Joint EUROCS-GCSS Workshop
Lisbon, Gulbenkian Foundation
28 - 31 May 2001
The workshop was hosted by
Pr. P. Miranda and funded by the Gulbenkian Foundation, the WCRP and EUROCS (funded by EU Community).
Local organizer: Pedro Miranda (Centro de Geofisica da
Univ. Lisboa)
Scientific organizers:
Jean-Luc Redelsperger (CNRM/GAME) , Chairman of EUROCS
Peter Duynkerke (IMAU),
Chairman of WG1/GCSS
Wojciech W. Grabowski (NCAR),
Chairman of WG4/GCSS
Diurnal Cycle of Deep Convection
All results
of the stratocumulus intercomparison of EUROCS/WG1 have been put on the web: http://www.phys.uu.nl/~wwwimau/EUROCS/eurocs.html
Please
have a look and give your comments.
Shallow Cumulus
Diurnal cycle of shallow cumulus convection.
1. Introduction
The diurnal cycle of
shallow Cumulus clouds has been studied using 3 Large Eddy Simulation (LES)
models and 7 Single Column Models (SCM's). The case has been set up by Andy Brown
(UK Met. Office), and is a idealization of observations made at the Southern
Great Plains ARM site on June 21 1997. For a complete case description with
results from both LES models and SCM's we refer to http://www.knmi.nl/samenw/eurocs
2. Large Eddy Simulation studies.
Within the EUROCS framework three groups (MPI, UKMO and KNMI) have
submitted LES results. The most notable conclusions from these simulations are:
i) LES models are well
capable of reproducing realistically the diurnal cycle of the shallow cumulus
topped boundary layer. The cloud cover amounts 20-30 % with a maximum around
noon. Cloud base is around 1000 m, the highest cloud tops are at 2500 m. These
results are in agreement with cloud observations during that day.
ii) Results of previous steady state cases of cumulus over sea still apply to the present case where a strong diurnal cycle is present .
iii) The sub-cloud layer structure seems hardly affected by the clouds on top.
3. Single Column Model studies.
Single column model simulation were done by 6 groups:
|
ECMWF |
(ECMWF-model) |
|
INM |
(HIRLAM and MESO-INM model *1) |
|
KNMI |
(RACMO model) |
|
MPI |
(ECHAM4) |
|
UL |
(MESO-NH) |
|
LMD |
(CT01 *2). |
The most notable results from the SCM's are:
i) Most models produce too high values for both cloud liquid water (2 to 5 times larger than LES and observation) and cloud cover (50-70 %).
ii) Most models have
difficulties representing the diurnal cycle. Two models already have clouds
very early in the simulation. Five models have difficulties with dissolving the
clouds after sunset.
iii) Most models under predict the height of the cloud top, some of them even with 1000 m.
iv) The two models with K diffusion performing mixing in the
cloud layer give reasonable profiles of temperature and humidity. On the
downside, both are rather noisy in the cloud layer.
v) Two of the models with a mass flux scheme give a too
strong drying at cloud base and a too strong moistening at the inversion. The
other mass flux schemes give reasonable to rather good profiles and temperature
and humidity.
vi) There is a surprisingly large difference in the wind speed between the different models.
The overall conclusion is that most state-of-the-art climate and weather
prediction models fail in reproducing realistically a simple diurnal cycle of a
convective day in the presence of fair weather cumuli.
......................................................................
Footnotes
*1 = MESO-NH model with modifications made by INM
*2 =
research model of S. Cheinet and J. Teixera
Diurnal Cycle of Deep
Convection
1. An
introduction by F. Guichard was given which provided an overview of the
problems of modelling the diurnal cycle of convection in large-scale models.
Results provided by J. Slingo and G.-Y. Yang [University of Reading], C. Jacob [ECMWF] and J.-F. Royer et al. [Météo-France/CNRM] showed that the
diurnal cycle of convection in the Tropics was significantly out of phase in
the Met Office, ECMWF and Météo-France global models respectively.
Output from the LMD climate model, provided by A. Lahellec and R.
Tailleux, also point to the same default in the LMD climate model over Northern
America. In contrast, results from C. Jones [Swedish Meteorological and
Hydrological Institute] showed that a mesoscale
limited area model (RCA05) was able to
reproduce several aspects of the observed diurnal cycle of convection over
Northern America, but the runs were performed at higher resolution than typical
in a GCM.
The introduction also described the basic case
being used in this study. The case is a 4-day convective period over the
Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site in the
central U.S for which a large amount of observations is available. This case is
also being used by the ARM Single Column Modelling (SCM) working group and GCSS WG 4.
The discussion focused on the first 24-hour period of the case because much of
the later periods were dominated by large-scale forcing terms not related to
the diurnal cycle. This issue had been raised at the previous EUROCS meeting
(January, ECMWF) where suggestions had been made that the experiments would
need refining for an in-depth investigation of the diurnal cycle of convection
over land.
J. Petch
and F. Guichard both presented results showing tests
using more idealized forcing which could form the basis for a more focused
inter-comparison case. Results from the Met Office SCM and CRM showed that
using the surface forcing alone from the ARM SGP case with interactive
radiation produced a diurnal cycle of precipitating convection. Results
from the CNRM CRM showed that a composite of the first days forcing, run for 4
days, also produced a diurnal cycle of convective rainfall. It also showed
that the effect of the large-scale forcing was to delay the initiation of
convection on each of the 4 days. Sensitivity to the start time of the
simulation was shown, with the main influence on the first 2 days.
2.
CRMs which have contributed to the inter-comparison include one from the
Met Office and two from CNRM. Comparisons of these were shown and the main
point stressed was the problems in predicting the onset of precipitation during
the first day. These problems were the focus of more detailed talks. J. Petch
showed that an increase in horizontal and vertical resolution could improve the
timing of the first precipitation event. Transport terms from the model showed
that a horizontal grid length of around 250 m was required to resolve the major
eddies involved in transporting water vapour into the free troposphere, this
also resulted in a much improved prediction of the timing of the first
precipitation event. F. Guichard presented work of M.
Tomasini which showed that timing could also be improved through improvement of
the sub-grid turbulence scheme and the introduction of a sub grid condensation
scheme.
3.
SCMs which have contributed to the inter-comparison include models from
LMD, Met Office, CNRM and ECMWF. All models where showed to precipitate too
early when compared to observations of the first days rain event. This is an
encouraging result because it suggests that SCM simulations of the first day of
ARM provide a relevant case study for investigating the problems seen in the
full 3-D versions of the large-scale models. In addition, R. Tailleux presented
sensitivity test to the triggering function and P. Bechtold showed some results
documenting the SCM methodology.
4.
Following discussions, it was decided to form a new inter-comparison case which
will be more focused on the diurnal cycle. The simulation would begin close to
sunrise and be a 4 day composite of the first day of the ARM case including the
large-scale forcing. To avoid resolution problems discussed, the CRMs
should use a horizontal grid length of 250 m and no larger than 250 m
anywhere in the vertical. The domain size would be halved to 250 km and the
CRMs would include their own interactive radiation. Several diagnostics will be
added to better describe transport terms and radiation fields. It is also
planned to use the first part of this run with a smaller domain and higher
resolution for an in-depth investigation of CRM performances regarding the
representation of the earlier stage of convection. This work involves
interactions with the group working on the diurnal cycle of cumulus.
Finally, C. Jakob, W.-K. Tao and J.-P. Lafore presented interesting cases from
LBA, TRMM (Southern America) and HAPEX (Western Africa) which could be
exploited in the longer term.
The idealized humidity convection case is designed to show up the
impacts of
mid-tropospheric humidity on convection. These impacts are thought to
play a role in important climate mechanisms such as the Madden-Julian
Oscillation, in the differences
between maritime and continental convection, and possibly in the diurnal cycle
of convection. Moreover they provide an important conceptual test of the
mixing-assumptions in convection scheme. The case-setup for the process-model
comparison (CRM/SCMs) is designed to maximize the relevance of those
comparisons to GCM applications.
J.L.Redelsperger first presented some high resolution CRM results from a
TOGA-COARE study (Redelsperger, Parsons, Guichard 2001) in which the moisture
deficit at heights around 2-4km was found to be a key predictor of convective
activity.
S.Derbyshire described the EUROCS inter-comparison case. The case set-up
was based on strong nudging of
mean profiles, appropriate to a situation where the large scales exert a
control on the profiles of the column in question. The target profiles are specified
with a potential temperature difference of 1K across the boundary layer and
potential temperature increasing at 3K/km above the boundary layer. Relative
humidity is specified at 80% in the layer 1-2km and at a value RHt above
2km. (The boundary layer, below 1km, is not normally forced in these runs.)
The Met Office CRM was run for a pilot run A and the main
intercomparison run B. Run A has instantaneous nudging of the mean profiles and
only warm cloud physics. In run A, changes in RHt were able to switch
convection between a non-precipitating shallow convection and vigorous deep
convection. Run B used a nudging timescale of 1 hour and "full"
3-phase cloud physics. In
run B, changes in RHt were again able to switch the convection between almost
non-precipitating shallow convection and vigorous convection, even though the
1hr timescale allows convection to stabilize the mean profiles
significantly.
J.Y.Grandpeix (LMD) described how he was using this case to review
his version of the Emanuel scheme.
The scheme uses a pdf for mixing between cloudy and environmental air, and this
pdf can be tuned towards the CRM results. The set-up of this case allows the
CRM and SCMs to develop their own characteristic mass-flux profiles according
to the environment, whereas in some other set-ups the mass-flux
profile and precipitation are controlled almost directly by the forcing.
S.Derbyshire presented a synthesis of the CRM and SCM results currently
available, concentrating on the dependence of surface precipitation on RHt. The
CRM gave surface precipitation increasing roughly linearly with RHt from 0.05
mm/hr at RHt=25% to 1.5 mm/hr at RHt=90%. The Met Office SCM was able to
capture the trend with RHt but precipitated too strongly in the drier cases.
The LMD SCM
(original formulation) had
only a weak trend with RHt, but this trend could be increased by changes to the
mixing-formulation.
In addition to the originally funded participants, a number of others
expressed interest in running the idealized humidity case, including J.M.Piriou
(Meteo France), P.Bechtold (OMP), C.Jakob (ECMWF) and P.Soares (University of
Lisbon) with SCMs. J.L.Redelsperger (CNRM) hopes to run the case with his CRM.
Furthermore W.Grabowski (NCAR) is interested in running different cases with a
similar methodology, and is considering proposing their adoption under
GCSS-WG4.
The immediate priority is to synthesize machinable results from the
various models. SCM modellers are asked to send in results in GCSS format by
the end of July. S.Derbyshire hopes to go to Boulder in October and present a
synthesis. The focus will then move onto consideration of sensitivity tests
within GCMs.
The joint
EUROCS-GCSS workshop was very successful and participants recommended that we
pursue such joint workshops in the future. In particular, a joint workshop
between PBL-clouds experts and deep convective clouds experts was appreciated
and especially useful. GCSS/WG1 and EUROCS will participate on similar case
studies of PBL clouds for the 2 coming years. Therefore, it was recommended
that joint GCSS/WG1 and EUROCS workshops should definitely continue in future.
People
working on diurnal cycle of deep convective clouds identified 2 tracks. The
fast track is to continue to work on the idealized ARM case. For this case study,
GCSS and EUROCS time frameworks are in phase. The slow track is to develop a
new case study from LBA experiments. As the EUROCS funding ends in February
2003, the LBA case will be not be included in the EUROCS plans. Of course,
EUROCS people are encouraged to participate in this case study.
Finally,
it has been decided that in early 2002 there is to be a second joint
EUROCS-GCSS workshop in Utretcht (hosted by KNMI). The precise date and format
will need to be discussed in the coming months. A GCSS/WG4 workshop will also
occur on 22-24 October 2001 in Boulder, Colorado. The two case studies for deep
convective clouds will be discussed and some EUROCS participants plan to
attend.
Two other
workshops were announced by W.K. Tao. The second TRMM Latent Heating Algorithm
Workshop, organized by M. Moncrieff, A. Hou and W.-K. Tao, is to take place on
October 10-12 2001 at NCAR. A Cumulus Parameterization Mini-Workshop, organized
by W.-K. Tao, D. Starr and Y. Sud, will take place on Nov 13-15 2001 at
NASA/GSFC. For more details on these two latter workshops contact W.K. Tao
directly - tao@agnes.gsfc.nasa.gov