Opened 8 months ago
Last modified 7 weeks ago
#2838 accepted project-task
help with enabling hourly plotting and analysis in JULES/FLUXNET suite, for CSSP/PORCELAIN
| Reported by: | pmcguire | Owned by: | pmcguire |
|---|---|---|---|
| Component: | JULES | Keywords: | hourly, daily, JULES, FLUXNET, Python, CSSP China, PORCELAIN |
| Cc: | chuntingford, kwilliams, rfu | Platform: | JASMIN |
| UM Version: |
Description
Patrick
Can you help with enabling hourly plotting and analysis in the JULES/FLUXNET suite u-al752 (Python post-processing), for the CSSP/PORCELAIN project? We want to do the RSME of the difference between the FLUXNET observations (ignoring the night-time data) and the JULES models for the 10 China FLUXNET sites. Currently, the daily averages are being plotted and analyzed instead of the hourly data.
Chris H
Attachments (2)
Change History (9)
comment:1 Changed 8 months ago by pmcguire
- Status changed from new to accepted
comment:2 Changed 8 months ago by pmcguire
- Cc chuntingford added; chg@… removed
comment:3 Changed 8 months ago by pmcguire
- Cc kwilliams added
comment:4 Changed 8 months ago by pmcguire
Changed 8 months ago by pmcguire
Changed 8 months ago by pmcguire
comment:5 Changed 7 months ago by pmcguire
The plot_diurnal.py code has been adapted for JASMIN, and the hourly data for the Chinese FLUXNET sites has been obtained from Karina.
Patrick
comment:6 Changed 7 weeks ago by pmcguire
The location of the hourly data for the Chinese FLUXNET sites is:
/group_workspaces/jasmin2/jules/pmcguire/fluxnet/kwilliam/suite_data/CN1.0/fluxnet_obs/subdaily_obs/
I didn't want to merge this CN1.0 subdaily data ('CN' means 'China') with the regular vn1.0 daily data for the non-Chinese FLUXNET sites.
The regular most up-to-date data for all the FLUXNET sites is currently:
/group_workspaces/jasmin2/jules/pmcguire/fluxnet/kwilliam/suite_data/vn1.1
Patrick
comment:7 Changed 7 weeks ago by pmcguire
- Cc rfu added
Hi Becky
I had adapted for ChrisH this plot_diurnal.py by Karina for plotting diurnal cycles for BET-Tr sites, so that it can work on JASMIN with the Chinese FLUXNET subdaily/hourly data. This is only meant to be some test code with the right JASMIN path and the right FLUXNET site names. The dry-year and wet-year years are not accurate.
The new file is plot_diurnal_JASMIN_TEST.py:
https://code.metoffice.gov.uk/trac/roses-u/browser/a/l/7/5/2/u-al752-jpegpaper/bin/plot_diurnal_JASMIN_TEST.py
The latest version of Karina's original code is at the link in a previous comment in this ticket.
I had branched the code at the link here from an earlier version of her code.
Patrick
From ChrisH (chuntingford): [edited for this ticket]
Chris & Karina talked about Lina's idea of only looking at daytime values, in the optimisation. That is, ignoring invariant night or twilight values. This might give strong clues to the true optimum.
Changing all the existing RMSE code with a sub-daily mask might take some time, and Karina has a better way. This is to adapt code below, that explicitly looks at the diurnal cycle. Here:
https://code.metoffice.gov.uk/trac/roses-u/browser/a/l/7/5/2/u-al752-jpegpaper/bin/plot_diurnal.py
Sample plots attached. One is for one month, all days of FLUXNET. The simpler plot is the diurnal cycle.
So if we can make some plots similar to "FR_Puediurnal_GPP.PNG", we will be in a strong position to really get at good parameters.
In summary, this builds on what we've been doing in Reading, but gets more directly at looking at sub-diurnal features. More "by eye" checking.