CTSM5.3 historical simulations: Known issues

[wwieder]

These thoughts come from from looking at diagnostics and ILAMB results posted with #69.

Issues we may want to address with the PPE or hand calibration

• Dead grasses (see maps below, but also posted here, related to FUN bug or fire parameterization of froot_leaf?)
• High LAI in:
  • tropical deciduous forests (related to fire parameterization of froot_leaf?)
  • larches (deciduous needleleaf trees)
• Permafrost extent (maybe we're OK, or consider modifying Sturm thermal conductivity parameterization)?
• Water stress in the Amazon (related to ad-hoc kmax parameter change to increase ET and reduce LAI)
• High vegetation biomass, especially in the tropics (increase rmort, modify LAI and GPP?) ?
• [Fire evaluation] including fire C loss & fraction burned area
• Crop model calibration (not listed here).

[samsrabin]

Could you point me toward the diagnostic plots of a run with "better" C4 Grass distribution?

[wwieder]

do you mean a run with living C4 grass?

[samsrabin]

Yeah, exactly.

[wwieder]

These are GSWP3 forced runs with CLM5.1 (therefore dated), where C4 grasses still don't look great, but less bad than they are now.

@olyson will look at the TRENDY runs that Adrianna did with 5.3 datasets, CRUJRA datm & the older parameter file (w/ out froot_leaf changes from fire work.

[djk2120]

I looked at some PPE runs we had (GSWP3, CLM5.1 default params) with the FUNbug and then with that rectified. Fixing the FUNbug reduced C4 LAI by around 33% on average, but most of your problem gridcells were already dead.

[olyson]

Survivability and Max TLAI for the TRENDY run (TRENDY2024_f09_clm60_S3):

[wwieder]

Thanks Keith. C4s look pretty dead before changes to froot_leaf. Hopefully this is something the PPE tools can help address?

[olyson]

I was looking for GSWP3 driven historicals that are newer than the CLM5.1 simulation that @wwieder linked above and saw that I had a historical that was used for nuopc validation. It used cesm2_3_alpha03a (ctsm5.1.dev038). Below are survivability and max LAI plots. Survivability looks better than the TRENDY simulation (84% versus 62% in the TRENDY simulation) and LAI is higher.
There obviously are science differences between these tags, but there are parameter differences for C4 grasses as well (including froot_leaf as we know), summarized here:

            NEW    OLD
cc_leaf     0.6    0.8
cc_other    0.5    0.8
fm_leaf     0.2    0.8
fm_other    0.2    0.8
froot_leaf  0.5    1.5
fsr_pft     0.31   0.32
kc_nonmyc   0.12   7.2
kn_nonmyc   7.2    0.12

[wwieder]

Thanks for digging in here Keith. My other suspicion was that the FUN bug that increased n costs killed the C4 grasses. In CLM5 they were strongly n limited, correcting the fun bug may have done them in?

[samsrabin]

@lifang0209 noticed a large excess of burned area on the Tibetan Plateau. From a comment on Issue #2728: Incorporate the impact of O2 concentration on fuel flammability, which proposes a parameterization to handle this:

The attached figure displays the spatial distribution of the annual burned area fraction. Observations (obs) represent the average of three satellite-based burned area products (GFED5s, FireCCI, and MODIS C6). Simulations show results for the new fire schemes both with and without an O₂ limit. The figure includes global burned area totals and spatial correlation coefficients (R) between observations and simulations. As illustrated, the O₂ limit reduces the overestimation of the burned area fraction in Tibet, and improves the spatial pattern simulation performance (global spatial correlation with observations increases from 0.76 to 0.79).

[wwieder]

Quick update on survivability from #74 that includes updates to:

• Nfix temperature function and
• Arctic FUNfrac_fixers
• Backs off arctic tuning from the 5.3.0
• Reverts froot_leaf in C4 grass and tropical deciduous trees (from Fang's tuning)
• Makes N update cheeper for C4 grasses

My assessment, compared to results in #70 (first firgure below), results from #74 (second figure below)

• Arctic grass and shrubs looks good
• Tropics look much better w/ C4 living and deciduous trees less absurd
• C3 grasses look really different that C3 generic crops. Why is this?
• Changes to boreal evergreen trees likely need to be backed off too, with lower N fix in the arctic.

[lifang0209]

Does "Reverts froot_leaf in C4 grass and tropical deciduous trees (from Fang's tuning)" mean you prefer to set froot_leaf back to 1.5? Setting froot_leaf=0.5 is beneficial for their growth and post-fire recovery.

[wwieder]

This may be true, but the parameter change seems to contribute to high LAI biases in deciduous trees (mean max LAI > 12!) and may contribute to C4 grass mortality (especially in Africa).

[wwieder]

Modifying froot_leaf and reducing C4 grass N uptake costs seems to help with these biases.