CONCLUSIONS OF THE UTRECHT WORKSHOP

PARALLEL SESSIONS S1 & S2: Individual Presentations of the participating institutes with a focus on the evaluation and improvements of physical parameterisations in the SCM/RCM/GCM, plans for the coming year and assessments on participation with upcoming 3d-GCM cases, Publications.

S1: 9:00-13:00 Stratocumulus

Ø LES and SCM Simulations of FIRE stratocumulus case: A sensitivity study

(A. Chlond and F. Mueller, MPI Hamburg)

Ø HIRLAM-SCM results (J. Calvo & D. Olmeda, INM)

Ø The impact of changing the large-scale horizontal advection in LES and SCM

(P. Goggins UKMO)

Ø Impacts of variability in subsidence on stratocumulus (M. Köhler ECMWF)

Ø Stratocumulus convection in the Arpege Climate model: 1D and 3D evaluation of 2 PBL models (H. Grenier, CNRM/GAME).

Ø LES of CBL and stratocumulus on a large domain (S. de Roode IMAU)

Ø Discussion, Action items, Future plans

S2: 9:00-13:00 Humidity Sensitivity Case

Ø Overview of results (S. Derbyshire, UKMO)

Ø Results from CRM (JL Redelsperger, CNRM/GAME)

Ø Results from SCM & GCM at LMD & issues on parametrization (JY Grandpeix LMD)

Ø ARPEGE SCM and GCM results and Improvements (JM Piriou CNRM/GAME)

Ø Discussion, Action items, Next steps, Future plans for SCM and GCM, Publications

S1: 14:00-19:00 Shallow Cumulus

Ø SCM Simulations with ECHAM of the ARM diurnal cycle (A. Chlond & F. Mueller, MPI )

Ø SCM simulations with the 1D-HIRLAM model (D. Olmeda & J. Calvo, INM)

Ø Statistical Cloud scheme in 1D-RACMO (G. Lenderink KNMI)

Ø Evaluation of Mass flux closures in 1D-RACMO (R. Neggers KNMI)

Ø The French Climat/Arpege SCM: The work done on conserved variables. (P. Marquet CNRM/GAME, presented by H. Grenier CNRM/GAME)

Ø A multiple mass-flux scheme for dry and moist convection: tests on BOMEX and ARM. (S. Cheinet LMD)

Ø Trigger function in the ECMWF-model. (A.P. Siebesma & C. Jakob KNMI, BMRC, ECMWF)

Ø Open Discussion, Action items, Future plans, Publications

S2 14:00-19:00 Diurnal Cycle Deep Convection

Ø Status of the case and results from 5 different standard RCM/GCM (F. Guichard CNRM/GAME)

Ø The predictability of the development of deep convection: Some very preliminary results. (J. Petch UK Met. Office) (Presented by S. Derbyshire)

Ø Diagnostic analysis of the cases (F.Guichard CNRM/GAME)

Ø 1D simulations of the diurnal cycle of land based deep convection with the Kain-Fritsch scheme in the HIRLAM 1D model (C. Jones SMHI)

Ø Boundary layer/Deep convection coupling: impact on the phase and intensity

of the diurnal cycle of deep convection, New concepts. (R. Tailleux LMD)

Ø Analysis of thermodynamical (in)stability of CRM simulations (JP Chaboureau CNRM/GAME)

Ø Results from ECMWF 1D model (JP Chaboureau & M. Koelher CNRM/GAME, ECMWF)

Ø Open Discussion, Action items, Future plans, Publications , Shallow-Deep cloud bridge

Wednesday 10 April (University of Utrecht, IMAU)

PLENARY SESSION:

Conclusions, issues and plans from parallel sessions

Ø Idealized case on humidity sensitivity : S. Derbyshire

Ø Diurnal Cycle of deep convection. F. Guichard

Ø Diurnal Cycle of stratocumulus. S. De Roode

Ø Diurnal cycle of shallow cumulus: G.Lenderink

Reports from the GCM PACIFIC case.

Ø Overview of the Pacific-Case and Compilation of the Results from 8 RCM/GCM groups

(A.P. Siebesma & G. Lenderink KNMI)

Ø 3d-GCM results with the ARPEGE Model (H. Grenier CNRM/GAME)

Ø 3d-GCM results with the Met.Office Model (A. Lock, S. Debyshire UKMO)

Ø 3d-LAM results with RACMO (G. Lenderink, KNMI):

Ø 3d-GCM results with ECMWF Model (M. Kohler, ECMWF)

Ø 3d-GCM results with the MPI Model: ensemble runs ( ???, MPI)

Ø Discussions on Future of the GCM Pacific Case

General discussion on future EUROCS plans, future workshop, cost statement, publications.

The After-EUROCS

Ø The CLIWA-NET EU project (E. Van Meeijard, KNMI)

Ø The CLOUDNET EU project (KNMI)

Ø The VI framework and the different possibilities (JL Redelsperger)

General conclusions

The general conclusion from the workshop was that beyond the 2^nd Year report, a lot of promising works are going on. A general feeling was also that more and more theoretical and physical insights are brought in the evaluation and improvement of cloud schemes.

During the final discussion, we discussed about the next workshop, the publications and the after-EUROCS. It was decided:

To have the next workshop at the end of November 2002 at INM (Madrid) (hosted by Javier Calvo). This workshop will be largely broadcasted towards the international GCSS community. The format of workshop will consist in plenary sessions.
To explore the possibility to have a special issue dedicated to EUROCS (Quaterly Journal of RMS, JGR, ...). More than 12 papers were yet identified. A referring EUROCS committed should be set up to assure the quality of submitted papers.
To explore the after-EUROCS in the coming months. Presentations of 2 EU projects CLIWA-NET and CLOUDNET at the end of workshop given us the opportunity to discuss the after-EUROCS. EUCREM (led by K. Browning) and EUROCS have successfully brought together a critical mass of the European scientific community working close together on the problem of cloud representation in NWP/Climate models. It is crucial to maintain this unique and powerful organization. The different possibilities in the VI EU framework were discussed and several short-term actions decided.

Stratocumulus

(Lead by P. Duynkerke, S. De Roode, H. Grenier)

LES

The LES models all capture the strong diurnal variation in LWP due to the forcing imposed by the shortwave heating of the cloud layer. Like the observations the maximum cloud thickness is found during the night, and the cloud deck gradually thins until noon. The turbulence structure of the boundary layer during nighttime and daytime agree rather well with the observations. However, small differences in the entrainment rates are the major cause of the differences in the simulated diurnal cycle of the liquid water path.

Current model results of a sensitivity test in which the mean subsidence rate has been changed will be analyzed in detail. A few new sensitivity tests will be performed. One includes the diurnal variation of the subsidence rate, which will be prescribed on the basis of ECMWF results recently analyzed by Martin Köhler. It is aimed to run this simulation for 5-7 days.

The set-up of the stratocumulus case, the available observations and the LES results will be compiled by the IMAU (S. De Roode) and will be reported in a scientific paper (Duynkerke et al.).

SCM results

There is broad disagreement in the SCM results. Some models predict a cloud layer that gradually dissipates, whereas other models maintain a solid cloud layer. It is currently unclear how to explain these differences. The analysis of the model results is complicated by the different physical packages that have been used. It has been decided that the SCMs participating in the intercomparison for the FIRE 1 case would re-run the case and perform five sets of simulation according to the following procedure:

All single-column modelers should do a 'STANDARD' simulation. In STANDARD 1A (1B) convection routines must be switched off (on). STANDARD 1A and 1B should exclude precipitation, and the longwave and shortwave codes should be programmed according to the set-up proposed by Peter Duynkerke for this case. This approach facilitates a direct comparison with the LES results, and these two simulations are intended to verify the performance of the PBL models at representing Sc convection excluding all other model components but vertical advection. As the PBL gets decoupled during the day, it is relevant to evaluate also Cu convection schemes in this context.
Note that these SCM simulations must be done both on a coarse vertical resolution as is currently implemented in the ECMWF model (60 levels vertical resolution), and on a finer vertical resolution. The latter vertical grid set-up will be provided by Geert Lenderink from the KNMI, and this experiment is intended to minimize the effect of numerical errors. Possibly we can show that stratocumulus-topped boundary layers can be more accurately represented if one increases the number of vertical levels in the boundary layer in GCMs.
Lastly, the model should be run with the physics package currently implemented in the GCM/RCM ('OPERATIONAL' simulation). We wish to evaluate the performance of the whole SCM representative of the physical package currently used in climate or NWP models. Therefore, modellers will run the SCM at the standard vertical resolution used in the 3D host model. Comparison between this simulation and the previous ones should enable us to figure out the role of radiation schemes, microphysical schemes and vertical resolution in the SCM performance on Sc convection.
These five sets of simulation are a minimum requirement intended to provide to participants a basis to understand model results and draw conclusions regarding Sc convection using K-diffusion schemes. From this firm ground, everyone is welcome to provide further simulations including his own improvements in terms of physical parameterizations.

The model results must be submitted to the IMAU before 15 July 2002 (jonker@phys.uu.nl). Herve Grenier has offered to take the lead in the analysis of the SCM results. A modified data protocol for the SCM results at http://www.phys.uu.nl/~roode/EUROCS/SCM_format.htm. Any comments on this format are welcome.

The SCM participants should summarize the following issues (1 page) to send to Herve Grenier (Herve.Grenier@cnrm.meteo.fr)

What are the positive/negative aspects of your SCM inter-comparison with LES; which aspects of the physics are causing this?
Sensitivity analysis of your SCM using different initial inversion strengths, tendencies, etc.
Improvements/new aspects of physics and/or numerics in your SCM.

GCM/RCM

The stratocumulus working group will join the regional climate simulation experiment set-up by Christian Jakob and Pier Siebesma. This case includes both stratocumulus and cumulus, and their transition (see below “Pacific intercomparison case”).

Diurnal cycle of shallow cumulus

(Lead by Pier Siebesma and Geert Lenderink)

Following is a summary of the updates of the various participating models on the Single Column Model (SCM) studies such as presented at the workshop. The LES results have been yet published and the GCM/RCM part is mainly discussed below (Pacific intercomparison case).

ARPEGE-climate version

The turbulence scheme has been replaced by a new scheme based on prognostic turbulent kinetic energy. This has reduced the amount of noise significantly. In addition, the Kain-Fritsch (KF) convection scheme (similar to the one used in MESO-NH) has been introduced, and changes to the condensation/evaporation of cloud liquid water have been made. Together, these changes led to a significantly better simulation of the diurnal cycle of Cu.

ECHAM

A new statistical cloud scheme has been introduced. Also changes in the numerical time integration have been made, and changes in the mixing formulation of the massflux convection scheme have been attempted. However, so far this did not lead to a significant improvement of the simulation, possibly caused by the noisy behavior of the turbulence scheme.

HIRLAM-SMHI

The cloud condensation and convection scheme (STRACO) has been replaced by a new cloud condensation scheme (Rasch-Kristjansson) and the KF convection scheme (with adaptations made by Colin Jones from the Rossby Climate Center). This reduced the amount of clouds from fully cloud covered in the STRACO version to realistic values (close to 20 %) with the new schemes.

RACMO

The formulation of the closure of the massflux scheme has been investigated. A closure based on the convective turbulent velocity scale of the subcloud layer combined with the cloud fraction seems to work very well, with a good timing of diurnal cycle of the clouds. A simple statistical cloud scheme in RACMO gives realistic values for the cloud cover.

ECMWF

The closure of the convection scheme has already shown to improve the simulation significantly. In addition, the triggering function for deep and shallow convection has been investigated and its positive impact in the full 3d_GCM has been demonstrated. Especially the cold temperature bias near the tropopause has been reduced substantially.

ACTIONS

HIRLAM, ARPEGE, ECHAM and ECMWF will send in their latest SCM updates to Geert Lenderink before June 1st. After this date the results will be documented in an intercomparison paper.

Diurnal cycle of continental deep convection

(Lead by Francoise Guichard & Jon Petch)

An idealized case has been designed from the ARM observed dataset, in order to address the abilities of models to represent the diurnal cycle of deep convection over land. 5 SCMs and 3 SCMs have been working on this case study and results from most of them have already been provided. Some additional work is still needed for improving the case (in particular replacing the actual surface heat fluxes by more realistic ones which have been derived directly from observations) and checking the consistency of model results (initial conditions, forcing...).

However, the key point is that the analysis of both the GCSS and idealized cases show that better results and consistency are found among CRMs than SCMs. It also appears that deep convection occurs earlier in many SCMs than it does in CRMs. When data on each time step were available (instead of 3-h mean values), it was found that convection was "directly" considered as deep in SCM runs, without any transition regime involving shallow cumulus (in contrast with CRM realizations) and that several fields (e.g. rainfall, cloud water path, cloud cover...) were quite noisy because of a frequent on/off behavior of the parameterizations (a feature improved in the new SCM version of the UK Met Office compared to the previous one for this case).

We also began to investigate the ability of models to represent the diurnal cycle of atmospheric stability, closely linked to the surface and boundary layer time evolution (F. Guichard). It shows distinct behaviors of the models, in particular for the CIN weaker in SCMs than in CRMs. This suggests that the links between convection and these parameters are treated differently in these 2 types of model. In parallel, more sophisticated stability parameters such as GCAPE and GCIN have been developed and their analysis, in particular with respect to convective regimes, is currently under study at LMD (R. Tailleux).

The development of the boundary layer properties (e.g. BL height) was found closely linked to surface heat fluxes values, through their interactions with the boundary layer scheme in the ARPEGE climate SCM (H. Grenier). Complex sensitivities of deep convection sequences to the triggering criteria and to the downdraught effects were highlighted with the HIRLAM SCM (C. Jones). That shows the essential role of the triggering function and the strong impact of parameterized downdraughts, through large rain evaporation rates, which will be contrasted with CRM ones.

A sensitivity of 2D CRM runs (UK Met Office model, J. Petch presented by S. Derbyshire) to the initial conditions showed that the amount of rain could significantly vary from one run to the other, depending on the initial random noise. The timing of rain was a more robust feature (afternoon/evening). This sensitivity could be linked to the relatively small size of the domain coupled to the relatively weak forcing of the idealized case. When deep convection occurred in the form of 2 deep cells developing as far away as possible from each other, rainfall rates were high and the opposite for cells developing close to each other, suppressing each other. Indeed, the scatter between the runs substantially decreased with more realistic enhanced surface heat fluxes. The analysis also pointed to the impact of the convection sequence on one day on the one happening the following day within this idealized framework, a feature also shared with MesoNH CRM runs. Based on a sensitivity study to the horizontal resolution for the "shallow convection" case of EUROCS (thanks to A. Brown), some pilot high resolution 3D runs (100m & 250m) have been performed with the same model (Met Office CRM, J. Petch) for the idealized case. They document the initial dry and shallow convective phases prior to the onset of deep convection, and should be helpful for in-depth investigations of these transition phases.

MesoNH CRM runs (CNRM J.-P. Chaboureau) have been analyzed in terms of CAPE/CIN, theta_v and BL height time variations, showing that the model is able to reproduce a diurnal cycle of these parameters, as observed. A sensitivity study to the resolution in 2D did not show a systematic delay of rainfall with decreasing resolution (except at 4 km resolution). These distinct sensitivities to the resolution between the CNRM & Met Office CRMs are possibly linked to different treatments of subgrid scale microphysics & turbulence schemes. Preliminary low resolution 3D CRM runs were also presented by J. -P. Chaboureau. ECWMF SCM runs performed by JP Chaboureau and M. Koehler at the ECMWF show the same "too early timing of rainfall", as pointed out by C. Jakob during previous EUROCS workshops as a characteristic behavior of the ECMWF SCM.

ACTIONS

During the course of the remaining 11 months, we plan to focus on:

(a) transition regimes (e.g., transition from shallow to deep convection)

(b) triggering aspects (links with CIN)

(d) other parameterization issues.

CRMs/LES

CRM results must be further analyzed, in order to describe/explain/understand the different convective phases, from (a) dry to (b) shallow and finally (c) deep convection. We wish to raise the following questions:

Why do we get a shallow phase lasting 1 to a few hours?

Which factors control the length of this phase? (E.g. what is the role of the moisture fields?)
What is the role of downdraughts on the boundary layer fields and on stability parameters?

In parallel, we wish to extend the work of J. Petch on high resolution runs to larger domains. This will allow including also the deep convective phase in one single simulation. This work is closely linked to GCSS WG1-WG4 activities and questionings towards a common case study.

SCMs/RCMs/GCMs

On the base of SCM results, we wish to address the following questions:

What is the ability of SCMs to reproduce the diurnal cycle of atmospheric stability?
Why is the CIN so low in several SCMs?
Is there a link with the treatment of downdrafts?
How strong are the links between convection and the atmospheric stability in SCMs and how do they compare with the ones in CRMs?
How do shallow and deep convections interact (or not) during the transition regimes?
What are the reasons for the "on/off" behavior of convection in SCMs?
How does the behavior of SCMs differ between the shallow cumulus case and this idealized case when deep convection is not turned off?

WORK TO BE DONE

To address these questions, we need to complete and document model datasets (deadline June 1^st) this includes the following:

To provide time-step values instead of 3-h mean values for both time and time-height series (partly done)
To provide time-height series of Q1, Q2 and evaporation of rain for comparison between SCMs & CRMs (partly done)
To collect information concerning convection schemes (mass flux schemes, closure, triggering criteria, entrainment formulation, references...) and the way boundary layer, shallow and deep convection schemes interact together.
Participants need to send feedbacks/expertise to Francoise Guichard on SCMs runs in the form of short texts (e.g., larger sensitivities they noticed, answers to the questions above...).

Final runs will be performed with more realistic surface heat flux dataset and the stability study must be completed. It will also include an analysis of theta_v profiles and diurnal variations, BL height, condensation level and cloud parameters (diagnostics partly computed from the time-height series provided by the participants).

We are interested in the results from SCMs runs of the shallow cumulus case when deep convection is turned on. We wish to take advantage of the work already done by participants to the shallow cumulus case in order to contrast these 2 realizations (shallow cumulus case and idealized case). These two cases differ mostly by their initial conditions and possibly by large-scale advection of sensible heat and moisture. This analysis should be possible for at least 3 SCMs which have been run for both cases (ECMWF, SMHI and CNRM).

We propose to perform systematic sensitivity tests of SCM runs, documenting the role of downdraughts and trigger criteria, (turned on or off). A set of runs would be proposed.

Finally the impact of modifications/improvements of parameterizations in RCM/GCM runs should be address in terms of their impact on the representation of the diurnal cycle of deep convection.

PAPERS

Several people wished to write papers in connection with the work, which has been done or is currently in progress within EUROCS. This includes:

"A general paper", by Guichard, Petch et al. reporting our motivation (illustrated by GCM diagnostics), presenting the idealized case, both CRM/SCM results (for standard runs). It could discuss in a broad sense the cumulus cloud field (cloud top height, cloud water path & fraction), the transition regime, the timing of rainfall, the diurnal cycle of stability, the impact of downdrafts & the role of triggering criteria. For this paper, a large part of the work has already been done.
"A SCMs or SCMs/GCMs paper” (Petch, Guichard et al). This paper largely relies on what will be done in the following year. It could explore specific areas of convection parameterizations through the previously proposed set of sensitivity tests, i.e. triggering, downdrafts, transition regimes.
"A CRM analysis of the idealized case" by Chaboureau et al. addressing the issues raised above (work in progress)
A "GCAPE/GCIN analysis" by R. Tailleux: which would be a more theoretical paper towards a discussion of stability indexes (work in progress)
"A joint CRM/SCM analysis" by Chaboureau & Koehler based on the CNRM-ECMWF collaboration which began at the beginning of this year (work in progress)
More?

Idealized humidity case

(Lead by Steve Derbyshire)

Background:

During year 2, considerable progress has been made with the idealized humidity case. The case has been run with many more models (now 2 CRMs and 5 SCMs). This workshop was the first opportunity to discuss those new results properly, and to hear about first results with GCMs.

Presentations:

Steve Derbyshire (SDB) presented the overall case summary and this presentation (as given at NASA-Goddard for the Cumulus parameterization workshop) was broadly accepted
Jean-Luc Redelsperger (JLR) presented his CRM runs with the COME-NH model, performed with the help of F. Guichard. He stressed the importance of the initial profiles in the boundary layer to decrease the spin up time, the general agreement with the CRM Met Office model, and the 2 distinct convective regimes found in his simulations.
Jean-Yves Grandpeix presented the results of changing the mixing pdf in the Emanuel scheme. Although the impact of this in the SCM is rather subtle, its impact in the GCM appears significant, and potentially beneficial.
Jean-Marcel Piriou described the recent development of the ARPEGE-NWP mass-flux scheme following previous comparisons with Met Office CRM work by Hugh Swann. The scheme includes a term, which explicitly relaxes the dilute parcel values towards a notional adiabatic parcel, in addition to a conventional entrainment term. The scheme has been adopted operationally because of its benefits to global modelling. The scheme seems able to model the present case rather well.
Results from Peter Bechtold and Isabelle Beau were briefly considered, but not discussed in detail as they were not able to attend the workshop
It appears that the latest version of the ARPEGE-CLIMAT SCM (with extensive revisions to convection and turbulent mixing suggested by J.F.Gueremy and H.Grenier) brings this SCM generally closer towards the CRM behavior, although there is a question about possible numerical noise in some of the mass-flux profiles.

Paper:

An outline for a paper based on this case was discussed and broadly accepted. We have a lot more detailed work to do on the CRM-SCM comparisons but the process of drafting a paper should help to drive us forward.

Summary of outcomes:

Considerable discussion of theoretical and conceptual side of convection parameterization (as also in the diurnal cycle session). This discussion is a good feature of EUROCS.
The agreement between the CRMs was considered good, although JLR model showed signs of a sharper transition from shallow to deep convection (possibly reflecting the impact of glaciation when cloud tops become sufficiently cold)
The SCMs showed a wide range of behavior. Various sensitivity tests have been performed.
Main diagnostics for comparison: Updraught mass fluxes, Q1,Q2, boundary layer, surface precipitation, ….
It was accepted that (as assumed in the case specification) it is reasonable for moisture convergence at upper levels to play a significant role in the overall moisture budget

Actions:

· SHD starts drafting paper on SCM/CRM work

· SHD looks at BL comparisons

· SHD contacts Julia Slingo re GCMs, climate and moisture sensitivity

· SHD to update web pages

· Optional additional case with intermediate 60% humidity

· GCM modellers evaluate GCM sensitivity of existing SCM modifications

· To produce diagnostics based on zonal qv and others (JMP and JYG please liaise and tell to SDB your thoughts)

· To aim for an interim report on GCM tests by July (as requested by JLR)

Afternotes:

· Peter Bechtold has submitted further results which SBD have not yet had a chance to analyze. Although not a funded participant, he remains very interested in the case.

· SDB would like to include some SCM sensitivity tests in the paper, not just “best” results. This will help to show that the case can be used to develop the SCMs. So please don’t write off your first versions completely!

PACIFIC Intercomparison Case for 3d GCM's

(Lead by Pier Siebesma and Geert Lenderink)

General conclusions

Five Models within the EUROCS framework (ARPEGE,ECMWF,RACMO,UKMO,ECHAM)and one external (Japanese Climate Model (JCM) have participated in an intercomparison study for GCM's over the Northern Pacific Ocean for July 1998. From this studies the following general conclusions can been drawn:

· All models strongly overestimate the liquid water path (LWP) and cloud cover (CC) in the trade wind regions.

· Most models strongly underestimate LWP and CC in the Stratocumulus regions.

· Strong differences in the subsidence rates are encountered.

More model specific conclusions:

ARPEGE-climate: Too deep PBL-height, too high integrated water vapour content (IWV).

ECHAM: Too high sea surface temperature (SST), too high IWV

RACMO, JCM: Overall too low cloud cover (CC) and too low LWP.

UKMO: Too high IWV and Outgoing Longwave Radiation (OLR) and too low LWP in the intertropical convergence zone (ITCZ).

General discussion

During the individual presentations a couple of drawbacks have been noticed:

· A monthly mean over only one strip does not have sufficient statistics. For instance, the difference between the monthly means over two neighboring strips within one model is of similar magnitude as the difference between the monthly means over the same strip for 2 different models (UKMO). Similarly the ECHAM model made several runs with initial conditions with different random perturbations also gave substantial differences between the ensemble members. These results suggest that longer averaging periods and/or areas should be used.

· Separate monthly means over nighttime and daytime give strong differences in cloud parameters such as cloud cover, liquid water path and ice path (RACMO). To address this more specifically, different monthly means over daytime and nighttime should be requested as output.

· The ITCZ is just at the edge of the required diagonal. Extension of the diagonal further to the south would allow a better analysis of the ITCZ.

· The prescribed diagonal strip crosses Hawaii. This was done deliberately since the presence of observational data here. Unfortunately this hinders the analysis since some models do resolve Hawaii while others do not. These differences show up in the subsidence fields and complicate the intercomparison between the models in the vicinity of Hawaii.

As a result of these drawbacks the following changes for the intercomparison are proposed:

Extension of the simulation period to 3 months: June/July/August 1998
Monthly means for different times: 0hr, 6hr, 12hr, 18hr
Extend the diagonal further south to 160 East, -5 South
Request to send in data for the complete rectangular spanned by the extended diagonal. This way it can be inspected how representative one single strip is for a 3 month average and the strip can be made thicker if necessary.

Actions

All the participants will be contacted and asked if they are able to redo the Pacific simulations with the modifications such as sketched above. If a majority reacts positively, results should be resubmitted before Sept 15 so that the results can be presented during the Madrid workshop in mid-November.