Opened 8 months ago

Last modified 3 months ago

#2632 accepted project-task

CALIBRE downscaling

Reported by: pmcguire Owned by: pmcguire
Component: Land Surface Modelling Keywords: CALIBRE, SDVGM, downscaling, ECMIP
Cc: t.l.quaife@… Platform: Other
UM Version:

Description

Can you get CALIBRE setup with downscaling and the SDVGM model?
SDVGM is currently on the met-cloud. The docs haven't kept up with the source code.
Then get it running with ISIMIP.
SDVGM is a land surface model with vegetation and terrestrial carbon balance.

Change History (12)

comment:1 Changed 8 months ago by pmcguire

  • Status changed from new to accepted

comment:2 Changed 8 months ago by pmcguire

Hi Tristan

I was able to compile successfully the sdgvm0 code. I tried running it. But I don't belong to the essc group and I don't have permissions

to read the subdirectories of /export/cloud/nceo/users/db903833/sdgvmDrivers/SDGVMdata/ .
Can you fix this?

comment:3 Changed 8 months ago by pmcguire

Thanks for fixing the permissions. It runs now.

Tristan says:
"The next job is get data from ISIMIP (https://www.isimip.org/) and convert it into the format of the driving data used by the model.

In the first instance just a single model for RCP2.6 would be ideal. I assume once it's done for one ISIMIP dataset it will be easy to repeat for any of them."

comment:4 Changed 7 months ago by pmcguire

I now can download ISIMIP forcing data (RCP2.6). I am working on reformatting this data and finding sufficient disk space for the data.

comment:5 Changed 4 months ago by pmcguire

I have written Python scripts to reformat the ISIMIP forcing data (RCP2.6) from NETCDF format to the text format for SDGVM (Sheffield Dynamic Global Vegetation Model). I have run the SDGVM model with this ISIMIP data.

Still need to:
— get the right CO2 driving data into the model for RCP2.6,
— find out what the minimum amount of rain for a day to be considered to be a rain day,
— repeat this exercise for other RCP's.

comment:6 Changed 4 months ago by pmcguire

Tristan has sent the CO2 driving data for RCP2.6.
This came from https://tntcat.iiasa.ac.at/RcpDb/dsd?Action=htmlpage&page=welcome
which Tristan thinks is the official source of the scenario data fro CMIP5.

comment:7 Changed 4 months ago by pmcguire

Hi Patrick,

I've spoken to the two main developers of SDGVM and neither of them can remember what the threshold used for the rainy days was. I'm not sure where that leaves us, but one thing to do might be to compare the number of rainy days calculated toward the first decade of an ISIMIP data set with the existing stats file. We wouldn't expect it to be the same - but it shouldn't be drastically different either.

Another approach might be to find a point where the weather is very sensitive to that threshold and then see how sensitive the model outputs for carbon variables (GPP, NPP, NEE) are. If the carbon's not that sensitive then we don't need to worry too much.

My hunch is that for the CALIBRE project itself we probably don't need to be too worried, although of course it's a good thing to look into.

Tristan.

comment:8 Changed 4 months ago by pmcguire

Hi Tristan

Thanks for asking around and for the recommendations.

I just did some spot checks for 8 of quasi-randomly chosen lines in the 259200 line stats/raindays.dat
files. I compared the ISIMIP raindays I computed from 2005-2099 (with 0.1 mm/day minimum) with the ERA40/30min rain days that were previously computed.
And I estimated the min # of raindays/month and the max # of raindays/month for the 12 months (averaged over the time span of the data).
These min and max raindays/month were very similar for both the ISIMIP and for the ERA40 for the 6 out of the 8 lines I looked at.
The average raindays/month (averaged over the 12 months) was close to the same for both datasets for 8 out of the 8 locations I looked at.

I didn't note any notable trends in the ISIMIP monthly averages from 2005-2099 in the pixels I checked.

So I tentatively conclude that the 0.1mm/day rainday threshold in my CDO script for ISIMIP is not too far off.
Patrick

comment:9 Changed 4 months ago by pmcguire

I have copied the RCP2.6=RCP3PD co2 data (CO2_RCP3PD_MIDYR_CONC.DAT) to CEDA JASMIN
IN THIS DIRECTORY:
/gws/nopw/j04/odanceo/pmcguire/sdgvm

and I have written an awk/bash script (co2_rcp2.6.bash) to convert this data from parts per million to parts per 100000.
The years from 2013 to 2500 were extracted by hand from the output file, and the bash script co2_rcp2.6splice.bash merges
the original SCRIPPS_spline_merged_ice_core_yearly.dat file with the new CO2_RCP3PD_MIDYR_CONC_2cut.DAT
to make a new SCRIPPS_spline_merged_ice_core_yearly_merged2_RCP3PD_MIDYR.DAT file

This co2 file can now be used by SDGVM, with an example input file modified to be input_calibre_test5_ISIMIP_JASMIN.dat as called by test1b.bsub .

comment:10 Changed 4 months ago by pmcguire

Hi Tristan

I found a discussion of the humidity > 100% for ISIMIP . See section 3.2 of this document.

Fact Sheet: Bias correction in the ISI-MIP
https://www.isimip.org/documents/16/Fact_Sheet_Bias_Correction.pdf

Patrick

comment:11 Changed 4 months ago by pmcguire

Good spot - thanks!

Given that one of their proposed solutions is simply to truncate the data at 100% then I think that gives us fair justification for doing the same.

Tristan.

comment:12 Changed 3 months ago by pmcguire

  • Type changed from task to project-task
Note: See TracTickets for help on using tickets.