Opened 7 months ago

Closed 7 months ago

#2919 closed help (answered)

fluxnet forcing number of rows

Reported by: awright Owned by: pmcguire
Component: JULES Keywords: fluxnet jules
Cc: Platform: JASMIN
UM Version:

Description

Hello Patrick,

I want to make scatter plots of JULES outputs (e.g. latent_heat in AT_Neu-r9227_branch-presc0.D.2006.nc to AT_Neu-r9227_branch-presc0.D.2012.nc) to its FLUXNET forcings (e.g. air temperature from /vn1.0/fluxnet/AT_Neu-met.dat). That is to find out how the forcings affect the model output. The forcing dataset does not have any date columns, but I assume they are hourly. Having said that, when I plot the air temperature, the cycle suggests that it might be half-hourly. Taking AT_Neu-met.dat as an example, I have 192913 rows, divided by 24*365 it will be 22 years of data (or 11 if half-hourly). Whereas in my model output, I have years 2006 to 2012 for the same site, which makes it 7 years of data. How can I find the corresponding dates in the fluxnet forcing datasets? Do you happen to have the forcing files with date columns? Any suggestions or hints will be appreciated.

Many thanks,

Azin

Change History (4)

comment:1 Changed 7 months ago by pmcguire

Hi Azin
If you look at the JULES namelists in the Rose-Cylc suite files, you can see what is going on.

First, note that ~/roses/u-al752/suite.rc file defines the RUNBEGDATE and RUNENDDATE variables for each site.

Second, note that these variables, RUNBEGDATE and RUNENDDATE are used in ~/roses/u-al752/app/jules/rose-app.conf in the jules_time namelist, for example.

Third, note that the jules_time and jules_drive namelists are amended in ~/roses/u-al752/app/jules/opt/rose-app-drive-AT_Neu.conf, for this particular site. Also note that the jules_drive namelist defines the data_start,data_end, and data_period values for the driving data, and the jules_time defines the timestep_len for the simulation. In this case, the driving data_period is the same as the JULES simulation timestep_len, which is 1800 seconds = 30 minutes.

So you can construct the scatterplots that you want by definining the time axis from those values.

It might take extra work to compare the daily average values in AT_Neu-r9227_branch-presc0.D.2006.nc to the half-hourly driving data in /group_workspaces/jasmin2/jules/pmcguire/fluxnet/kwilliam/suite_data/vn1.0/fluxnet/AT_Neu-met.dat. An easier way might be to compare the hourly data in AT_Neu-r9227_branch-presc0.H.nc to the output hourly driving data in AT_Neu-r9227_branch-presc0.H.nc. There are hourly variables (that are very similar to the original driving data) for precip, t1p5m_gb, q1p5m_gb, and sw_down in that file. They are hourly averaged (2 time steps of averaging, in this case), and the temperature and humidity variables are averaged over surface-tile type. If you do the scatterplots with the hourly output data, for the latent heat vs. temperature, for example, you can avoid having to define the half-hourly time axis and doing the daily averaging. There are other output variables in the daily-average output NETCDF files, though.

Does this help?
Patrick

comment:2 Changed 7 months ago by awright

Yes Patrick, Thank you very much!

All the best,

Azin

comment:3 Changed 7 months ago by pmcguire

Excellent!
I will close the ticket now.
Patrick

comment:4 Changed 7 months ago by pmcguire

  • Resolution set to answered
  • Status changed from new to closed
Note: See TracTickets for help on using tickets.