You’re welcome!
Patrick
Hi Tristan
I see the latent heat flux, the sensible heat flux, and the net surface-atmosphere exchange (NSAE) in the bundled eddy-covariance data for the NEON sites. I think the NSAE is the same thing as Net Ecosystem Exchange (NEE) for CO2. These are 3 out of the 5 data columns in Karina’s energy+carbon observational data sets for the FLUXNET sites. I think I will go with that for now. The other two are GPP and RECO (respiration), which could involve looking at the night-time fluxes, in order to partition the NEE between GPP and RECO. That processing of the partitioning could be deferred for the future.
Patrick
Thanks Patrick. NEON do their own partitioning of the data. I am not sure where to access it, but I will take a look when I get chance.
Hi Tristan:
I will look again for the GPP & RECO data from NEON. I think I saw a refereed journal article about it somewhere.
I was able to get some code working for producing a daily text file for each of the 9 sites that has the Net Ecosystem Exchange (NEE) of CO2, the Sensible Heat Flux (SH) of temperature, and the Latent Heat Flux (LE) from water vapor. I will work next on getting these overlain on the JULES model curves. This way, we can have more confidence that JULES is at least working halfway right, and I have the right units and UTC time shifts, etc.
Patrick
Hi Tristan:
I have been able to make some plots for Sensible Heat Flux (SH), Latent Heat Flux (LE), and Net Ecosystem Exchange (NEE) for the 9 NEON sites, with the observed fluxes overlain on the JULES model fluxes, with the JULES FLUXNET suite u-al752. They are below.
I think the units for both the observed fluxes and the model fluxes might be right, since there is some correspondence between the monthly-average curves for the observed & model fluxes. But it is still possible that there are mistakes in the code, here and there, which we would need to sort out.
The thick blue lines are the monthly average observed fluxes, whereas the thick dashed blue lines are the 1-standard-deviations of the daily observed fluxes over the month. The thin blue lines are the daily observed fluxes.
Correspondingly, the thick black lines are the monthly average JULES-model fluxes, whereas the thick dashed black lines are the 1-standard-deviations of the daily JULES-model fluxes over the month. The thin black lines are the daily JULES-model fluxes.
The gap filling is forward-fill (from the last known value). The gaps are rather easily seen by the flat lines in the monthly-average data.
The correspondence of observed vs. model for the NEE for these NEON sites is somewhat tenuous. But there are some correspondences. Maybe, there needs to be some outlier-filtering applied of the daily data, prior to the monthly-averaging, or maybe, there is something wrong with my JULES settings, or maybe, something else is wrong. And the plotting could of course be done better. I switched off the auto-scaling for the y-range, and chose the y-range manually, due to some of the extreme values. This also had the advantage of keeping the y-range the same for the different NEON sites.
Not sure what I will do next. Maybe I will add a few more NEON sites next.
Patrick
Latent Heat Flux
Sensible Heat Flux
Net Ecosystem Exchange
Thanks Patrick. I’ll have a think about these.
Hi Tristan:
I have added 5 more NEON sites to the JULES FLUXNET/LBA/NEON suite u-al752:
{CLBJ, KONZ, NIWO, ONAQ, ORNL}. That makes 14 in total. I ran the suite, which runs JULES for about 8 spinup cycles, and made this 3 sets of plots, comparing JULES model to NEON observed for Latent Heat Flux (LE), Sensible Heat Flux (SH), and Net Ecosystem Exchange (NEE). Similar concerns apply for the correspondence of model to obs NEE, as was discussed here on 5 December. After the AGU and the New Year, I will continue working on this suite. See the plots below.
Patrick
Latent Heat Flux
Sensible Heat Flux
Net Ecosystem Exchange
1 Like
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.
Hi Tristan:
I just added 9 more NEON sites. That makes 23 in total. This includes all 20 of the Core Terrestrial sites for NEON. I am running the revised JULES FLUXNET/NEON/LBA suite now. There might be some run-time bugs to sort out. But after that, I will try to get the new NEON plots to you here.
I do note that there were inadvertently overrides in the prior plots here, which overrode the vegetation class to be C3 grass for all the NEON sites. I fixed this in the new version, so that there are evergreen (EBF, ENF), deciduous (DBF,DNF) trees, or shrub/scrub (SH) as appropriate. Also, the vegetation tile fractions for the different NEON sites probably need to be estimated more accurately somehow, but we can work with my default tile-fraction assignments for now, based upon the IGBP cover.
After this run is complete, I probably need to do several things:
- figure out why the NEE model curves aren’t agreeing better with the NEE observation curves;
- Try to get GPP and RECO observed/partitioned data;
- Try out the Ameriflux version of the NEON site data, which includes data already processed to CSV files;
- Estimate the vegetation tile fractions for each site.
Patrick
Hi Tristan & Natalie:
Here are the plots for the 23 NEON sites.
I changed 3 of the sites from my setting of EBT (Evergreen Broadleaf Tree) to a setting of BET-Tr (Broadleaf Evergreen Tree - Tropical). Maybe this and similar IGBP settings for the other sites need to be changed further, as well.
Patrick
Latent Heat Flux
Sensible Heat Flux
Net Ecosystem Exchange
Hi Tristan:
I spoke with Natalie Douglas today again. She is making good progress with the u-al752 suite for the LBA FLUXNET site. She chose that site since I had only been able so far to get the NEON-sites working without GPP observational data, and since the LBA FLUXNET site has GPP observational data. She is now wanting me to get the NEON sites working with GPP observational data. I will work on that. I had previously only been able to get the NEE observational data into the NEON version of the suite. I have yet to find the GPP observational data in the NEON data files.
Patrick
Hi Patrick,
I will make some enquiries about this. I am sure there is a product that exists.
Hi Tristan:
Thanks! Much appreciated.
I can also try to find out more about GPP observational data myself.
But any help you can give or get would be quite helpful.
Patrick
FYI - I have drawn a blank so far.
Hi Tristan:
Thanks. I will try to look further into myself. Whom did you contact already, or did you look at some online resource?
Patrick
Mike Dietze, Boston. I was chatting with him about it at AGU. I sure he knows the answer, but is not responding to email.
Response from Mike below.
My take away from this is that (a) there is software called OneFLUX that does the job - but I am not sure we should go down that route as could be quite time consuming and (b) the AmeriFlux data base also contains NEON sites, including derived GPP. If we can get the data from AmeriFlux, to me, that would seem adequate.
Sorry for the slow reply. Alexis in my group was going to work on getting OneFLUX up and running on our own system, but she’d decided to leave with a MA after finishing her first paper (a basic characterization of our flux forecasting system) so she’s not going to get to this and I’m not going to have the bandwidth to pick up on this. The OneFLUX code is public on Github. Trevor’s the only one I know who’s actively been using it. I think OneFLUX is what Ankur recommended as a starting point too, but there’s definitely things that could be done better.
I think if I were working with older data I’d probably download the NEON flux data through Ameriflux (their partitioning, which is now based on OneFLUX) rather than NEON’s own data, but for our purposes the latter has the critical difference of a 5 day latency rather than a multi-year latency on Ameriflux.
Thanks, Tristan:
I will try to implement the AmeriFlux solution, as you and Mike suggest.
But it might take a few weeks for me to do that.
Patrick