EUCLIPSE Toulouse meeting April 2012
Roel Neggers
WP3 – SCM resultsWP3 – SCM results
Mapping the behavior of subtropical marine Mapping the behavior of subtropical marine boundary-layer clouds under weakening boundary-layer clouds under weakening
inversionsinversions
Contents
Case update – where are we now?Case update – where are we now? The bigger picture: A further analysisThe bigger picture: A further analysis
Kappa – TCC diagramsKappa – TCC diagrams
i) Single cases i) Single cases
ii) Ensemble SCM: internally resolved composite transitionsii) Ensemble SCM: internally resolved composite transitions
iii) CMIP5 station dataiii) CMIP5 station data
Case update
List of participantsList of participants
Short reminder of general model performanceShort reminder of general model performance
A remaining issue in the ASTEX case setup A remaining issue in the ASTEX case setup (spinup effects in surface evaporation & low level wind)(spinup effects in surface evaporation & low level wind)
List of participants
Name Affiliation Model ASTEX Composite cases
Eric Basile Météo France AROME
ARPEGE-NWP
Isabelle Beau Météo France ARPEGE-CLIMAT
Vincent Larson UWM CLUBB
Sara dal Gesso
Roel Neggers
KNMI EC-Earth
RACMO
Suvarchal Kumar MPI-M Hamburg ECHAM6 Expected soon
Irina Sandu
Martin Köhler
ECMWF
DWD
IFS cy36r1
Hideaki Kawai JMA JMA
Anning Cheng NASA LaRC LaRC
Heng Xiao UCLA UCLA-AGCM
Marie-Pierre Lefebvre Météo France LMDZ
Wayne Angevine NOAA WRF TEMF
Cisco de Bruijn
Wim de Rooij
KNMI HARMONIE EDKF Expected soon
HARMONIE EDMF Expected soon
Jennifer Fletcher University of Washington NCEP GFS
Sungsu Park University of Washington NCAR CAM5
Ian Boutle UK Met Office UKMO
New: * Case updates from LMD, NCEP-GFS, ECHAM6 and HARMONIENew: * Case updates from LMD, NCEP-GFS, ECHAM6 and HARMONIE * CMIP5 physics versions from LMD (AR4) and Met Office (HadGem2)* CMIP5 physics versions from LMD (AR4) and Met Office (HadGem2)
Reminder: General model performance (Exeter meeting)
Spaghetti:Spaghetti:
This looks nasty … should this really be our main This looks nasty … should this really be our main message, or can we do something better / more message, or can we do something better / more instructive?instructive?
Remaining issue in ASTEX case
Spinup effects in surface evaporation as produced by most SCMsSpinup effects in surface evaporation as produced by most SCMs
Low level wind speedLow level wind speed
Further analysis in EC-Earth SCM: hodographsFurther analysis in EC-Earth SCM: hodographs
m m
Budget analysis of V at 10m:Budget analysis of V at 10m:
Surface momentum flux (u’w’ + v’w’):Surface momentum flux (u’w’ + v’w’):
Witin the first three hours, the depth of the vertical mixing in Witin the first three hours, the depth of the vertical mixing in momentum is much smaller in most SCMs compared to the LESmomentum is much smaller in most SCMs compared to the LES
Where to go from here?
Q: How can these transition case studies contribute?Q: How can these transition case studies contribute?
Q: Are our single case studies representative of GCM climate?Q: Are our single case studies representative of GCM climate?
Idea: To establish “fingerprints” of relevant fast physics across Idea: To establish “fingerprints” of relevant fast physics across a hierarchy of modelsa hierarchy of models
Can we identify similar patterns in GCM and SCM Can we identify similar patterns in GCM and SCM studies?studies?
The question: What “key diagnostic” do we choose?The question: What “key diagnostic” do we choose?
EUCLIPSE goal: attribution of GCM behavior to sub-grid physicsEUCLIPSE goal: attribution of GCM behavior to sub-grid physics
Kappa-TCC Kappa-TCC spacespace
Further analysis
CTEI parameter
Kuo and Schubert (QJRMS, 1988)Moeng (JAS, 2000)
Lock (QJRMS, 2009)LES results
ll qqtt
Sandu and Stevens (JAS, 2011)
““Cloud cover is universally large when Cloud cover is universally large when < < 0.2 and universally small when 0.2 and universally small when > 0.4 in > 0.4 in broad agreement with previous studies. broad agreement with previous studies.
However, there is considerable scatter in However, there is considerable scatter in the value of cloud cover over the value of cloud cover over intermediate values of intermediate values of , and our , and our simulations neither support a threshold in simulations neither support a threshold in cloud cover associated with a critical cloud cover associated with a critical value of value of nor a one to one relationship nor a one to one relationship between between and cloud cover. and cloud cover.
That cloud cover should generally That cloud cover should generally decrease as decrease as increases is reasonable, as increases is reasonable, as larger values of larger values of imply the entrainment imply the entrainment of relatively dry air. of relatively dry air.
The relative scatter however indicates The relative scatter however indicates that other processes also play an that other processes also play an important role.”important role.”
More LES resultsMore LES results
Arno Nederlof, Bachelor’s thesis, TU Delft, August 2010
DALES simulations of multiple transitions (variations on ASTEX)
Let’s not use Let’s not use as a threshold as a threshold
Let’s use this PDF as a benchmark Let’s use this PDF as a benchmark that GCMs and SCMs have to that GCMs and SCMs have to reproducereproduce
How good are they at this, really?How good are they at this, really?
Calculating kappa in a discretized column
ll
Method 0Method 0: diagnose : diagnose over a single over a single model level at cloud topmodel level at cloud top
Method 1Method 1: diagnose : diagnose over a single over a single model level at the level of the model level at the level of the strongest strongest ll –gradient –gradient
Method 2Method 2: as 1, but now : as 1, but now is is diagnosed over the inversion diagnosed over the inversion layer, defined as the height-range layer, defined as the height-range in which the in which the ll –gradient is still –gradient is still significant (>10% of max)significant (>10% of max)
Method 3Method 3: as 2, but now correcting : as 2, but now correcting for the tropospheric lapse rate in for the tropospheric lapse rate in ll and and qqtt (downward extrapolation)(downward extrapolation)
Using Method IUsing Method I
Pretty much reproduce earlier Pretty much reproduce earlier results: cloud transition results: cloud transition occurs within range 0.1 < occurs within range 0.1 < kappa < 0.5kappa < 0.5
LES
Method IMethod ISCM
Sensitivity to Kappa-Sensitivity to Kappa-methodmethod
MetOffice SCM
Two messages:Two messages:
* * Use kappa-method that best reflects the way the inversion is treated in the top-Use kappa-method that best reflects the way the inversion is treated in the top-entrainment modelentrainment model
* Correcting for tropospheric gradients (both * Correcting for tropospheric gradients (both qqtt and and ll) tends to shift the pdf to ) tends to shift the pdf to larger kappa values. Except in ASTEX, where the tropospheric larger kappa values. Except in ASTEX, where the tropospheric qqtt gradient is gradient is significant!significant!
Ensemble SCM
The composite transitions were based on 497 observed/diagnosed The composite transitions were based on 497 observed/diagnosed transitions in the NE Pacific transitions in the NE Pacific (Sandu et al, ACP 2010; Sandu and Stevens, JAS (Sandu et al, ACP 2010; Sandu and Stevens, JAS 2011 )2011 )
Would it not be a great idea to simulate each of the 497 transitions Would it not be a great idea to simulate each of the 497 transitions individually?individually?
Benefits:Benefits:
* Composites become internally resolved* Composites become internally resolved
* Increased statistical significance * Increased statistical significance Single realizations might not represent “typical behavior” as Single realizations might not represent “typical behavior” as seen in GCMseen in GCM E.g. numerical effects like grid-locking at PBL topE.g. numerical effects like grid-locking at PBL top
* More honest comparison to GCM results* More honest comparison to GCM results
* This could facilitate the attribution of GCM behavior to sub-grid * This could facilitate the attribution of GCM behavior to sub-grid physicsphysics Serving as an intermediate stepServing as an intermediate step
Results: first impressionA selection from 497 NEP transition A selection from 497 NEP transition
casescasesSimulated with EC-Earth DualMSimulated with EC-Earth DualM
Dry CBLDry CBL
NightNightDayDay
Deep ConvDeep Conv
Resolved composite-internal variability
The single realization The single realization (composite experiment) (composite experiment) is making many is making many excursionsexcursions
In contrast, the ensemble In contrast, the ensemble mean is smoothly mean is smoothly varying, and shows a varying, and shows a weak diurnal cycleweak diurnal cycle
EC-Earth EC-Earth (cy31r1)(cy31r1)
2
Comparison to LES
The time-development of the The time-development of the ensemble-mean compares ensemble-mean compares much better to the LESmuch better to the LES
Should we use multiple instead of Should we use multiple instead of single SCM realizations to single SCM realizations to evaluate the representation of evaluate the representation of this “knife-edge” regime?this “knife-edge” regime?
SlowSlow
ReferenceReference
FastFast
Kappa-TCC diagrams
Different BL physics produce different Different BL physics produce different PDFsPDFs
Various modes can be distinguishedVarious modes can be distinguished
EC-Earth (cy31r1)EC-Earth (cy31r1) EC-Earth DualMEC-Earth DualM
Conditional sampling
Filtering out deep convective eventsFiltering out deep convective events
Comparing scatterplots to PDFs
Over-plotted: SCM results for ASTEX & composite cases (black Over-plotted: SCM results for ASTEX & composite cases (black symbols)symbols)
Can we see these PDFs as “fingerprints”?
Proof of principle: a sensitivity test on a key closure in the PBL Proof of principle: a sensitivity test on a key closure in the PBL schemescheme
Entrainment efficiency at cumulus top (Wyant et al., 1997)Entrainment efficiency at cumulus top (Wyant et al., 1997)
Ri
A
M
M
c
e
In EDMF DualM, the top-level of cumulus mass flux is replaced by K-diffusion, In EDMF DualM, the top-level of cumulus mass flux is replaced by K-diffusion, with K parameterized as a function of with K parameterized as a function of MMee
This is necessary to let the scheme represent cumulus rising-into This is necessary to let the scheme represent cumulus rising-into stratocumulus (ATEX)stratocumulus (ATEX)
How does the kappa-TCC diagram change when we do not do this?How does the kappa-TCC diagram change when we do not do this?
qqtt
cM
Sensitivity test: result
Fair weather cumulus (TCC ~ 20-40 %) occurs much more Fair weather cumulus (TCC ~ 20-40 %) occurs much more frequently, at the cost of Strato-cumulusfrequently, at the cost of Strato-cumulus
Moving up the model hierarchy: GCM
ESM output at selected gridpoints (CMIP5 ESM output at selected gridpoints (CMIP5 AMIP)AMIP)
CMIP5 AMIPPeriod : 200509 - 200608Period : 200509 - 200608
Location : Eastern Pacific ( cfSites 1-29 )Location : Eastern Pacific ( cfSites 1-29 )
Comparing PDFs
The SCMs “kind of” reproduce the PDF as diagnosed from the The SCMs “kind of” reproduce the PDF as diagnosed from the associated GCMassociated GCM
But can we make a qualitative statement? But can we make a qualitative statement? Come up with a metric? Come up with a metric?
Problem: the SCM PDF is ‘undersampled’:Problem: the SCM PDF is ‘undersampled’:
Many situations are not encountered in the SCM casesMany situations are not encountered in the SCM cases
A simple metric
bin at mode theof valuekappa:
axis TCCon index -bin :
1 2
n
n
NRMS
n
n
GCMn
SCMn
Idea: we can at least compare the location of the mode (maximum) in kappaIdea: we can at least compare the location of the mode (maximum) in kappa
A benefit: There is no contribution at those A benefit: There is no contribution at those nn where the (incomplete) where the (incomplete) SCM-PDF has no samplesSCM-PDF has no samples
SCMn GCM
n
A cross-comparison of SCMs and GCMs
RMS is always smallest for the GCM and its own SCMRMS is always smallest for the GCM and its own SCM
SCM “best matches” its native GCMSCM “best matches” its native GCM
Conclusions
Ensemble SCMEnsemble SCM
By simulating all 497 individual transitions individually, the By simulating all 497 individual transitions individually, the composite-internal variability becomes resolvedcomposite-internal variability becomes resolved
Where cloud parameters in single SCM realizations typically show a Where cloud parameters in single SCM realizations typically show a lot of random-like excursions, the diurnal cycle is much better lot of random-like excursions, the diurnal cycle is much better manifested in the SCM ensemble meanmanifested in the SCM ensemble mean
Kappa-TCC diagramsKappa-TCC diagrams
Seem to be an effective tool for characterizing and inter-comparing Seem to be an effective tool for characterizing and inter-comparing the behavior of fast parameterized physics across a hierarchy of the behavior of fast parameterized physics across a hierarchy of simulationssimulations
In these diagrams, SCM results for transition cases are In these diagrams, SCM results for transition cases are representative of GCM behavior (attribution)representative of GCM behavior (attribution)
(How) would these diagrams change in future climate?(How) would these diagrams change in future climate?
Outlook
With the kappa-TCC analysis of single SCM, ensemble SCM and GCM With the kappa-TCC analysis of single SCM, ensemble SCM and GCM results there is enough material to make an interesting & results there is enough material to make an interesting & innovating inter-comparison paper.innovating inter-comparison paper.
Things still to do:Things still to do:
i) ASTEX spin-up issue i) ASTEX spin-up issue Fix? Reruns required? Fix? Reruns required?
ii) Forcings and boundary conditions for 497 transitionsii) Forcings and boundary conditions for 497 transitions
Available to everyone who is interested in simulatingAvailable to everyone who is interested in simulating To be put on To be put on www.euclipse.eu
iii) CMIP5 station data for the remaining three CMIP5 models participating in iii) CMIP5 station data for the remaining three CMIP5 models participating in EUCLIPSEEUCLIPSE
EC-Earth, CNRM (only profiles) and LMDEC-Earth, CNRM (only profiles) and LMD
iv) Add kappa-TCC derived from observations? Is this possible (jumps)?iv) Add kappa-TCC derived from observations? Is this possible (jumps)?