I’m having a problem with running simulations with land use change in payu.
Payu configurations with land use change call a pre.sh script which calls a python script
Re-scales the new cover fraction so that they add up to 1.
Does nothing more if the old and new cover fractions are the same.
Inserts the new land cover fractions into the restart file.
Initialise variables for any new tiles that are created by the changed land-use with grid averages.
Set variable values to 0 for tiles that transition to 0 cover fraction.
Re-scales most of the 8xx code variables proportionately for the new cover fractions.
2 is basically always untrue due to poor handling of missing values in the old and new fractions.
4 should only be necessary for certain variables such as soil and snow temperature.
5 might also be unnecessary, since CABLE won’t use them anyway.
6 Seems problematic to me, since CABLE already does this in subroutines like newplant() and it’s not something that should be done twice.
Also, it does not explicitly insert the previous year’s land use change into the correct code, but instead relies on 6. to do something similar.
This is quite a bit more than the old ksh scripts for running ACCESS, which as far as I can tell, simply insert the new fractions into the restart file (code 216), and move the old ones to the previous year (code 835). This is ok, since land-use change that creates new tiles never occurs in existing simulations.
I’m doing some simulations with very large land use change, where all crops are turned into tree PFTs. This means, new tree PFT tiles are created and need to be initialised. When looking at the total wood carbon pool of the 4 forest PFTs, there are these strange large drops at the start of each year. This happens only for grid cells where land-use change occurs. I would have expected the wood carbon pool to start at the same point as the simulation without land use change.
The orange line is the large drops in cwood.
The blue line I simply switched off land use change after the first year.
The green line is just a regular pre-industrial simulation.
The red line is with land use change turned on but it runs the update script to insert the regular 1850 land cover maps each year. I was surprised that it is slightly less than the regular pre-industrial simulation. Maybe this has always happened in other simulations but no one noticed because the land use change is usually so small?
I managed to mitigate the problem partially by fixing some of the obvious issues above with 2. and 6. But I’m not exactly sure what the cause is otherwise.
Below is the script i used to create the no-crops land use map:
And the land use map on gadi