Difference between input and output PCT_NAT_PFT data

[Andrie]

Dear Team,

I am running fully-coupled CESM2.1.2 simulations with CLM 5.0 (and BGC-Crop) under an SSP scenario. In my simulations, I am adding tree cover and thus removing shrub/grass/crop to make space for the trees. I ran the first two years of simulations for the baseline (fixed tree cover at 2015 level) and the linear afforestation simulations. While the simulations ran through without an error I am a bit confused about the output. I was writing out the PCT_NAT_PFT variable to double-check that the model is changing forest cover etc as expected. What is surprising is that there seems to be an offset between the input and output data. In the input data there is more trees to start with but less shrub and grass (I was keeping the bare soil as it is).

The below plot shows a timeline of the global mean changes in bare soil/forest/shrub/grass showing the input data in faint lines and the output data in strongly coloured lines, where the baseline data is dashed and the afforestation data is a solid line. I multiplied the natpfts with the natveg fraction and summed over the relevant natpfts (i.e. natpft 1-8 = forest etc). Here, the difference between the input and output is clearly visible. While the changes are around 0.02% per gridcell on a global scale they can be a bit larger when plotting the spatial maps. Note that there are only 2 points on this graph as I so far only ran 2 years of simulations to check.

Is this something I should worry about? Do you have any idea why this happens? Did the same happen in any of your simulations at some point?
The PCT_NAT_PFT variable is not dynamic right (in the sense that theoretically we would expect the same input and output)?

Thanks a lot for your help!

 

[oleson]

There are checks in the dyn_subgrid code to make sure that landunits add up to 100% within some tolerance, and pfts add up to 100% within some tolerance also. I wonder if there are some small adjustments being made accordingly. You could check to see if anything else is changing in the surface dataset, e.g., are crops static?
You could pick a gridcell with the largest differences and put some write statements in the code to see if any adjustments are being made.

 

[lawrencepj1]

Hi Andrie / Nora?,
My guess is that the reason for the difference between the first and second year is coming from the restart file you are using. The output variables are carried over from the previous run and only updated when the model reads the values in. Are you changing the crop area as this will impact the output natural vegetation amounts as well.