Experiment Proposal: Extending sensitivity tests on marine low cloud and radiation bias in BARRA-C2 (4.4km, RAL3.2)

Atmosphere Experiments
, ,
1

It is an extension to the previous experiments proposed here

Summary

We found that reducing a key parameter Cloud Droplet Number Concentration (CDNC) can reduce model bias in low clouds and radiation for a case study period 1-2 Jun 2020. Here I would like to extend it for the whole month (another 28 days) for the control experiment and two tests with lower CDNC. This will enable us to investigate potential adverse impacts on other variables and slightly longer timescales.

Experiment Name
Extension of sensitivity tests on marine low cloud and radiation bias in BARRA-C2 (4.4km, RAL3.2)

People
Qinggang Gao

Total KSUs required
380

Total storage required
4.2TB

Details:

Model: ACCESS-rAM3
Configuration: https://code.metoffice.gov.uk/svn/roses-u/d/q/7/0/0
Initial conditions: ERA5
Run plan: 3 simulations, each extended for 28 days
Total KSUs required: 3 x 28d x 4.5 KSU/d = 380 KSU
Time required: 3 * 28d * 6 hours/d = 504 hours
Total storage required: 3 x 28d x 50Gb/d = 4.2TB

2

I support this use. Could you outline why this time period has been selected?

3

Thank you Matt.

We found that the bias in low clouds and radiation is consistent throughout all years, months, and hours. So it does not matter which time period we choose.

Though, I chose the year 2020 because the WCRP hackthon run a global UM simulation with RAL3.2 at ~4.4 km grid spacings for the year 2020. This point does not matter now because there was a big problem with dust/clear-sky radiation effects in that simulation.

I chose the month Jun because trade cumuli are more prevalent in the austral winter season over the Great Barrier Reef region. It will be our main focus for next steps when running sub-km-scale simulations for aerosol-cloud interactions.

4

@qinggangg, I believe the Met Office k-scale team are re-running that run - so the fixed version might be available for comparison at some point. I will be having a meeting with them over the next couple of weeks, I can find out if you’d like.

5

Thank you @bethanwhite That would be fantastic. It would be nice to know the global implication of this low cloud and radiation bias.

6

Hi @qinggangg you should check in with me re: I/O settings and job queue/executable for your suite.

You might be able to reduce 4.5 kSU/day by a significant amount.

7

Hi @Paul.Gregory That sounds great. Would you be at office tomorrow anytime? Or would you be available for a Zoom meeting today? What are the main ways you found useful to reduce computational requirements?

8

I’ll be in the office tomorrow.

9

Could we meet at 9:30 am in the Boardroom 406? I sent an invitation. Other times except 1:30-2pm work for me as well.

1 Like