Improve replacement of urban and pft parameters

[s-poll]

Currently, the eCLM land cover surface file is generated using the LUH2 dataset and then in a further step replaced with the target land cover (e.g. globecover2009).

Replacing the land cover means that there are grid points that had no land or urban parameters in the original surface file, but now require them. These are currently replaced with average values of these parameters from the original surface file.

The replacement of urban and land cover parameters should be based on the raw data of the parameters.

[mvhulten]

I have locally rebased this changeset on misc-fixes (this was trivial) and will test replace_domfile2tsmp.r.

For replace_surdat2tsmp.r and replace_topo2tsmp.r I probably need to fix #21 first.

Even though I have the rights to write in /p/scratch/cslts/poll1/sim/euro-cordex/tsmp2_workflow-engine/dta/geo/eclm/static/, maybe I do not want to. I suppose I should consider the replace_*.r scripts for now as unfinished, and use my own writable directory somewhere?

[s-poll]

A prior correction of #21 is not necessary.

The scripts are meant so that you have to adjust the paths before execution. This includes in particular the output directory

[mvhulten]

How can I enter this incomprehensible, loonatic R command that I came up with on a Unix prompt?

> install.packages("ncdf4", configure.args = c(ncdf4 = "--with-nc-config=/p/software/default/stages/2025/software/netCDF/4.9.2-gompi-2024a/bin/nc-config"))

I tried this bizar command:

$ R CMD INSTALL --configure-args="--with-nc-config=/p/software/default/stages/2025/software/netCDF/4.9.2-gompi-2024a/bin/nc-config" ncdf4
Warning: invalid package _ncdf4_                                                                                                                                            
Error: ERROR: no packages specified

Should this not do the same thing as the command from the R prompt?

---

Reminder to self: I've rebased for my consistent development; not going to push the rebased local branch (dev_replace_tsmp2_REBASED), but rather cherry-pick that anew onto the the original dev_replace_tsmp2.

[kvrigor]

@mvhulten If Rscript is in your $PATH you can try this:

cat << EOF > install.R
#!/usr/bin/env Rscript
install.packages("ncdf4", configure.args = c(ncdf4 = "--with-nc-config=/p/software/default/stages/2025/software/netCDF/4.9.2-gompi-2024a/bin/nc-config"))
EOF

chmod +x install.R && ./install.R

[kvrigor]

Forget my previous post... this solution should work:

module --force purge
module use $OTHERSTAGES
module load Stages/2025
module load Intel
module load ParaStationMPI
module load netCDF
module load R

CRAN_MIRROR="https://ftp.gwdg.de/pub/misc/cran/"
R_LIBS="$HOME/R-packages"
mkdir -p $R_LIBS
Rscript -e "install.packages(\"ncdf4\", repos = \"${CRAN_MIRROR}\", lib=\"${R_LIBS}\")"

[mvhulten]

I am able to run all three R scripts succesfully (only netCDF path variables must be changed), but since yet another language—and R is one that I really hate—would increase the overall complexity, Stefan might rewrite the R scripts in the tolerable Python 🙏.

Replacing the land cover means that there are grid points that had no land or urban parameters in the original surface file, but now require them. These are currently replaced with average values of these parameters from the original surface file.

The replacement of urban and land cover parameters should be based on the raw data of the parameters.

I'm not sure if I understand. Are these grid point near the boundaries only? Then you'd (still) get discontinuities maybe.