Hi,
Now I would like to seek some advice on collocating UM AOD with MODIS AOD. Even though I found a similar topic down below, that did not seem to include helpful information for me. Similarly to AERONET, probably I should use L2 (instantaneous) data instead of L3 daily average data. For UM, I have 3 hourly AOD data. Here is the work flow I'm thinking;
1. collocate UM 3 hourly AOD with MODIS L2 AOD for a day. The MODIS variable name is Optical_Depth_Land_And_Ocean.
2. probably I will create a global 1x1 degree grid and take grid average AOD for each grid for a day.
3. Do 1 and 2 for all days for a month or two in year 2008 and then take monthly averages of both MODIS and collocated UM output.
4. Compare monthly averages.
Can you see any problem or improvement here? I think I can do steps 1, 3 and 4 using CIS? Can I do step #2 with CIS as well? Or alternatively, should I use L3 daily data and somehow assume measurement times? They should be
For step #1, I tried doing this;
$ cis col aod550_total:aod550_total_teafwa_pb20080101.nc /group_workspaces/jasmin2/crescendo/Data/MODIS/Terra/L2/MOD04_L2.A2008001.0000.006.2015005072958.hdf -o aod550_total_teafwa_pb20080101_col_MOD04_L2.A2008001.0000.006.nc
the output file was created but all values were missing, just like my earlier attempt of collocating with AERONET data. I tried specifying variable name in MODIS;
$ cis col aod550_total:aod550_total_teafwa_pb20080101.nc /group_workspaces/jasmin2/crescendo/Data/MODIS/Terra/L2/MOD04_L2.A2008001.0000.006.2015005072958.hdf:variable=Optical_Depth_Land_And_Ocean -o aod550_total_teafwa_pb20080101_col_MOD04_L2.A2008001.0000.006.nc
but the result is the same.
Could you please give me any advice here? I copied both aod550_total_teafwa_pb20080101.nc and MOD04_L2.A2008001.0000.006.2015005072958.hdf to JASMIN home directory: /home/users/myoshioka/temp/ .
Thanks,
Masaru
Hi Masaru,
I think this is a similar issue as the AeroNet: the time points from the MODIS data are 2008-01-01 00:00:07 to 2008-01-01 00:05:06, but the Model's first tilmestep is 00:20:
>>> import cis
>>> c=cis.read_data('/home/users/myoshioka/temp/aod550_total_teafwa_pb20080101.nc', 'aod550_total')
print(c.coord(>>> print(c.coord('time').points)
[ 149019.01388889 149019.13888889 149019.26388889 149019.38888889
149019.51388889 149019.63888889 149019.76388889 149019.88888889]
>>> print(c.coord('time'))
DimCoord([2008-01-01 00:20:00, 2008-01-01 03:20:00, 2008-01-01 06:20:00,
2008-01-01 09:20:00, 2008-01-01 12:20:00, 2008-01-01 15:20:00,
2008-01-01 18:20:00, 2008-01-01 21:20:00], standard_name='time', calendar='gregorian', var_name='time')
So the interpolation is returning fill values because it won't extrapolate by default. You could include the previous (and next) day in the collocation. All the later MODIS files should work though.
Otherwise I think your workflow looks good though!
Hi Duncan,
You are right! Because I had discarded all data for 2007 as spin-up, I duplicated aod550_total_teafwa_pb20080101.nc and created aod550_total_teafwa_pb20071231.nc. I changed the values of time manually. For info here is how;
================================================
ncdump aod550_total_teafwa_pb20080101.nc > aod550_total_teafwa_pb20080101.cdl
cp aod550_total_teafwa_pb20080101.cdl aod550_total_teafwa_pb20071231.cdl
edit data manually with text editor. In this case I subtracted 24 from all time values.
ncgen -b aod550_total_teafwa_pb20071231.cdl
This creates aod550_total_teafwa_pb20071231.nc
================================================
Then I did collocation like this;
cis col aod550_total:aod550_total\_teafwa_pb20071231.nc,aod550_total\_teafwa_pb20080101.nc /group_workspaces/jasmin2/crescendo/Data/MODIS/Terra/L2/MOD04_L2.A2008001.0000.006.2015005072958.hdf -o aod550_total_teafwa_pb20080101_col_MOD04_L2.A2008001.0000.006.nc
This seems to have worked!
cis plot aod550_total:aod550_total_teafwa_pb20080101_col_MOD04_L2.A2008001.0000.006.nc --vmin=0.0 --vmax=0.8 &
shows collocated data. For record MODIS data was plotted like this;
cis plot Deep_Blue_Aerosol_Optical_Depth_550_Land_Best_Estimate:MOD04_L2.A2008001.0000.006.2015005072958.hdf --vmin=0.0 --vmax=0.8 &
See the plots here;
https://drive.google.com/open?id=1c08T3VT4YIym5r5iuKnLGb0Gv6jdph4R
Nest step will be try and see if I can practically loop over hundreds of MODIS data and collocate UM data on them. If that is successful, I will think about how to construct global data. Or it might be better to combine all MODIS data for a day first and then collocate the UM data. I'll play around and decide it.
Thank you for your help!
Masaru