BISMG:SarahS/weekly reports

Aims
Testing the ability of glimmer-cism to predict changes in ice thickness and velocity for Antarctica over the period 1980 to 2004.

Simulating changes in the Antarctic ice sheet for the next 200 years using data from regional climate models for the A1B and E1 SRES scenarios.

A sensitivity study of key model parameters in glimmer-cism will be carried out and I will also help support an ice model inter-comparison project with Tamsin Edwards.

This work forms part of the ice2sea project.

Week ending 13th November 2009
Ran glimmer-cism using the glint driver. Forcing with one year NCEP data repeated for 10 years. Initial conditions 1) Ice surface elevation 5km provided by Anne Le Brocq (5km resolution) and 2)  Bed Topography from BEDMAP1_plus  (5km resolution).

I had to run on the 20km resolution because I encountered a run time error about insufficient virtual memory when I ran on a 5km resolution. Perhaps 5km is too fine a resolution to run the model or I need to look for a bug. Prepared ERA-40 monthly 2m temp and total precipitation data for input into glimmer-cism for the years 1958-2002. Changed the driver to read in ERA-40 data.

Week ending 20th November 2009
I made changes to the floation equation in glimmer-cism to account for firn snow. These are the changes

1.	In glide_io.F90 : read in firn snow from netCDF file

2.	In glide_vars.def : add extra variable data%geometry%firn to the list

3.	Change density of ice to value used in Anne's data(original value 910kgm-3, new rhoi=918kgm-3)

4.	In glide_setup.F90 insert new subroutine called glide_calclsrf_firn(thck,topg,eus,firn,lsrf)

!*FD Calculates the elevation of the lower surface of the ice, !*FD by considering whether it is floating or not, modify to   !*FD include firn

use glimmer_global, only : dp   use glimmer_physcon, only : rhoi, rhoo implicit none

real(dp), intent(in), dimension :: thck !*FD Ice thickness real(dp), intent(in), dimension :: topg !*FD Bedrock topography elevation real, intent(in)                     :: eus  !*FD global sea level real(dp), intent(in), dimension :: firn !*FD Firn depth real(dp), intent(out), dimension :: lsrf !*FD Lower ice surface elevation

real(dp), parameter :: con = - rhoi / rhoo

where (topg-eus < con * thck + firn) lsrf = con * thck + firn elsewhere lsrf = topg end where

end subroutine glide_calclsrf_firn

5.	In glide_mask.F90 modify subroutine glide_set_mask

!Identify points where the ice is floating or where there is open ocean including firn

where (topg - eus < con * thck + firn) mask = ior(mask, GLIDE_MASK_OCEAN) elsewhere mask = ior(mask, GLIDE_MASK_LAND) endwhere

Test pixels to check if calculated usrf and lsrf equals data provided by Anne.

Looks ok!

Week ending 27th Nov 2009
Installed an updated version of glimmer-cism.

Modified the code to read in fixed temperature and accumulation data from Anne's dataset.

I’ve been comparing the re-factored temperature solver to the original code to get a handle on what changes have been implemented.

Week ending 4th Dec 2009
Started a new wiki section for glimmer-cism notes.

Made modifications to glimmer-cism Makefile to remove python.

Some of the week has been spent writing a paper on dust modeling and tuning.

Week ending 13th Dec 2009
Made a list of the subroutines contained in the glide files and what these subroutines do.

Anne sent me the latest version of her dataset with a paper describing the data. Read the paper F. Pattyn (2003) A new three-dimensional higher order thermomechanical ice sheet model..

I still have nt been able to resolve the reason why I get insufficient virtual memory when I run the model on a 5km resolution. Running on bluecrystal gives the same error. The debugger tells me the error occurs at subroutine glide_temp_timeevolvetemp. This subroutine is not present in the code so it must be generated at run time. The _mp suggests some kind of parallelisation. Ill see if I can increase the ulimit next (as suggested when i Google the error message) and see if using 4 processors on blue crystal helps.

Continued to write up a paper.

Week ending 18th Dec 2009
The model now runs on a 5km resolution. The stack limit on the machine needed to be changed to unlimited (ulimit –s ulimited)

Modified the code to read in geothermal heat flux instead of calculating it. All of Anne's data is read in glide_io.F90.

Read some papers:

Rignot and Thomas 2002- Mass balance of polar ice sheets

Alley et al. 2005 - Ice sheet and sea level changes

Davis et al. 2005 - Snowfall-driven growth in East Antarctica Ice sheet mitigates recent sea level rise

Finished first draft of dust model description paper.

Week ending 8th Jan 2010
Compared the run time for glimmer-cism at 5km and 20km resolution.

2 years at 5km = 15m41.000s

2 years at 20km = 0m25.792s

I am going to run on 20km for testing.

Read Characteristics of the Antarctica SMB, 1958-2002 using a RCM (RACMO2), Van De Berg et al., (2005)

Currently running a 20k simulation and outputting total ice volume get an idea of how long it takes to spin up the model.