ice_cov

[songmirong@lasg_iap_ac_cn]

hello everyone
I'm learning CAM3 recently, and I draw some plots using the netCDF files of intro-model.When I use the boundary data of sea-ice to plot,I found there are some negative numbers of ice_cov in sst_HadOIBl_bc_128x256_clim_c031031.nc. What I'm confused is why there should be minus about ice_cov? A piece of the data are like this :
-0.01155236, -0.01565171, -0.03008045, -0.03927081, -0.04363815,
0.002425901, 0.01498926, -0.0547794, -0.06508866, -0.09573206,
-0.2511888, -0.2600976, -0.180498, -0.1722315, -0.1613573, -0.1842992,
-0.1900707, -0.1835952, -0.1639596, -0.1694418, -0.2112418, -0.2200596,
-0.2309477, -0.183575, -0.105267, -0.05079828, -0.00119349, 0.004840307

 

[ilit_shlezak@weizmann_ac_il]

Hi,
I have the same question.
The ice_cov variable units defined as [fraction], so what is the meaning of a negative value?
Thanks,
Ilit

 

[olson]

I did not write the algorithm but had it explained to me by the developers
when I ran into the same problem. So you're hearing it third-hand:


The model reads in the monthly climatology and then time-interpolates the
monthly mean values to the current model time. It's this time-interpolated
value that's actually used by the model simulation. But there is a subtle flaw
in the interpolation of monthly mean data: if you were to reconstruct the
monthly mean by averaging all the interpolated values over one month of a model
run you often will not get back the original monthly mean. So the model is
not being forced by the "true" monthly mean of ice fraction. This is not an
issue for that part of the time series that is linearly increasing/decreasing
or staying steady from month to month. But, for example, there are significant
errors when spanning a local min or max in the time series.

Say, for example, there is a point that has 0.3 ice fraction in Aug, 0.01 in
Sept and 0.4 in Oct. Then, if you were to run the model during the entire
month of Sept and average up all the resultant interpolations you would get
about 0.0875 rather than 0.01.


The solution was to modify the boundary dataset for all the months such that
the model was always being forced by the true monthly mean. This means that
sometimes ice fraction had to be made less than 0 or greater than 1 in building
the modified dataset (IIRC, the actual results of the interpolation themselves
are always bounded by 0 and 1 during a model simulation - the model never sees
a negative ice fraction).


So if you want to plot the *actual* monthly mean of ice_cov, the boundary
dataset has a companion "pre-diddled" (not MY choice of words) field that
IS the actual monthly mean. An "ncdump" of the dataset gives:

float ice_cov(time, lat, lon) ;
ice_cov:long_name = "BCS Pseudo Sea-ice concentration" ;
ice_cov:units = "fraction" ;
float ice_cov_prediddle(time, lat, lon) ;
ice_cov_prediddle:long_name = "Sea-ice concentration before time diddling" ;
ice_cov_prediddle:units = "fraction" ;


"ice_cov_prediddle" is the field you want to plot.

 

[ilit_shlezak@weizmann_ac_il]

Thank you Jerry for the detailed answer.

I am using the file 'sst_HadOIBl_bc_64x128_1949_2001_c020411.nc' from the T42 datasets. This file doesn't contain the 'prediddle' variables (ice_cov and SST).
In my case, besides plotting the field, I need also to modify it in some locations.
Any idea how to do it in a way that will keep the averages correct?
Even if I had the 'prediddle' variable (which I have for my LGM run), I would need to change them both, and I assume that they need to be changed together somehow.

Thanks,
Ilit

 

[olson]

If you have access to "bluefire" at NCAR, the boundary dataset:

/fs/cgd/csm/inputdata/atm/cam/sst/sst_HadOIBl_bc_64x128_1949_2004_c050526.nc

does contain the true monthly mean as well as the modified. It is the same dataset as
the one you're using except extended to the year 2004

If you don't have access let me know and I can get it to you.


And, as far as modifying it is concerned, look in the directory of CAM code:

models/atm/cam/tools/icesst

There is a README and tools to help you with the time-diddling of a dataset that you create

 

[ilit_shlezak@weizmann_ac_il]

I don't have access to "bluefire" at NCAR.
Can you help me with getting the file?
Thanks,
Ilit

 

[olson]

I put the file on ftp

ftp to "ftp.cgd.ucar.edu"
login as anonymous and use your email address as the password
cd to "pub/olson"
type "bin" (for binary)
get sst_HadOIBl_bc_64x128_1949_2004_c050526.nc