis’ and a pure data assimilation cycle
possible changes in this sequence:
• additional soil moisture analysis (SMA)
• GME, LME and LMK cannot run at the same time on the computer
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 47 -
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 48 -
from: R. A. Houze, Jr.: Cloud Dynamics International Geophysics Series Vol. 53
Schematic model from a Colorado storm casestudy (Raymer Hailstorm)
Deep moist convection2.8 km
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 49 -
Explicit Convection in LMK (case study '26.08.2004')
LMK, Testsuite 1.7RadarLM
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 50 -
Monatsmittel der Niederschläge ‚August 2004‘
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 51 -
Monthly mean of precipitation (area average over Germany)
Ground Obs. LMK LM
Juli 2004 103 mm 82 mm (TS 2.2) 99 mm
Aug. 2004 88 mm 70 mm (TS 2.2) 84 mm
Sept. 2004 64 mm 55 mm (TS 2.2) 56 mm
Okt. 2004 52 mm 42 mm (TS 2.1) 41 mm
Nov. 2004 72 mm 77 mm (TS 2.1) 68 mm
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 52 -
Testfall 03.06.2005, 0UTC + 21h
bessere Böenvorhersage im LMK aufgrund höherer horizontaler Auflösung undohne Konvektionsparametrisierung
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 53 -
Vorteile beim Abschalten der Konvektionsparametrisierung realistischerer Tagesgang des Niederschlags
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 54 -
Physics (I)• Radiation• Shallow Convection• Coriolis force• Turbulence
DynamicsRunge-Kutta [(phys) + (adv) --> fast_waves ]
‚Physics (I)‘-tendencies n(phys I)
Physics (II)• Cloud Physics
Physics-Dynamics-couplingfields n=(u, v, w, pp, T, ...)n
fields n+1=(u, v, w, pp, T, ...)n+1
fields *=(u, v, w, pp, T, ...)*
‚Physics (II)‘-tendencies n(phys II)
+ n-1(phys II) NEW!
- n-1(phys II) NEW!
Descr. of Advanced Research WRF, v. 2 (2005)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 55 -
LMK
Radar
MC2
model intercomparison:MAP IOP 2b(„19./20.09.1999“)heaviest RR during MAP(300mm / 30 h)
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 56 -
1: MM5
2: WRF
3: MOLOCH
4: Méso-NH
5: LMK
6: Aladin oper. 1999
7: Arpège oper. 1999
8: EZMW oper. 1999
9: EZMW oper. 2002 with MAP-reanalysis
--> LMK performs not bad !:• bias less than for the most other models
• correlation & RMSE better than for operational forecasts, worse than for (tuned?) science models.
LMK
LMKLMK
Bias (mm) RMSE (mm) Correl. coeff.model intercomparison:MAP IOP 2b(„19./20.09.1999“)heaviest RR during MAP(300mm / 30 h)
precipitation scores:
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 57 -
6-class-cloud-microphysics scheme in LMK
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 58 -
Testsuite 1-6with simple shallow convection param.
Testsuite 1-5without shallow convectionpressure jump problem
operational LMthe ‚truth‘ (?)
Need for a shallow convection parameterization:
10./11.04.2006
Aktionsprogramm 2003
AP 2003: LMK - 59 -
ground obs. radar obs. LM
LMK w/o DA LMK + DA LMK + DA
24h precipitation sum ‚18.07.2004, 06...30 UTC‘