General Info

GCSS WG2 (Cirrus) Model Inter-comparison Study

Feb 26, 2007

 

Introduction

 

We are pleased to introduce the next GCSS WG2 (cirrus) inter-comparison.   The previous inter-comparison study by WG2 illustrated that different cirrus models produce vastly different results for IWP in time even for an idealised case study.   The order of magnitude variation was not anticipated based on the published literature leading up to the inter-comparison.    A degree of grouping was noted separately for bin and bulk models and a strong dependence was noted on the fall-velocity of ice.   Significantly different behaviour between these groups of models was noted in terms of the fall-velocities and the vertical profiles of ice.    The last inter-comparison was an idealised case and so did not provide for an opportunity to compare results to observations.

 

To build on the past results, there is much interest in a cirrus case that is well grounded in observations. This current inter-comparison has made this a focus.   The aim in choosing the new case has been to use a case with a wide range of observations available, including extensive remote sensing and in-situ observations.     We have chosen the cirrus observed on March 9th 2000 during the IOP at the ARM Southern Great Plains (SGP) site for this new inter-comparison.    

 

Case Overview

 

There was an IOP for the month of March in 2000 at the SGP ARM site.   Two key objectives in the IOP were:

•         Document the 3d distribution/evolution of the cloud field as it advects through a volume

•         Characterize the evolving cloud properties in a CRM grid square

There was cirrus observed on the 5th, 6th, 9th, and 13th.       The 9th stands out in terms of having good aircraft observations and remote sensing results.     The cirrus observed by the MMCR radar for the 9th is shown below:

Fig 1 MMCR Radar for March 9th 2000 (Mace)

 

On the 9th March, there was a SW jet to the west of the ARM SGP Central Facility (CF).   The cirrus developed in disturbance just the west of the CF and advected directly over the ARM site.  The cloud arrived at the CF at 14:00 (UTC) and is observed for several hours as it advected over the site.  

 

In situ observations of cloud were taken by the UND Citation using PMS, CPI, CVI, FrostPoint Hyg., and Turb probes.   Ground based remote sensing was performed using ARM operational and supplemental instruments.    Mace has performed analysis of this case and has supplied IWC, number concentration, effective radius, mean length, and mass squared and mean mass weighted fall velocity profiles for the case.  Analysis of the size distributions and CVI observations have been provided from Heymsfield.

 

The ARM radiosondes were launched every three hours from five sites.  A sample profile is shown in Fig 2.  

 

Fig 2 Sample profile

 

The radiosonde data was used in conjunction with profilers in an objective analysis (OA) by Zhang et al.    The OA shows a weak lifting at the CF at the time of the cirrus formation.   Regional models, however, do not show a lifting and exhibit a quiescent atmosphere or very slight downdraft.    Since the large scale flow upstream from where the cirrus forms encounters the Rocky Mountains, we have investigated the origin of the cirrus to most likely result from gravity waves resulting from the flow over the mountains.   The 3DVOM model has been run (Ross and Yang) to assess the gravity wave influence and it agrees with the OA in terms of the strength and location of the weak regional updraft at the CF site and upwind where the cirrus forms.   We use the gravity wave results as a forcing to generate the cloud for the inter-comparison.   A plot of the gravity wave spatial distribution is shown in Fig 3

 

Fig 3.  Gravity wave forcing from 3DVOM (Ross and Yang) at cloud height.  The scale is in m/s.

 

 

Model Simulations

 

Satellite observations show that the cirrus forms at about 14:00 (UTC) on March 9th, upstream from the CF.   It then advects for 3.5 hours, intensifying especially as it crosses the CF at about 17:30-1800.   

 

In order to keep the simulations as simple as possible and not computationally demanding, we have chosen to do 2D model simulations using a 10km x 20km (horizontal x vertical) domain for the period from formation of the cirrus cloud to when it crosses the CF.    The simulation is semi-Eularian in that the whole domain at time zero is exposed to the conditions of the cirrus at the formation location.   The domain is then forced by the spatial variation of the gravity wave as the domain advects to the CF.   In our model, we specify the start time to be when the cloud forms upwind from the CF.   We apply a gravity wave forcing according to the results from 3DVOM model results.   A plot of the gravity wave forcing taken from Fig 3 between the location where the cirrus forms, through the CF, and downstream is shown in Fig 4.   This is the forcing that is applied in our 2D simulation.    On the vertical is ascent rate in m/s and the horizontal is the spatial coordinate.   The left end of the plot is the location of the cirrus formation and the right side is the location of CF.

 

Fig 4.  Gravity wave forcing from 3DVOM (Ross and Yang).

 

The initial profiles are taken from the CF site and projected upstream to the location where the cirrus forms, accounting for the influence of the gravity wave on the profiles.   We know that the cloud forms at this point and so a slight adjustment to the temperature (within the variation noted at different radiosonde locations) was applied so that the water vapour (wv) is slightly ice supersaturated.     We specify a small ice super-saturation of 10% at the start time and then allow the simulation to evolve under the influence of the gravity wave forcing shown in Fig 4.  We chose 10% so that the cloud forms at the correct location and yet it does not significantly influence the cloud amount at later times.   We tested this by omitting the GW influence in a test run and it shows a weak cloud formation and quick dissipation.   Any weaker and the cloud is delayed in forming.    Our main simulations are run with the GW forcing and show the formation begins at the first time-step and development of the cloud coincides with the GW forcing, as expected. We begin the run with initial random perturbations specified in the case setup.   

 

We use a Fu-Liou radiation model for the solar (6 bands) and IR (12 bands) simulation.  We expect participants to switch off their radiation schemes and use our heating rates as an applied forcing.    We provide the heating rates with updates every 5 minutes.   We have performed runs for updates every 2, 4, and 5 minutes and the resulting profiles of IWC agree with one another and with the on-line radiation runs.  

 

Part 1 of this inter-comparison will be to study the formation and the resulting cloud first observed at the site.    Part 2 of this inter-comparison we will update the GW forcing to correspond to a later time in the observed cloud at the CF and shown in Fig 1.

 

LEM simulations at Leeds

 

Our simulations at Leeds have been performed using the UK Met Office LEM model run with the 10km x 20km domain with variable vertical resolution and 100m horizontal resolution.  This resolution was confirmed to be adequate by Quante through analysis of the turbulence measurements from the citation aircraft measurements.  The model uses a double moment bulk microphysics scheme with an online coupled Fu-Liou radiation model.    We have compared the simulation results to the CF MMCR radar returns and note that there is reasonable agreement with the profiles of IWC (magnitude and vertical profile).   Our simulations take about 3-4 hours to run the case on a Linux PC (Pentium IV, 2.8).    We expect participants to be able to setup and run the model in a few days of work.   We are now beginning more detailed comparisons with the observations, such as those shown from remote sensing in Fig 5 below.

Fig 5.  Remote sensing results for March 9, 2000 (Mace).

 

Participation

 

This case study is appropriate for LEM, CRM, and SCM simulations.    We will give an initial four month period for participants to run the case (See the timeline on the main page).   We will then have an informal meeting at the IUGG in Perugia, Italy in 2-13 July, 2007 where we will discuss the first analysis of the results.   

 

Please refer to the download page for the case setup and requested output.  

 

Acknowledgements

 

The Chair would like to thank Mr Huiyi Yang (Leeds) for establishing the case with the help of data and analysis from Dr Jay Mace (Utah), Dr Andy Ross (Leeds), Dr Markus Quante (GKSS), and Dr Andy Heymsfield (NCAR).    Thanks for support from the previous pan-GCSS Chair Dr Christian Jacob (BOM) and the current chair, Dr Piers Siebesma (KNMI).