Section 1

March 2000 - 28 February 2001

Web Page: http://www.cnrm.meteo.fr/gcss/EUROCS/EUROCS.html

1.1 Objectives

The project mainly aims to improve the treatment of cloud systems in global and regional climate models, which directly relates to the RTD priority 2.1.3 "improved model treatment of physical processes (in particular clouds)." In addition, benefits will also be gained for hydrology and severe weather issues (RTD 1.5.2 "Improved flood and drought forecasting").

Clouds probably remain the largest source of uncertainty affecting evaluations of climate change in response to anthropogenic change. That explains for a large part why the range of simulated temperature changes in response to a CO2 doubling (1.5 to 4.5 C) is quite invariant for almost 20 years (e.g. Report of Intergovernmental Panel on Climate Change: Dickinson et al. 1996). The recent interest to develop capability to predict regional climate changes stress the importance to better represent clouds in models. For example, the prediction of distributions of mean and extreme precipitation is highly dependent on the cloud representation in models. The climate community must thus give more consideration to the problem of cloud representation in General Circulation Models (GCMs) and Limited Area Models (LAMs) which are applied to climate issues. As most of European models applied to climate issues are also used to produce short and medium range forecasts, prediction of severe weather systems will directly benefit from EUROCS.

EUROCS concentrates its efforts on 4 major and well identified deficiencies of climate models in representing clouds:

· stratocumulus over ocean

· diurnal cycle of cumulus

· diurnal cycle of precipitating deep convection over continents

· sensitivity of deep convection development on the moisture profile.

These issues (identifying our 4 case studies) are considered of great magnitude as they lead to major deficiencies in the predicted global and regional climates. Some of these issues have been addressed in the past, but others as of yet have not.

The specific aims of the project are:

· To produce comprehensive 4-D data sets using several and Large Eddy Simulations (LES) and Cloud Resolving Models (CRMs) on cases defined above which address critical problems associated with the prediction of cloud in regional and global climate and Numerical Weather Prediction (NWP) models.

· To use the LES/CRM data sets to investigate deficiencies in climate and NWP models using 9 different Single Column Models (SCMs) as a test bed. Specific issues to be addressed will be the general failure of climate regional and global models to predict stratocumulus amounts, the diurnal triggering of boundary layer convection and deep precipitating convection over land, and the lack of sensitivity of deep convection development on moisture profile in these models. Once the reasons of deficiency are identified, physically-grounded corrections will be brought in SCMs.

· To improve climate and NWP models ability to represent both the mean structure (horizontally and vertically) and time variability of cloud water and cover for the critical cases defined above. Six different European climate models will be used in EUROCS making it of large benefit to the whole community.

· To bring together a critical mass of the scientific community across Europe working in various areas of Cloud and Climate research, with the focused aim of improving cloud representations in climate and NWP models.

1.2 Participants and their modeling plans

The strategy used in EUROCS to address these issues is based on the use of a hierarchy of models and observations to integrate cloud studies across the full range of scales. Numerical models range from regional and general Climate Models (CMs) through Single Column models (SCMs) to Cloud Resolving Models (CRMs) and Large Eddy Simulations (LES). Observations collected will vary from global satellite measurements to local observations of individual clouds through lidar and millimetric radars. Table 1.1 illustrates how the ten participants are involved in various aspects of this research by listing the models they are using.

Table 1.1 Partners and model involvement.

Partners	CRM/LES	SCM	CM
CNRM-GAME	Yes	Yes	Yes
ECMWF	No	Yes	Yes
UKMO	Yes	Yes	Yes
IMAU	Yes	Yes	No
INM	Yes	Yes	No
SMHI	No	No	Yes
LMD	No	Yes	Yes
MPI	Yes	Yes	Yes
KNMI	Yes	Yes	No
UL	No	Yes	No

1.3 Milestones

To date, the project is on task and progressing as expected. Table 2 lists the deliverables as set up by the project. The expected date and whether that deadline was achieved is noted.

Table 1.2 Project deliverables and current status.

Target Date	Deliverables	Status
April 2000	D1 - Choice and organization of observations of stratocumulus case to initialize and evaluate SCMs and LES D2 - Choice and organization of observations of diurnal shallow convection case to initialize and evaluate SCMs and LES D3 - Choice and organizations of observations of diurnal deep convection case to initialize and evaluate SCMs and CRMs	Done Done Done
June 2000	D4 - Coice of idealized deep convection case with different mid-tropospheric vapor profiles to initialize SCMs and CRMs	Done
February 2001	D5a - Reference simulation with LES of stratocumulus case and inter-comparison with observations	Done
August 2001	D5b - Ensemble simulations to study the sensitivity to initial conditions and forcings	On task
February 2001	D6a - Simulation with LES of diurnal Shallow convection case and inter-comparison with observations	Done
August 2001	D6b - Ensemble simulations to study the sensitivity to initial conditions and forcings	On task
February 2001	D7a - 2D simulation with CRMs of diurnal deep convection case and inter-comparison with observations	Done
June 2001	D7b - 3D simulation with CRMs of diurnal deep convection case and inter-comparison with observations	On task
February 2001	D8 - Simulations of idealized deep convection studying the sensitivity to mid-trospheric humidity	Done
February 2001	D9 - First drafts of articles on EUROCS activity on CRM	Underway
May 2001	D10 - Dataset from LES runs for stratocumulus case to evaluate and to improve SCMs D11 - Dataset from LES runs for diurnal shallow convection case to evaluate and to improve SCMs	On task On task
May 2001	D12a - Dataset from 2D CRM runs for diurnal deep convection case to evaluate and to improve SCMs	On task
August 2001	D12b - Dataset from 3D CRM runs for diurnal deep convection case to evaluate and to improve SCMs	On task
May 2001	D13 - Dataset from CRM run for idealized deep convection case to evaluate and to improve SCMs	On task
February 2002	D14 - Improvement of cloud and boundary layer parameterizations with SCMs to represent stratocumulus D15 - Improvement of cloud and boundary layer parameterizations with SCMs to represent diurnal shallow convection D16 - Improvement of deep cloud parameterizations with SCMs to represent diurnal variations of deep convection D17 - Improvement of deep cloud parameterizations with SCMs to represent effect of various mid-tropospheric humidity profiles D18 - Comparison of characteristics of stratocumulus, diurnal cycle shallow and deep convection as simulated by current GCMs, LAM and available satellite observations D19 - Drafts of articles on EUROCS activity on CRMs and SCMs
April 2002	D20 - Implementation of improved cloud schemes in GCMs
August 2002	D21 - Climate runs with improved cloud schemes in GCMs
October 2002	D22 - New SCM runs in using CRM datasets to fix new potential problems in climate runs with improved cloud schemes in GCMs
Dec. 2002	D23 - New climate runs with last improvements in cloud representations
February 2003	D24 - Comparison of characteristics of stratocumulus, diurnal cycle shallow and deep convection as simulated by improved GCMs, LAM and available satellite observations
February 2003	Drafts of papers in referee journal describing EUROCS results

1.4 Deviations from Work Plan

The following matrices clearly show the progress being made as regard the case studies. Each case study being explored in this project is represented by an individual matrix that shows which partner is in charge of set-up as well as who all other participating partners are. An X indicates that the data for that model has been collected and disseminated. A date shows the expected time for the data associated with those models. It should be noted that the Climate Model diagnosis will not take place until the final stages of the research.

As can be seen with the matrices, all but one partner was on target. Due to unexpected hiring delays INM were unable to collect their CRM/LES data on time. But this did not interfere with the schedules because results from the three other models being run by other participating partners were sufficient. INM's contribution adds to the data base.

STRATOCUMULUS CASE

Group	Case Set-up	CRM/LES Runs	CRM/LES Data Dissemination	SCM Runs	SCM Data Dissemination	CLIMATE MODELS Diagnosis
CNRM-GAME				X	May 2001
ECMWF				X	May 2001	X
UKMO		X	X
IMAU	X	X	X
INM		May 2001*	May 2001*	May 2001	May 2001
SMHI
LMD				X	X
MPI		X	X	X	X
KNMI/UL				X	X

CUMULUS CASE

Group	Case Set-up	CRM/LES Runs	CRM/LES Data Dissemination	SCM Runs	SCM Data Dissemination	CLIMATE MODELS Diagnosis
CNRM-GAME					May 2001
ECMWF				X	X
UKMO		X	X
INM		May 2001*	May 2001*	X	X
SMHI
LMD					May 2001
MPI		X	X		May 2001
KNMI/UL	X	X	X	X	X

* Due to unexpected hiring delays the CRM runs are currently being performed. The runs are expected to be finished by May 2001. Results will be presented at the Lisbon Workshop.

Diurnal Cycle of Deep Convection

Group	Case Set-up	CRM/LES Runs	CRM/LES Data Dissemination	SCM Runs	SCM Data Dissemination	CLIMATE MODELS Diagnosis
CNRM-GAME	X	X	X		May 2001
ECMWF				X	X
UKMO	X	X	X		May 2001
INM					May 2001
SMHI
LMD				X	May 2001

Idealized Humidity Case

Group	Case Set-up	CRM/LES Runs	CRM/LES Data Dissemination	SCM Runs	SCM Data Dissemination	CLIMATE MODELS Diagnosis
CNRM-GAME					May 2001
ECMWF					May 2001
UKMO	X	X	X	X	X
INM					May 2001
SMHI
LMD				X	May 2001

Section 2

EXECUTIVE PUBLISHABLE SUMMARY

Contract n°

EVK2-CT-1999- 00051

Reporting period:

March 2000 – February 2001

Title

EUROCS : EUROPEAN PROJECT on CLOUD SYSTEMS in CLIMATE MODELS

Objectives: EUROCS aims to improve the treatment of cloud systems in global and regional climate models. In addition, benefits will also be gained for hydrology and severe weather issues. Clouds probably remain the largest source of uncertainty affecting evaluations of climate change in response to anthropogenic change. The recent interest to develop capability to predict regional changes of climate, stresses the importance of better representing clouds in models. EUROCS concentrates its efforts on 4 major and well identified deficiencies of climate models: stratocumulus over ocean, diurnal cycle of cumulus, diurnal cycle of precipitating deep convection over continents, and sensitivity of deep convection development on the moisture profile. These issues are considered of great magnitude as they lead to major deficiencies in the predicted global and regional climates. The strategy used to address these issues is based on the use of a hierarchy of models and observations to integrate cloud studies across the full range of scales. Numerical models range from General Circulation Models (GCMs) through Single Column models (SCMs) to Cloud Resolving Models (CRMs) and Large Eddy Simulations (LES).

Scientific achievements

General achievements

-The choice and organization of observations to initialize and evaluate SCMs and LES have been achieved for: stratocumulus, diurnal shallow convection, diurnal deep convection, and idealized deep convection cases with different mid-tropospheric vapor profiles.

-Reference simulations with LES of stratocumulus and cumulus cases and inter-comparison with observations have been successfully performed.

-Two-dimensional simulation with CRMs of diurnal deep convection and inter-comparison with observations and simulations of idealized deep convection studying the sensitivity to mid-tropospheric humidity have been successfully performed

-Preliminary analysis of simulations with SCMs on each case study have allowed to identify quite general deficiencies in standard cloud schemes. That demonstrates that the setup of cases is well suited to address the major issues of climate models in representing clouds.

Detailed achievements

i) Stratocumulus case

The LES models all capture the strong diurnal variation in liquid water path, the observed turbulence profiles and clearly show a decoupled boundary layer during daytime and a well-mixed boundary layer during the night. The diurnal variation in boundary layer properties leads to a strong diurnal variation in the surface fluxes, an effect which will be analysed in detail in combination with the SCM and GCM results.

Preliminary results from three SCMs have been compared with the observations and LES results. Two SCMs predict a too low liquid water path during the night and one predicts a rather large liquid water path but with a timing out of phase compared with the observations.

ii)Cumulus case

The LES models are capable of reproducing realistically the diurnal cycle in the presence of shallow cumulus clouds. LES models do behave similarly and give approximately the same timing as the observations for the onset and offset of the clouds. The cloud cover amount reaches about the same maximum value as observed. The diurnal variation of cloud base height is similar to observations (ceilometer). The sub-cloud layer structure seems hardly affected by the clouds on top, meaning that the issue of the parameterization is less complicated.

Most SCMs produce a total cloud cover of 50-60%, in contrast to the 25% in the LES results. The integrated liquid water path is consequently also high (2-5 times too large). In addition, neither of the SCMs produces a realistic diurnal cycle, in particular the clouds remain after sunset.

iii)Diurnal deep convection

A first survey of GCMs involved in EUROCS have shown that deep convection frequently occurs too early in the daytime. Results from CRMs showed that they compare reasonably well with observations. However, it was difficult to predict the exact occurrence of precipitating events. This type of simulation involves strong interactions with surface and boundary layer processes, i.e. processes that occur on smaller scales than deep convection and thus may not be well resolved by CRMs. Sensitivity tests to the horizontal resolution (from 2km to 250m) have shown that the timing of rainfall and cloud cover is sensitive to these parameters but with an evidence of convergence of results for high resolution . This sensitivity to the resolution could be explained by turbulent flux differences in the boundary layer. The timing of precipitation was also found to be sensitive to the turbulent scheme. These two results suggest that a proper CRM simulation of deep convection over land require high horizontal resolution and/or an advanced turbulent scheme.

Preliminary results of two different SCMs have shown that precipitation occurs too early. These results suggest that this weakness shared by GCMs, could be properly analyzed within the joint CRM/SCM framework.

iv) Sensitivity to humidity profile

A CRM pilot run have confirmed that changes in the mid-tropospheric humidity can switch the convection regime between shallow non-precipitating convection and vigorous deep convection, with changes to the shape of the convection profiles as well as the magnitude and depth. For the formal inter-comparison of models there is also a clear trend for stronger convection in the moister cases, which pushes the temperature profile closer to the moist adiabat. Despite the differences in magnitude and profile shape, the depth of convection was almost the same in all the cases. The CRM simulations converge rapidly to a quasi-steady convective ensemble.

To compare with the pilot CRM, a simple SCM was used. It is based on a highly-simplified plume model, including an "ensemble" option resembling the Arakawa-Schubert ensemble. The ensemble version matches the shapes of the CRM profiles more closely than does the standard single-plume version. Significant variations are found in surface precipitation as a function of humidity, and between different SCMs and the CRM. These differences require further analysis but show the potential of this approach to test the schemes.

Socio-economic relevance and policy implications

The economic and social impacts of this project will result in the expected improvements in cloud parameterizations. In improving cloud parameterizations, we would improve cloud and precipitation events of importance to human activities, such as agriculture, transport, some kinds of industry, as well as the atmospheric pollution forecast which depends on the ability to make precise weather forecasts. The study on the deep convection case should also contribute to improve the forecast of summer convective systems. Heavy precipitation and severe flood events associated with these systems are recognized as a threat to social and economic life in many regions of Europe.

Conclusion

At the end of the first year, the project is on the right track and on time. The choice of case studies with the observations to initialize and evaluate all the hierarchy of models has been achieved.

Reference simulations with explicitly resolved cloud models have been successfully performed. Preliminary analysis of simulations on each case study has led to the identification of quite general deficiencies in standard cloud schemes used in climate models. This point was considered crucial as it demonstrates that the se-up of cases is well suited to address the major issues of climate models in representing clouds.

Because the EUROCS programme constitutes the largest European contribution to GEWEX Cloud System Studies (GCSS is a non-funded programme of WCRP), EUROCS has been active and influential in a

co-ordinated manner within GCSS.

Keywords: CLOUDS, CLIMATE, PRECIPITATION, BOUNDARY LAYER

Section 3

STRATOCUMULUS OVER THE OCEANS

Gathered and Prepared by:

Peter G. Duynkerke, Margreet C. van Zanten and Arjan van Dijk

Case Leader: IMAU

Participating Partners: CNRM-GAME, ECMWF, UKMO, INM, SMHI, LMD, MPI, KNMI, and UL

WEB Page: http://www.phys.uu.nl/~wwwimau/EUROCS/eurocs_art.html

3.1 Motivation

The primary objective of EUROCS is to improve the treatment of cloud systems in general circulation models (GCMs) and regional circulation models (RCMs). The present research focuses on the representation of stratocumulus over the ocean in these three-dimensional models. This report section addresses the full diurnal cycle of stratocumulus off the coast of California based on observations of the First ISCCP (International Satellite Cloud Climatology Project) Regional Experiment (FIRE).

The extensive areas of sub-tropical marine stratocumulus clouds off the West coast of large continents lie beneath the descending branch of the Hadley circulation. The large static stability associated with the cold sea-surface temperatures (SSTs) and atmospheric subsidence results in extensive marine stratocumulus cloud decks with an areal coverage of about 40 - 90%. Due to their high albedo, these stratocumulus decks have a significant cooling effect on the underlying ocean. This cooling is only slightly offset by the downward infrared emission from the cloud.

In atmospheric GCMs, however, the amount of subtropical marine stratocumulus is usually under predicted, even when the observed SSTs are prescribed (Jakob, 1999). As a result, in such models the net surface heat flux into the ocean is overestimated in these regions. In a coupled atmosphere-ocean model this can lead to positive SST biases of about 5K. It is now well established that these model errors have a significant influence on the (sub) tropical circulations (Philander et al., 1996; Ma et al., 1996; Nigam, 1997). Furthermore, the stratocumulus cloud cover can act as feedback (or a “thermostat”) which reduces the intensity of tropical convection and damps the tropical circulation as a whole (Miller, 1997). Therefore, it is vitally important to consider the full life cycle of stratocumulus.

The goal of this study is to assess the quality of the representation of stratocumulus in the GCM (weather and climate) and RCM models. We will therefore perform an intercomparison between FIRE I observations and model results. The FIRE I experiment (Albrecht et al., 1988) provided a comprehensive observational set of data on marine stratocumulus on San Nicolas island (SNI) during July 1987. These observations will be used in LES and SCM simulations to study the diurnal variation of stratocumulus (section 3.3 and 3.4). Moreover, cloud parameters (Minnis et al., 1992; Rossow and Schiffer, 1999) derived from satellites will be used to verify the representation of stratocumulus in GCMs and RCMs over a larger region (section 3.5). We will focus on the monthly-mean diurnal variation and monthly mean values of cloud parameters and surface energy balance.

In Duynkerke and Teixeira (2001) the ECMWF Re-Analysis (ERA) results for July 1987 are compared with stratocumulus observations for the same period made during the FIRE I experiment. They concluded that the ERA cloud thickness is overestimated and cloud cover and liquid water path are strongly underestimated compared with the FIRE I observations. As a result the ERA downwelling shortwave radiation at the surface is much larger than observed. They argued that the large difference between the ERA downwelling shortwave radiation at the surface and that observed in FIRE I can lead to large biases in the sea-surface temperature in a coupled ocean-atmosphere model. Therefore, this study focuses on representation of the diurnal variation of cloud properties (boundary layer depth, liquid water path, optical depth, etc.), surface energy balance and implications for a coupled atmosphere-ocean model, turbulence dynamics (entrainment, decoupling, etc.) in the boundary layer, and shortwave optical properties of inhomogeneous 3-dimensional cloud fields.

3.2 Case Description

3.2.1 Surface based observations

The FIRE I experiment (Albrecht et al., 1988) provided a comprehensive observational set of data on marine stratocumulus during July 1987. From the observations the following parameters will be used: cloud cover, liquid water path, cloud-base and cloud-top height, and downwelling shortwave and longwave radiation at the surface (Duynkerke and Teixeira, 2001).

A description of the temporal evolution of the cloud conditions on San Nicolas Island during July 1987 can be found in Blaskovic et al. (1991), Betts (1990) and Albrecht et al. (1990). Hignett (1991) describes the detailed turbulence profiles throughout the boundary layer, taken on 14 and 15 July 1987. These turbulence data have been successfully modelled with a 1.5-order turbulence closure scheme (Duynkerke and Hignett, 1993). The overall picture is of a cloud layer that progressively thins during the late morning into the afternoon, both by raising cloud base and lowering the inversion, but that thickens again during the evening.

3.2.2 Satellite observations

Betts et al. (1992) and Minnis et al. (1992) have derived cloud parameters from the Geostationary Operational Environmental Satellite (GOES). Figure 3.1 shows the low-cloud faction derived from GOES data over the Extended Time Observations (ETO) area (Betts et al., 1992) for July 1987. The ETO data were analysed on a 2.5^o latitude-longitude grid covering the area between 10^o - 40^o N and 110^o - 145^o W. Minnis et al. (1992) analysed the data on finer grids: a 1^o and 0.33^o latitude-longitude grid, respectively. They compared the satellite derived values, on the 0.33^o latitude-longitude grid, with observations performed on San Nicolas Island. They concluded that the diurnal variation of satellite retrieved values of cloud amount, optical depth and cloud-top height shows very good agreement with the surface-based observations on San Nicolas island. The satellite data of Betts et al. (1992) and Minnis et al. (1992) were thus verified at SNI and this data will now be used over the ETO region to verify the GCM and RCM predictions (section 3.5).

3.2.3 Large Eddy Simulations and Single Column Model Simulations

The LES and SCM model simulations were started on 14 July 0800 UTC (= 00 LT) lasting at least 37 hours (37 LT). The initial and boundary conditions were based on the observations described in Blaskovics et al. (1991), Hignett (1991), Duynkerke and Teixeira (2001). Both LES and SCM use the same initial conditions and vertical resolution. From the measured potential temperature and specific humidity profiles the initial potential temperature and humidity profiles were determined. The goal is to simulate as closely as possible the typical monthly-mean diurnal variation. Therefore the initial conditions were not

Figure 3.1 Mean low-level cloud cover for July 1987 from GOES for the ETO region.

exactly tuned to the conditions on 14 July 1987 but were more representative of the monthly-mean conditions.

3.3 Large Eddy Simulation results

We have received LES results from 3 out of the 4 models. All three models have simulated the case defined and have performed sensitivity simulations with respect to sea-surface temperature, inversion properties, long wave radiative heating at cloud base and grid resolution. In Figure 3.2 we have compared the liquid water path of the three LES models with the observed values. The observed values are the hourly-mean values during 14 and 15 July 1987 (denoted as 0 to 48 hours) and the hourly monthly-mean diurnal variation. The LES models all capture the strong diurnal variation in LWP due to the forcing imposed by the shortwave heating of the cloud layer. The UKMO model has more liquid water because the model generates less entrainment of warm and dry air from above the inversion into the boundary layer. The LES results are also compared with the turbulence profiles observed by Hignett (1991) and clearly show a decoupled boundary layer during daytime and a well-mixed boundary layer during the night. The diurnal variation in boundary layer properties leads to a strong diurnal variation in the surface fluxes, an effect which will be analysed in detail in combination with the SCM and GCM results.

Figure 3.2 The observed and modelled liquid water path (LWP) as a function of time for 14 and 15 July 1987 (denoted from 0 to 48 hours). The open circles are the hourly mean observed values whereas the filled dots are the hourly monthly-mean values. The LES results are from IMAU (full line), MPI (long dashed) and UKMO (short dash).

3.4 Single Column Model results

Results from three SCM have been disseminated and have been compared with the observations and LES results. The SCM include the KNMI, LMD and MPI model. In Figure 3.3 the liquid water path (LWP) of the three SCMs is compared with the observed values. The KNMI and LMD model predict a too low liquid water path during the night (24 hours). The MPI model predicts a rather large liquid water path. Further analyses of the SCMs will be performed when all results have disseminated.

3.5 Climate Model simulations

In order to make an inventory of the quality of General Climate Models (GCMs) and Regional Climate Models (RCMs) we will collect the data of present day models and compare these with the SNI and satellite data. The GCM model can be either a climate model or a weather-forecast model. The models have to simulate July 1987 and output has to be supplied over the ETO region: between 10^o - 40^o N and 110^o -145^o W. The output should be given at the model resolution. We will focus on the monthly-mean diurnal variation of the surface energy balance and cloud properties. Moreover, we will document the vertical structure of the boundary layer over the ETO region. At the moment we have the data from the ECMWF model and we have started analysing the results. Figure 3.4 shows the monthly mean low cloud cover from the ECMWF model. This can be compared with the observed cloud cover retrieved from a satellite as shown in Figure 3.1. From a comparison of Figures 3.1 and 3.4 it can be concluded that the ECMWF model is under predicting the cloud cover by about 20%. In the future we will perform a more detailed analysis as was done for ERA in Duynkerke and Teixeira (2001).

Figure 3.3 The same as Figure 3.2 but the lines are now SCM results: KNMI (full line), LMD (long dash) and MPI (short dash).

Figure 3.4 Mean ECMWF low-level cloud cover for July 1987 for the ETO region.

4.6 Conclusions and future work

Most LES results are now available and analysed. The results will be available on the world-wide-web in the near future. With the LES models, sensitivity simulations have been performed to the sea-surface temperature, inversion properties and grid resolution. New sensitivity simulations will be performed with respect to subsidence and wind shear at the inversion. Moreover, LES simulations will be performed on a large domain (tenths of kilometres) and a case will be set up to study the transition of stratocumulus into cumulus. Results of three SCMs have been disseminated and have been compared with the LES results and observations. The SCMs should perform sensitivity analysis with respect to different boundary conditions, forcings and vertical resolutions. All other SCM results should be submitted before the workshop in Lisbon from 28 to 31 May 2001. At the time of this workshop most SCM and GCM results should be analysed.

Section 4

DIURNAL CYCLE OF SHALLOW CUMULUS OVER LAND

Gathered and Prepared by:

A. P. Siebesma and G. Lenderink

Case Leader: KNMI

Participating Partners: UL, CNRM-GAME, ECMWF, UKMO, INM, SMHI, LMD, and MPI

WEB Page: http://www.knmi.nl/samenw/eurocs/ARM/

4.1 Introduction and Motivation

Underestimation and an inadequate phase of the diurnal cycle for cloud cover over land are known to occur in climate and weather numerical models. This can have dramatic consequences on the behaviour of the boundary layer. During summer seasons, diurnally driven boundary layer clouds are commonly seen over continental regions and play a crucial role in determining the surface energy budget. On a given day failure to capture the variation of shallow cumulus clouds adequately, may lead to errors in predicted surface temperature and moisture and so contribute to failures of predicting the onset of more vigorous deep convection. On seasonal timescales, persistent biases in the surface radiation budget can lead to systematic errors in continental scale hydrological budgets which are important for predicting impact of seasonal and anthropogenic climate variability. Using the strategy of a hierarchy of models and observations such as pointed out in the EUROCS proposal we will present the intermediate results on the above raised issues.

4.2 Case description

The case was released on the internet http://www.knmi.nl/samenw/eurocs/ARM/ early September 2000. It is based on the one originally set up by Andy Brown (UK Met. Office) and is composed on observations made at the Southern Great Plains (SGP) ARM site on June 21 1997. On this day, cumulus clouds developed at the top of an initially clear convective boundary layer. The large scale forcing was weak compared to the surface forcing. Non-precipitating cumulus clouds started to form around 14:30 UTC (8.30 local) and disappeared at around 23.30UTC (17.30 local). A maximum cloud cover of around 30% was reached. Because this is the first inter-comparison study on the diurnal cycle of cumulus everything is kept as simple as possible but yet realistic. For this reason, surface fluxes, the tendencies due to radiation and large-scale advection, are prescribed all during the simulation. The required simulation starts at 11.30 UTC, just before sunrise with a stable dry boundary layer based on radiosonde observations. The simulation is required to run from 11.30 UTC June 21 throughout the whole day until 2.00 UTC June 22, well after sunset when all the clouds should have dissolved again. For a detailed case description plus the required output we refer to the web address mentioned above.

4.3 Large Eddy Simulation studies

Within the EUROCS framework three groups (MPI, UKMO and KNMI) have submitted LES results. Three important conclusions can be drawn from these results:

· LES models are well capable of reproducing realistically the diurnal cycle of the shallow cumulus topped boundary layer

· Results of previous steady state cases of cumulus over sea still apply to the present case where strong diurnal cycle is present

· The sub-cloud layer structure seems hardly affected by the clouds on top.

Concerning the first conclusion, all three models behaved very similar. They all transformed an initially stable boundary layer into a convective dry boundary layer without clouds, developed subsequently a conditionally unstable layer with clouds and finally, just after sunset slid back into a stable clear nocturnal boundary layer. Directly related is the correct timing of the appearance and disappearance of the cumulus clouds as well as the time evaluation of cloud cover, cloud base height, and cloud top height. In Figure 4.1 we show the time evaluation of the total cloud cover of the three LES models. Note that all three models do behave similarly and actually give approximately the same timing as the observations for the onset and offset of the clouds. Also the cloud cover amount reaches about the same maximum value as observed. Cloud base height (not shown here) is for all three models increasing from about 600m around the onset at 15.00 UTC to 1250m at 23.30 just before the clouds disappeared. Again these results compare quite well with observational ceilometer measurements at the central facility at the ARM site.

Figure 4.1. Total cloud cover as a function of time of the 3 LES models.

Concerning the second conclusion, previous inter-comparison studies based on BOMEX and ATEX show cloud cover and mass flux profiles that have a maximum value around cloud base and are then monotonically decreasing with height. Furthermore those studies show typical lateral fractional entrainment rates of 1~3 10^-2 m^-1 and slightly higher fractional detrainment rates. We repeated these calculations for the present case for each model result during each hour of the simulation and found similar results. Note that this is quite an interesting result, since throughout the simulation the mean vertical profiles do change substantially. So apparently these results are robust fingerprints of shallow cumulus that not only occur during controlled steady state conditions but also under fast varying conditions such as the present diurnal cycle.

Finally, concerning the last conclusion it turns out that the sub-cloud layer has quite the same characteristics as the dry convective boundary layer. For instance, the vertical velocity variances s_w² of the three models, when non-dimensionalised appropriate, do exhibit the same behaviour as in the dry convective boundary layer. Similarly the ratio of the minimum buoyancy flux and the surface buoyancy flux is almost all the time from sunset to sundown quite close to –0.2, the same value as found in typical dry convective boundary layers! This means that the top entrainment mechanism and hence the ventilation of moisture and heat into the free atmosphere is not strongly affected by the clouds. The only difference is that in the case of a dry convective boundary layer heat and moisture is deposited in a rather thin entrainment layer while in the case of a cumulus topped boundary layer it is redistributed in a much deeper cloud layer. This finding makes the issue of the parameterization of the cumulus-topped boundary layer much less complicated.

The overall conclusion is that the results of the LES models are such that they can serve as an excellent 4D-dataset that can be used as a test-bed for evaluation and improvement of the parameterizations of these processes in GCMs.

4.4 Single Column Model results

In order to diagnose the results of the parameterizations in various SCM’s the first step that has been undertaken was to get a good description of the parameterizations of all the participating GCMs. This necessary step was initiated in June 2000. At the moment, we have received parameterization descriptions of the ECMWF, LMD, ARPEGE (CNRM/GAME),, RACMO (KNMI), ECHAM (MPI) and the Meso-NH (INM and UL) models. A summary of these descriptions can be found on the web at http://www.knmi.nl/samenw/eurocs/ARM/participant.html. It appears that many of these models are rather similar in many respects, though on a more detailed level they differ distinctly. All models have a turbulent diffusion scheme, attempting to do the mixing in the sub cloud layer and some, if formulated in moist conserved variables, as well in the cloud layer. Except LMD and ECMWF, all models use a moist 1.5 order turbulence closure. Some models (LMD and ARPEGE) however do not seem to mix in moist conserved variables and are in this sense inconsistent. Most models use a (bulk) mass flux scheme to do mixing in the cloud layer. The models differ most with respect to computation of the cloud fraction and cloud liquid water (the cloud scheme). The models ranged from prognostic schemes for both cloud liquid water and cloud fraction to fully diagnostic schemes for both variables.

At the moment of this writing results were received by following models: ECMWF, ARPEGE, RACMO and Meso-NH (two versions from INM and UL). However, not all the requested output was sent in by the participants so a very detailed analysis is at the moment not possible. This task was in fact scheduled for the second year in the initial proposal. Nevertheless, from the first results a rather consistent picture emerges. The results basically confirm results earlier obtained from inter-comparisons for cumulus clouds. Most models produce a total cloud cover of 50-90 %, in contrast to the 25 % in the LES results (see Figure 4.2). The integrated liquid water path, which is shown in Figure 4.3, is consequently also high (2-5 times the value in the LES models). In addition, neither of the models produces a realistic diurnal cycle; the clouds remain after sunset. Except the ECMWF model, they all produce profiles of temperature and moisture that are too well mixed and the cloud top height is underestimated. This is a consequence of the apparent strong activity of the turbulence scheme, producing profiles close to the well-mixed stratocumulus regime. This indicates that either the moist mixing in the turbulence scheme is overestimated or the activity of the mass flux scheme underestimated. Some participants reported that the mass flux scheme failed to turn on. In this respect, the ECMWF model was the only exception. In this model, the mass flux scheme seems to be overactive. The ECMWF produced a very realistic cloud top height,

Figure 4.2 Total cloud cover as a function of time of the single column models (note that the time is local time).

Figure 4.3 Integrated liquid water path as a function of time of the single column models (note that the time is local time).

but overestimated the transport by the mass flux scheme. As a consequence, a too strong drying and warming near cloud base occurs and a too strong moistening and cooling near the inversion. This produces a maximum in the cloud fraction at the inversion; this feature was also observed in the other models - though for a different reason - and contrast highly with the LES which have maximum cloud fraction near cloud base.

So there are at the moment two concluding remarks:

· Concerning the mixing, it appears that many of the deficiencies are caused by an improper balance between mixing of the (moist) turbulent scheme and the mixing of the mass flux scheme. Except the ECMWF model, in all models the turbulence scheme is too active in the cloud layer. The ECMWF model does the opposite and has a too active mass flux scheme.

· Concerning the cloud fraction, all models are strongly overestimating cloud cover and liquid water paths. Also they collectively fail to dissolve the clouds at the end of the day.

4.5 Plans for Climate Models

In Section 4.4 of this report, a case for climate model over persistent stratocumulus areas off the coast of California are planned. In addition we plan to do additional climate model runs for the same period over an area more upstream towards the Hawaii Islands where the stratocumulus regime is breaking up into more scattered cumulus fields. As a first step in this direction the ECMWF model will produce output, which will be analysed in spring 2001. Subsequently a prescription for the other participating models will be set up.

4.6 Conclusions and future work

The LES model results compared well enough with the observations in order to be used as 4D Data sets for evaluating parameterization issues. Therefore, not much more additional effort is required on this part. Many of the results found for steady state marine cumulus were reconfirmed by the present case.

For the SCM results, despite the fact that not all requested output was sent in we already identified some quite general deficiencies concerning both the mixing and the clouds. This needs and will be further explored at the next workshop in Lisbon. By this time all required output of the participating SCMs will be analyzed. Special attention will be paid to the vertical resolution since it seems that vertical resolution might play a major role in the results. To this purpose we ask for all SCMs output at three different vertical resolutions: 1) the operational, 2) ECMWF resolution (at present: highest operational resolution) and 3) a very high resolution of effectively 100m.

Further future plans are twofold: Firstly to address the stratocumulus/cumulus issue, we are conceiving a wide range of LES runs in which the initial conditions are varied from typical stratocumulus to cumulus. This can be used for the SCMs tot test whether they are capable of selecting the right cloud type, given the initial conditions. Secondly as mentioned in section 4.5 we are setting up a case for 3D climate models in the trade wind region over the Pacific Ocean.

Section 5

DIURNAL CYCLE OF DEEP CONVECTION OVER LAND

Gathered and Prepared by:

F. Guichard and J. Petch

Case Leaders: CNRM-GAME and UKMO

Participating Partners: ECMWF, INM, SMHI, and LMD

WEB Page: http://www.cnrm.meteo.fr/gcss/EUROCS/deepdiurnland.html

5.1 Introduction and Motivation

The diurnal cycle of deep convection plays an important role in the energy and water budgets, both locally and globally. Convectively generated clouds (as any cloud) strongly interact with both solar and thermal infra-red radiation. At night, they contribute to the greenhouse effect by efficiently trapping thermal infra-red radiation so that night time surface temperature is warmer under cloudy conditions. During the day the reflection of insolation generally dominates over the greenhouse effect which has an opposite effect on surface temperature. Therefore, it is important to correctly predict the time of the day during which these clouds are present in order to get an accurate radiative budget, both at the top of the atmosphere and at the surface. Clouds also strongly modulate the magnitude of surface heat fluxes (sensible plus latent) as well as their partition (sensible versus latent) via their radiative impacts and precipitation (a larger part of rainfall is re-evaporated over land in the daytime than during night-time). The diurnal cycle of deep convection is generally stronger over land than over ocean, with a phase difference between oceanic and land areas (Figure 5.1). Convective precipitation exhibits an early morning maximum over open oceans (Gray and Jacobson 1977) but there is no consensus toward a theory explaining this phenomenon (see Parsons et al. 2001 for a review). The situation is somewhat simpler over land where the diurnal cycle of deep convection is strongly linked to daytime boundary layer heating in response to solar radiation, with a maximum of precipitation in the afternoon to early evening.

Recent studies have shown that it is difficult for GCMs to capture the diurnal cycle of deep convection, both in terms of magnitude and phase, and over land as well as over ocean convection (Dai et al. 1999, Lin et al. 2000, Royer et al. 2000, Yang and Slingo 2001). Royer et al. (2000), using the CLAUS (Cloud Archive User Service) data set show that the diurnal cycle in their GCM is too weak with precipitation occurring several hours too early as compared to observations (Figsures 5.1 and 5.2). Lin et al. 2000 also show that they can get a reasonable GCM simulation of the diurnal cycle or of the monthly mean, but not both at the same time. As an archetype, the magnitude of the diurnal cycle of convection over South America is the largest in the world (Yagai et al. 2001). In the ECMWF operational model, it corresponds to a large-scale feature over the Amazon region. Figure 5.3 shows the deviation of 6-hour averages of precipitation from the daily total taken from 24 to 48 hour forecasts with the model for 1-7 February1999. The four panels show the rainfall anomaly from the 7-day all-day average for rainfall occurring between 12-18 UTC (8-14 LST, top left), 18-00

Figure 5.1 Mean convective rainfall for JJA 1992 computed by applying the DCA index of Hendon and Woodberry to the CLAUS satellite archive (top : in mm/day), relative amplitude ratio (middle) and phase (bottom : hour of the maximum in local time) of the first harmonic of its diurnal cycle (from Royer et al. 2000).

5.2 Case description

As a first goal of this study, we need to evaluate the ability of the models to reproduce deep convection over land, over a domain typical of a GCM grid size, for a time period of a few days. This implies the simulation of a real case, for which relevant observations are readily available, both to set up the case study and to validate the simulations.

This is the reason why we have chosen to focus on a case of mid-latitude deep convection over the ARM (Atmospheric Radiation Measurement) SGP (Southern Great Plains) site in Oklahoma, for which a large amount of data are available. We chose our case study from the IOPs (Intensive Observation Periods) dedicated to ARM-SCMs studies. Of special interest is the fact that large scale advection derived from observations have been computed from these data. However, the diurnal cycle of deep convection over the Southern Great Plains is particularly complex. It does not frequently exhibit a well-defined afternoon maximum. Instead, convective rainfall is strongly modulated by mesoscale convective systems generated outside of the domain, in the North-East and further advected in the SGP at night. A careful examination of the different IOPs led us to retain the 4-day period June 27-31 1997, as this period appears less influenced by advection of large convective systems generated outside of the area. This period also presents the advantage of strengthening the links with the GCSS-WG4 (Gewex Cloud system Study - Working Group 4) and ARM communities, as this period is also part of an ensemble of periods chosen for an inter-comparison of models (Krueger et al. 1999).

Moreover, from this case study, we can further design a more idealised set-up, better suited to address issues relevant to the diurnal cycle of deep convection. An interesting approach consists of choosing one particular day, or in compositing a few appropriate days, in order to build a generic 24-hour time period. This can be further used to perform simulations over 3 or 4 days with the composite surface and large scale forcing of this 24-hour period repeated for each day of the simulation. So far, we have retained the 27th of June 1997 as a good candidate for this generic 24-hour time period. The reason why we want to perform a 3 to 4 days simulation with this generic 24-h day is to assess the reproducibility of the phenomenon by the models and its sensitivity to variations of the thermodynamic fields.

Necessary information to perform the simulation of this case has been made available on the EUROCS web site. In brief, the models (both CRMs and SCMs) are initialised from an area average sounding (to which a white noise in temperature is added in CRMs in order to initiate convection). Time-varying large-scale temperature and moisture advection are prescribed as deduced from a variational analysis (Zhang and Lin 1997). CRMs lateral boundary conditions are cyclic, so that SCMs and CRMs simulations are performed with exactly the same large scale forcing. Time varying surface sensible and latent heat fluxes are also fixed to values derived from an average over several point measurements in the domain. The mean winds are nudged towards observed mean values with a time scale of two hours, as it is very difficult to diagnose large-scale advection of momentum from observations.

Data files and programs reading these data have also been made available on the EUROCS web page, as well as an ensemble of figures documenting the case. Diagnostic outputs have been defined following the list proposed by GCSS-WG4 (including time series of precipitation, precipitable water, cloud water, temperature and moisture fields, cloud mass fluxes...). Also included is an ensemble of additional model output more dedicated to the analysis of the diurnal cycle, including in particular convective fluxes of heat and moisture.

5.3 Cloud Resolving Model results

Results from three different CRMs have been analyzed for the case study described above: UKLEM (UKMO), Meso-NH, and Come-NH (CNRM/GAME). Two of these CRMs also took part in the GCSS-WG4 Case 3 CRMs inter-comparison (Xu et al. 2001) - the analysis of the simulated diurnal cycle was not performed in detail as this study mostly focused on time-mean properties. This exercise showed that CRMs compare reasonably well with observations and that the results are very consistent among CRMs. The performances of the two CRMs (cited above) are comparable to the ones of the five other CRMs involved in the inter-comparison. However, this exercise also showed that it was difficult for CRMs to predict the occurrence of precipitating events with a very precise accuracy. Indeed, some of the models even missed the first event.

The (too) simple design of the simulation might be partly responsible for this problem (initialization from a homogeneous sounding, a lack of spatial variability in the forcing). However, as noted in section 5.1, this type of simulation also involves strong interactions with surface and boundary layer processes, i.e. processes that occur on smaller scales than deep convection. For the above inter-comparison, the horizontal resolution of the models was 2 km, i.e. too low to explicitly simulate boundary layer eddies, so boundary layer processes are for these resolutions mostly handled by the turbulent scheme.

We began to investigate this area and the first results are quite encouraging. Sensitivity tests to the resolution were performed with UKLEM, from 2km to 250m (simulations noted S_{2000 m}, S_{1000 m} , S_500
m , S_{250 m}). The timing of rainfall (and therefore cloud cover) was found sensitive to the horizontal resolution. This problem was not simply related to the initialization technique, as this feature was robust (see Figure 5.4; it appears every day for a four-day simulation using the observed surface flux and no large scale forcing). Figure 5.4 also shows that the difference of the rainfall timing among the simulations with various resolutions decreases as the resolution increases. In other words, there is evidence of convergence at higher resolution. The sensitivity to the resolution was also shown to be a consequence of different turbulent flux in the boundary layer.

The timing of precipitation was also found sensitive to the choice of mixing length L used in the turbulent scheme. Simulations were performed in using the standard formulation in CRM relating L to the grid size (simulation S_l(_D)), and using the formulation of Bougeault and Lacarrere (1989) (simulation S_l(BL)), which is expected to provide a better representation of turbulent processes in the boundary layer for mesh size larger than the kilometre (Figure 5.5). The first event is missed in S_l(_D), but is reproduced in S_l(BL₎.

These results from 2 CRMS suggest that a proper CRM simulation of deep convection over land requires high horizontal resolution and/or an advanced turbulent scheme.

Figure 5.4 Time series of precipitation for 2D CRM simulations with various horizontal resolutions. Simulations were performed with large scale advection of heat and moisture set to 0 - units on the y-axis are mm/h. (UKLEM CRM used at UKMO)

Figure 5.5 Time series of precipitation for 2D CRM simulations with two different turbulent schemes. (Meso-NH CRM used at GAME/CNRM)

5.4 Single Column Model results

Results from two different SCMs have also been provided (UKMO and ECMWF SCMs ). A complete analysis is under way. First results show that precipitation occurs too early in both models. These results suggest that this weakness, shared by many GCMs, could be properly analyzed within this joint CRM/SCM framework.

Finally, initial investigations also suggest that the diurnal cycle of the boundary layer is not fully satisfying for both CRMs and SCMs (too weak diurnal variations of temperature and too vertically homogeneous moisture profiles). A likely cause to this problem is the treatment of surface fluxes. One participant will investigate the impact of replacing fixed surface fluxes by a time varying prescribed surface temperature coupled to a simple surface scheme using observed Bowen ratio.

5.5 Conclusion and future work

This work focuses on the diurnal cycle of deep convection over land. It aims at investigating the ability of a hierarchy of models (GCMs, SCMs and CRMs) to reproduce this atmospheric feature. In a first step, an overview of GCM results have shown that deep convection frequently occurs too early in the daytime in GCMs.

In a second step, a case-study for which relevant observations are available has been chosen. It corresponds to mid-latitude deep convection over the ARM SGP site in Oklahoma. The simulation set-up and a list of required model diagnostics has been made available to participants on the web. The results from 2D simulations have been analysed. They show that the timing of rainfall is sensitive to both the horizontal resolution and the turbulent scheme in CRMs, better resolution and turbulence scheme leading to improved results. They also show that the rainfall simulated by SCMs occurs too early, suggesting that this weakness, generally shared by GCMs, could be properly analysed within this joint CRM/SCM framework. Finally, first investigations also suggest that the simulated diurnal cycle of the boundary layer needs to be improved.

From this case study, we are currently designing a more idealised set-up, better suited to address issues relevant to the diurnal cycle of deep convection, e.g. by compositing a few appropriate days so as to get a generic 24h time period. We will also extract new diagnostics from the CRM and SCM simulations in order to better understand the role of boundary layer processes. In parallel, additional outputs from standard GCMs will be collected in order to get an accurate view of the problems encountered by these models with the diurnal cycle of deep convection.

Section 6

IDEALIZED HUMIDITY CASE

Gathered and Prepared by:

Steve Derbyshire

Case Leader: UKMO

Participating Partners: CNRM-GAME, ECMWF, INM, SMHI, and LMD

WEB Page: http://www.cnrm.meteo.fr/gcss/EUROCS/deephumidity.html

6.1 Motivation

The sensitivity of moist convection to mid-level humidity is an important issue in climate variability (e.g. Parsons et al 2000), and provides an idealized test of convection representation in weather and climate models. Johnson (1997), reviewing tropical observations from the TOGA-COARE experiment, shows clear associations between observed convection strength and relative humidity profiles. The more strongly convecting cases tend to have significantly moister profiles in the mid-troposphere (whereas in the boundary layer the relative humidity varies little between strongly and weakly convecting cases).

Links between convection and humidity are thought to play a role in the Madden-Julian Oscillation (an important component of the climate system which is often not well modelled), in the differences between convection over oceans and over continents (including Europe), and in the differences between cyclonic and anticyclonic weather in mid-latitudes.

The sensitivity of convection to humidity is also relevant to certain technical model problems. In weather and climate models, numerical "gridpoint storms" can occur when resolved-scale saturation is approached in a convectively unstable column. Modellers usually try to prevent this undesirable phenomenon by tuning the rate of parametrized convection (e.g. Betts and Miller 1993), but the need for such tuning may partly reflect deficiencies in the response of current convection schemes to humidity.

Parametrization of convection requires not just observed associations but a clear and quantitative understanding of mechanisms. The tropical observations might be explained by various mechanisms:

(a) loss of buoyancy within clouds by entrainment of subsaturated environmental air.

(b) radiative impacts on temperature.

In this project therefore we seek to use idealized CRM-SCM comparisons to help separate and quantify the mechanisms, especially the entrainment mechanism (a).

Most current convection schemes attempt to represent entrainment in some form. But the values used differ widely between schemes, and many schemes do not use a single

entrainment profile but rather assume a range of mixing within the convective cloud-field. Such variation is represented in different ways, following conceptual models due to Arakawa and Schubert, Kain and Fritsch or Emanuel. Hence it may not be appropriate to test these schemes by evaluation of a "standard" entrainment-profile derived from CRMs. Instead, here we seek to use the humidity-sensitivity as a benchmark for all the schemes, which is linked to entrainment but not tied to any specific formulation.

Here we examine the convective humidity-sensitivity in a quasi-steady scenario with profile nudging. The quasi-steady scenario avoids complications of time-development, which are studied in other components of EUROCS. Profile nudging effectively incorporates a simplified model of the dynamical feedback on the temperature profile when the convecting region is smaller than the characteristic dynamical response scale. In this scenario the dynamics act to balance much of the convective heating (Q1); see Sobel & Bretherton (2000) and Randall & Cripe (1999) for related discussion. Inclusion of such nudging is intended to show the humidity impact more clearly and make the CRM-SCM setup more relevant to the performance of convection schemes in GCMs.

6.2 Case description

Two cases have been run with the UKMO CRM, a pilot simulation (A) and the main intercomparison case (B). Each case involves a range of humidities.

6.2.1 Run A (pilot Cloud Resolving Model run)

The CRM pilot run was based on simple warm-rain physics (Kessler), with rapid relaxation of potential temperature and specific humidity to target profiles (at heights above 1km). The target humidity profile was varied at heights above 2km. A sea-surface temperature was prescribed, with a potential- temperature difference of 1K across the boundary layer and no wind-forcing. A sketch of the specification is shown in Figure 6.1and basic results in Figure 6.2. The pilot run confirms that changes in the mid-tropospheric humidity can switch the convection regime between shallow non-precipitating convection and vigorous deep convection, with changes to the shape of the convection profiles as well as the magnitude and depth.

Figure 6.1 Sketch of target specification for pilot run A, for mean potential temperature q and relative humidity RHw. Dashed portions represent regions where no nudging is applied. The specification for run B is broadly similar but differs in detail.

Figure 6.2 Mass flux profiles with various mid-tropospheric humidities in the pilot run A, using warm-rain physics only and strong nudging. With 50% RH, shallow convection is found with a small mass flux, but with only 80% RH, a much stronger precipitating convection with a different shaped profile (peaking at 4-5 km).

6.2.2 Run B (main intercomparison case)

For the formal intercomparison of different models it was decided to use 3-phase cloud physics in the CRM (since the SCMs are designed to allow for ice processes). The relaxation timescale was specified as 1 hour, which is longer than a typical timestep in the SCMs (e.g. 30 minutes in the UKMO SCM). A moderate mean wind profile was specified, to help promote surface evaporation even in the absence of convective gusts (these are represented in the surface scheme for some but not all models). The case setup is given on the Web page. Following are only given the key points. Radiation schemes in the CRM and SCMs are switched off. The target relative humidity RHt is specified as 80% for z between 1km and 2km. Above 2km, RHt is specified as (a) 25% (b) 50% (c) 70% (d) 90%. Here RH=max(RHw,Rhi). The simulations to be run for 12 hours (likely then to be statistically steady)

6.3 Cloud Resolving Model results

A summary of the CRM results in the intercomparison case B is shown in Figure 6.3. Again there is a clear trend for stronger convection in the moister cases, which pushes the temperature profile closer to the moist adiabat (in contrast to the more nearly linear profile of the target profile). These differences are seen in the fluxes of sensible and latent heat, with changes to both the magnitudes and profile shapes.

The more strongly convecting cases gave cooler and moister boundary layers, consistent with the expected influence of precipitating downdraughts. Despite the differences in magnitude and profile shape, the depth of convection was almost the same in all the cases. Positive sensible heat-flux (and probably buoyancy-flux) are found up to 8-9km, with a significant `overshooting layer' of penetration of non-buoyant updraughts.

In case B the equilibrium rates of convection are attained more quickly than in the pilot case A, possibly owing to the influence of the mean wind on surface evaporation.

6.4 Single Column Model results

The pilot run A can been compared with idealized mass-flux convection
schemes. It is found that schemes which allow for variation in mixing
can reproduce the profile shapes better than a simple fixed-entrainment scheme.

For the intercomparison case B the initial results for surface precipitation as a function of mid-level humidity are shown in Figure 6.4, including the CRM results for comparison. The P3SCM (Met Office) agrees overall quite well in many respect, including the increased precipitation at higher RHt. At low RHt though the P3SCM maintains significant precipitation, whereas P3CRM (Met Office) is almost non-precipitating. The P7SCM, using a version of the Emanuel scheme, appears to show lower precipitation. These initial results require further assessment before drawing any definitive conclusions.

6.5 Plans for Climate Models

Plans for implementation in climate models are at an early stage because the CRM-SCM comparisons require further assessment. However it is expected that conclusions from the CRM-SCM work will be tested in climate models in two main ways:

i) in the "closure" which controls the overall rate of convection.

ii) in the convective profile shapes, especially in schemes which use a PDF for mixing.

Figure 6.3 Results from CRM (run B) for 25%, 50%, 70%, and 90% target humidities. The moister cases show stronger convection, pulling the termperature profile closer to the curved moist adiabat, cooling and moistening the boundary layer through precipitation and changing the magnitude and shape of the flux profiles for potential temperature () and moisture ().