search

ESCOMP / CTSM

Community Terrestrial Systems Model (includes the Community Land Model of CESM)
http://www.cesm.ucar.edu/models/cesm2.0/land/
Other
310 stars 313 forks source link

Dead PFTs in PPE2_BHSon simulations #1274

Open wwieder opened 3 years ago

wwieder commented 3 years ago

Issue:

Decid. Temp trees and C4 grasses are dead at the end of the historical period in PPE2_BHSon simulations.

Notes

At the end of AD spinup decid. temperate trees and C4 grasses look OK in LMBIRCH, PPE2_BHSoff and PPE2_BHSon simulations.

Ideas:

Ideas to try in parallel to explore this and other possibilities:

olyson commented 3 years ago

BL Deciduous Temperate Trees are still at 84% survivability at the end of the pAD.
C4 grass is at 74%. I'll have some spatial plots once timeseries generation is done.

olyson commented 3 years ago

I've generated some survivability timeseries plots for the last 200 years of the pAD:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_1850pAD_TLAI.png

and the historical:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_hist_TLAI.png

The BL deciduous temperate and boreal are stable in the pAD, but then die off in the historical.

The C4 grass declines only slightly in the historical but I think the damage was done earlier in the AD/pAD.

dlawrenncar commented 3 years ago

Interesting. Looking through diagnostics plots, it looks to me like the vegetation in eastern US (probably temperate deciduous trees) is unhealthy at the beginning of the historical.

https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/set6/set6_crbStock_Eastern_US.png

This suggests to me that the decreasing area of live trees over the historical is actually probably reflecting trees dying that were already on verge of dying at end of spinup. So, I think that this result doesn't go against our previous hypothesis that with BHS on, a couple of PFTs are showing enough of an increase in N limitation associated with cooler soil temperatures that these plants are not healthy or not surviving.

For reference, here are the full set of diagnostics: https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/setsIndex.html https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/setsIndex.html

On Fri, Feb 5, 2021 at 2:48 PM Keith Oleson notifications@github.com wrote:

I've generated some survivability timeseries plots for the last 200 years of the pAD:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_1850pAD_TLAI.png

and the historical:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_hist_TLAI.png

The BL deciduous temperate and boreal are stable in the pAD, but then die off in the historical.

The C4 grass declines only slightly in the historical but I think the damage was done earlier in the AD/pAD.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774306523, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVE4OKVZJ53A2S5S44TS5RRTPANCNFSM4XFF5SSQ .

dlawrenncar commented 3 years ago

Talked with Keith and looked at postAD spinup. Vegetation carbon drops rapidly at beginning of AD spinup. Looks like the AD to postAD worked fine, but then TOTVEGC just drops everywhere over the course of about 40 years, ending up at a value that is 10% or so of what it should be. Strangely, LAI, GPP, NPP all look fairly reasonable, so hard to understand where this is coming from. Keith is continuing spinup simulations with BHS off, with BHS off and reverting NSTEM and related params back to default. With this set of spinups, should be able to establish if the issue is arising with BHS, with new parameters, or with this tag, which does have some other changes in it relative to the last tag that worked normally.

On Fri, Feb 5, 2021 at 3:29 PM David Lawrence dlawren@ucar.edu wrote:

Interesting. Looking through diagnostics plots, it looks to me like the vegetation in eastern US (probably temperate deciduous trees) is unhealthy at the beginning of the historical.

https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/set6/set6_crbStock_Eastern_US.png

This suggests to me that the decreasing area of live trees over the historical is actually probably reflecting trees dying that were already on verge of dying at end of spinup. So, I think that this result doesn't go against our previous hypothesis that with BHS on, a couple of PFTs are showing enough of an increase in N limitation associated with cooler soil temperatures that these plants are not healthy or not surviving.

For reference, here are the full set of diagnostics: https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/setsIndex.html https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/setsIndex.html

On Fri, Feb 5, 2021 at 2:48 PM Keith Oleson notifications@github.com wrote:

I've generated some survivability timeseries plots for the last 200 years of the pAD:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_1850pAD_TLAI.png

and the historical:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_hist_TLAI.png

The BL deciduous temperate and boreal are stable in the pAD, but then die off in the historical.

The C4 grass declines only slightly in the historical but I think the damage was done earlier in the AD/pAD.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774306523, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVE4OKVZJ53A2S5S44TS5RRTPANCNFSM4XFF5SSQ .

dlawrenncar commented 3 years ago

Looking a bit further, it looks like there is possibly a crash in mineral nitrogen in the first few years of postAD. Been looking at Amazon where signal is very easy to see. NPP does drop and allocation drops, so I think maybe this is carbon starvation. Strange that GPP, LAI recover to normal levels by year 30-40. New runs will be instructive, hopefully.

On Fri, Feb 5, 2021 at 4:15 PM David Lawrence dlawren@ucar.edu wrote:

Talked with Keith and looked at postAD spinup. Vegetation carbon drops rapidly at beginning of AD spinup. Looks like the AD to postAD worked fine, but then TOTVEGC just drops everywhere over the course of about 40 years, ending up at a value that is 10% or so of what it should be. Strangely, LAI, GPP, NPP all look fairly reasonable, so hard to understand where this is coming from. Keith is continuing spinup simulations with BHS off, with BHS off and reverting NSTEM and related params back to default. With this set of spinups, should be able to establish if the issue is arising with BHS, with new parameters, or with this tag, which does have some other changes in it relative to the last tag that worked normally.

On Fri, Feb 5, 2021 at 3:29 PM David Lawrence dlawren@ucar.edu wrote:

Interesting. Looking through diagnostics plots, it looks to me like the vegetation in eastern US (probably temperate deciduous trees) is unhealthy at the beginning of the historical.

https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/set6/set6_crbStock_Eastern_US.png

This suggests to me that the decreasing area of live trees over the historical is actually probably reflecting trees dying that were already on verge of dying at end of spinup. So, I think that this result doesn't go against our previous hypothesis that with BHS on, a couple of PFTs are showing enough of an increase in N limitation associated with cooler soil temperatures that these plants are not healthy or not surviving.

For reference, here are the full set of diagnostics: https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/setsIndex.html https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/setsIndex.html

On Fri, Feb 5, 2021 at 2:48 PM Keith Oleson notifications@github.com wrote:

I've generated some survivability timeseries plots for the last 200 years of the pAD:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_1850pAD_TLAI.png

and the historical:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_hist_TLAI.png

The BL deciduous temperate and boreal are stable in the pAD, but then die off in the historical.

The C4 grass declines only slightly in the historical but I think the damage was done earlier in the AD/pAD.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774306523, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVE4OKVZJ53A2S5S44TS5RRTPANCNFSM4XFF5SSQ .

olyson commented 3 years ago

Diagnostics for the first 100 years of postAD of PPE compared to lmbirch postAD:

https://webext.cgd.ucar.edu/I1850/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD.471_500-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE_1850pAD.271_300/setsIndex.html

wwieder commented 3 years ago

It's hard to tell what ecosystem responses are leading vs. following, but Amazon also has big changes in 2m air temperature righ https://webext.cgd.ucar.edu/I1850/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD.471_500-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE_1850pAD.271_300/set6/set6_landf_Amazonia.pngt off the bat in the 'new' simulation. It would be interesting to see if this occurs in the BHSoff simulation too.

You don't see the temperature changes in NE US, but N costs https://webext.cgd.ucar.edu/I1850/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD.471_500-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE_1850pAD.271_300/set6/set6_cnFlx_Eastern_US.pngseem to spike in both regions https://webext.cgd.ucar.edu/I1850/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD.471_500-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE_1850pAD.271_300/set6/set6_cnFlx_Amazonia.png with the 'new' simulation (see lower right panel).

On Fri, Feb 5, 2021 at 6:33 PM Keith Oleson notifications@github.com wrote:

Diagnostics for the first 100 years of postAD of PPE compared to lmbirch postAD:

https://webext.cgd.ucar.edu/I1850/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE_1850pAD.471_500-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE_1850pAD.271_300/setsIndex.html

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774372702, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB5IWJAP3647UP53PWPBJ4LS5SL7LANCNFSM4XFF5SSQ .

olyson commented 3 years ago

I've put spinup plots (AD and pAD) for these recent simulations on the bottom of the landing page:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/index.html

The BHSOff (BHS turned off) simulation has the same problem as the leafbiomassesai simulation, where TOTVEGC drops rapidly in the first few decades of the pAD spinup. So this appears to be a problem with the tag rather than with BHS. I have another pAD simulation in progress with BHSOff but with the original nstem values.

olyson commented 3 years ago

BHSOff with original stem values has the same problem.

dlawrenncar commented 3 years ago

Thanks. In this case, I am wondering if this might be related to this bug fox for wood c:n ratios. The fact that this seems to be affecting just deadstemc (I think) and then propagating through a few other things makes me suspicious.

https://github.com/ESCOMP/CTSM/issues/443

Unfortunately, I'm not well versed enough in github to sort out quickly when this bug was integrated into the code (I'd like someone to teach me!), but I think it is possible that it wasn't included in the so-called LMBIRCH simulations.

I will say, though, that looking through that issue, it does look like Will did some test simulations with the fix and things looked fine. But I don't know if those tests were done with the actual tag or with code mods. Seems possible that there is a bug in the bug fix (?). Or, that there is something particular about coming out of AD spinup mode that intersects with this bug fix that was never really tested/evaluated.

On Sat, Feb 6, 2021 at 12:53 PM Keith Oleson notifications@github.com wrote:

BHSOff with original stem values has the same problem.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774532623, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVFKNL56GGUCABMXULLS5WM2RANCNFSM4XFF5SSQ .

olyson commented 3 years ago

It looks like that bug fix went into ctsm1.0.dev109. The lmbirch simulation used ctsm5.1.dev006 and the PPE simulations are using ctsm5.1.dev021. I checked the code itself and it looks like the changes are in both those code bases.

I ran a pAD with BHS off with ctsm5.1.dev021 (the base tag for the PPE tag), initialized from the BHSOffnstemorig run, and TOTVEGC was fine. So seems like we should be able to figure this out by looking at the differences between PPE.n02_ctsm5.1.dev021 and ctsm5.1.dev021 code, and the parameter files.

dlawrenncar commented 3 years ago

Ok. Not sure I understand for sure, but I think you are saying that it’s not the wood cn bug, then.

On Feb 6, 2021, at 3:35 PM, Keith Oleson notifications@github.com wrote:



It looks like that bug fix went into ctsm1.0.dev109. The lmbirch simulation used ctsm5.1.dev006 and the PPE simulations are using ctsm5.1.dev021. I checked the code itself and it looks like the changes are in both those code bases.

I ran a pAD with BHS off with ctsm5.1.dev021 (the base tag for the PPE tag), initialized from the BHSOffnstemorig run, and TOTVEGC was fine. So seems like we should be able to figure this out by looking at the differences between PPE.n02_ctsm5.1.dev021 and ctsm5.1.dev021 code, and the parameter files.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774553719, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVEDN57S3NJQJVJS6HDS5W72RANCNFSM4XFF5SSQ .

olyson commented 3 years ago

Right.

wwieder commented 3 years ago

I'll start with a quick solution to this problem, then explain more:

two parameters rf_cwdl2_bgc = 0 and rf_cwdl3_bgc = 0 Changing both of these to 0.5 may help?

I've been thinking a bit about the woodCN hypothesis. These changes should make much of an impact on the the veg side (which should actually be less N limited with the changes in the woodCN bug fix). Instead, they would potentially impact the decay of coarse woody debris (CWD), which now has a much higher C:N ratio (500, as parameterized instead of 200-250, which was happening previously). This would cause a big immobilization of N as CWD moves into litter pools that would be much higher coming out of AD mode (when CWD pools are 10x larger). The colder soil temperatures in BHS could exacerbate this issue, by reducing N availability, which would be consistent with the feature of the PPE tag of a big crash in mineral AN stocks.

One way to get around this would be to introduce a CO2 flux when CWD is transferred to litter pools. Right now we're supposedly representing the physical fragmentation of wood into litter, but this transfer has no associated CO2 flux (which seems pretty absurd). Putting 50% of this C back to the atmosphere would greatly reduce our N demand associated with decomposition.

wwieder commented 3 years ago

Now that I've tried piecing more of the conversation above together it sound like woodCN may not be the culprit?

One this I don't understand is why it looks like decid trees are OK in 1850, but dying in the historical period? Something else seems to be going on for C4 grasses, however? Maybe we need to talk on monday?

swensosc commented 3 years ago

I have two single point pAD simulations set up (I hope correctly). One for the eastern US, one for amazonia. The eastern US is mainly pft 7, the amazonia has a small amount of pft 14 (none of the gridcells in the amazon have a large percentage of pft 14). My eastern US died due to wallclock, so I'm rerunning, but here are plots for amazonia ( focusing on pft 14). The first is 400 years, the second zooms in on the first 10. [image: amazonia_full.png] [image: amazonia_10yr.png]

On Sat, Feb 6, 2021 at 5:35 PM will wieder notifications@github.com wrote:

Now that I've tried piecing more of the conversation above together it sound like woodCN may not be the culprit?

One this I don't understand is why it looks like decid trees are OK in 1850, but dying in the historical period https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL_hist_TLAI.png? Something else seems to be going on for C4 grasses, however? Maybe we need to talk on monday?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774566605, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGRN57C2HHDS4JPX2YW62NTS5XN3VANCNFSM4XFF5SSQ .

wwieder commented 3 years ago

Hi @swensosc look like these didn't load correctly into github. maybe just email them?

dlawrenncar commented 3 years ago

@wwieder Just to reiterate, we have determined that nothing is right in 1850. Vegetation biomass is order magnitude low for all PFTs. The problem arises after AD spinup over the first 40 years of post AD spinup. And Keith has isolated to differences between two tags, both of which I believe have the Wood CN bug in it, so it is something in those tags in between. Need to start exploring the differences.

olyson commented 3 years ago

Ok, I've been comparing code. I see one thing that is incorrect in the PPE.n02_ctsm5.1.dev021. In CNFireLi2021Mod:

       rootc_col(c) = rootc_col(c) + (frootc(p) + frootc_storage(p) + &
               frootc_xfer(p) + deadcrootc(p) * 10._r8 +       &
               deadcrootc_storage(p) + deadcrootc_xfer(p) +    &
               livecrootc(p)+livecrootc_storage(p) +           &
               livecrootc_xfer(p))*patch%wtcol(p)

This should be (or at least it was in previous tags):

       if (spinup_state == 2) then
          rootc_col(c) = rootc_col(c) + (frootc(p) + frootc_storage(p) + &
               frootc_xfer(p) + deadcrootc(p) * 10._r8 +       &
               deadcrootc_storage(p) + deadcrootc_xfer(p) +    &
               livecrootc(p)+livecrootc_storage(p) +           &
               livecrootc_xfer(p))*patch%wtcol(p)
       else
          rootc_col(c) = rootc_col(c) + (frootc(p) + frootc_storage(p) + &
               frootc_xfer(p) + deadcrootc(p) +                &
               deadcrootc_storage(p) + deadcrootc_xfer(p) +    &
               livecrootc(p)+livecrootc_storage(p) +           &
               livecrootc_xfer(p))*patch%wtcol(p)
       endif

The ctsm5.1.dev021 tag has:

       rootc_col(c) = rootc_col(c) + (frootc(p) + frootc_storage(p) + &
               frootc_xfer(p) + deadcrootc(p) * spinup_factor_deadwood +       &
               deadcrootc_storage(p) + deadcrootc_xfer(p) +    &
               livecrootc(p)+livecrootc_storage(p) +           &
               livecrootc_xfer(p))*patch%wtcol(p)

Presumably, spinup_factor_deadwood is being set to 1.0 in pAD mode, which might explain why the ctsm5.1.dev021 run seems to behave correctly.

I'm running a test now with the corrected code. Hopefully that's the only problem.

olyson commented 3 years ago

That changed answers, but didn't fix the decline in TOTVEGC.

olyson commented 3 years ago

There are similar problems with CNGapMortalityMod and CNVegStrucUpdate. I've fixed these and the results are looking promising (no decline in TOTVEGC). I'll continue and post a new pAD spinup plot.

dlawrenncar commented 3 years ago

Good work Keith. Perhaps these errors came in during the aborted attempt to change the factor of 10 to a constant and the subsequent reversion wasn't complete. Or, maybe they came in incorrectly with the updated fire code. Once this is sorted out, probably need to think through how to catch this type of error ... if it's possible.

On Sun, Feb 7, 2021 at 12:00 PM Keith Oleson notifications@github.com wrote:

There are similar problems with CNGapMortalityMod and CNVegStrucUpdate. I've fixed these and the results are looking promising (no decline in TOTVEGC). I'll continue and post a new pAD spinup plot.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-774734044, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVHYFTMHMGNQQS2NBETS53PMFANCNFSM4XFF5SSQ .

olyson commented 3 years ago

I've posted a plot of the new spinup, looks pretty normal:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/spinup_plots/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE3_1850pAD_Spinup-0.png

and survivability timeseries for the first 200 years:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL3_1850pAD_TLAI.png

olyson commented 3 years ago

3.5% disequilibrium:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/spinup_plots/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE3_1850pAD_Spinup-0.png

olyson commented 3 years ago

Started the historical this morning.

ekluzek commented 3 years ago

@olyson latest case: clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE3_hist has changes to the following files:

modified:   biogeochem/CNFireLi2021Mod.F90
modified:   biogeochem/CNGapMortalityMod.F90
modified:   biogeochem/CNVegStructUpdateMod.F90
modified:   biogeophys/CanopyFluxesMod.F90

The changes in CanopyFluxes are just to remove the iWUE changes. In CNVegStructUpdateMod.F90 they also include setting biomass to the hex 0_r8 rather than double precision 0._r8. We were concerned about what it was doing before.

So to make sure I don't change answers I'll replicate one of @olyson cases and bring in his changes and then make sure anything I do on top of them don't change answers for the base case. I shouldn't have to backout the iWUE changes for example, but better show that it doesn't change the case. In the past I just do a one month test for this sort of thing, which I think should be sufficient here as well.

olyson commented 3 years ago

FYI, the code I'm using is PPE.n02_ctsm5.1.021. We didn't remove the iWUE changes, they just weren't in there yet. And the 0_r8 was in the code previously, we just moved that section of code up out of the woody if statement. I agree that it should be 0._r8, which I think you've already done in the latest tag. Also, note that CNVegSTructUpdateMod also contains mods to htop for spinup.

olyson commented 3 years ago

The new historical (clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE3_hist) is done. The diagnostics landing page is here:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/index.html

Standard diagnostics of the new run compared to the lmbirch tag are here:

https://webext.cgd.ucar.edu/I20TR/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE3_hist/lnd/clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE3_hist.1995_2014-clm51_lmbirchwkattgen01ctsm51d006_2deg_GSWP3V1_nocrop_PPE2_hist.1995_2014/setsIndex.html

BL deciduous trees and C4 grasses appear to be mostly healthy at present day, although BL deciduous tree survivability did decline a bit during the historical:

https://www.cgd.ucar.edu/tss/clm/diagnostics/clm5.1_dev/pft_level_plots/plot_checkfordead_timeseries_PPEFINAL3_hist_TLAI.png

ILAMB results are in progress.

Here's a table of survivability at the end of the historical for the new run (first row) and lmbirch (last row):

PPEFINAL3 97.104 93.279 98.523 98.485 99.186 94.735 75.171 87.047 96.990 64.218 88.092 72.255 84.214 68.316 LMBIRCH 96.969 92.945 98.254 98.014 99.181 94.628 78.761 87.878 99.578 63.254 90.782 78.014 83.333 73.441

dlawrenncar commented 3 years ago

Thanks. I've been looking through plots and I don't see anything that is surprising. It does look like 2m air temperature is somewhat cooler across a lot of the globe, especially the high latitudes. As far as I am concerned, though, everything looks fine and this is ready to be passed over to Forrest and Nate for a new cluster analysis.

wwieder commented 3 years ago

i'm looking forward to seeing the ilamb results. I'm surprised how productive these new simulations are, I think global GPP is 15-20% higher now than in CLM5.

Are the PHS parameter changes likely responsible for the decreased latent heat flux? And now the model is colder... is this just because of BHS?

On Fri, Feb 12, 2021 at 12:48 PM David Lawrence notifications@github.com wrote:

Thanks. I've been looking through plots and I don't see anything that is surprising. It does look like 2m air temperature is somewhat cooler across a lot of the globe, especially the high latitudes. As far as I am concerned, though, everything looks fine and this is ready to be passed over to Forrest and Nate for a new cluster analysis.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-778415829, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB5IWJHZCCQECVFMKM5EUG3S6WA2LANCNFSM4XFF5SSQ .

olyson commented 3 years ago

ILAMB results are now posted here:

https://webext.cgd.ucar.edu/I20TR/clm50_ctsm10d089_2deg_GSWP3V1_luna3_hist/lnd/_build_CTSM5.1G/index.html

olyson commented 3 years ago

Erik, I think the changes I made to CNFireLi2021Mod would also need to be applied to CNFireLi2016Mod and CNFireLi2014Mod, for other versions of the model, right?

ekluzek commented 3 years ago

Yes, that's likely correct. I'll make sure that's handled correctly in the final version. It won't matter if you only using clm5_1 for the PPE branch right now. I'm working on making sure the PPE branch matches your simulation results. So I'll be able to make sure this is all straightened out in the end version.

dlawrenncar commented 3 years ago

LAI is now more biased, but that was pretty much true with the LMBIRCH simulation too. GPP is higher, but from global perspective is actually closer to the GBAF/FLUXNETMTE product. Don't think should place too much weight on that one way or other. Still looks good to me.

On Sat, Feb 13, 2021 at 2:57 PM Erik Kluzek notifications@github.com wrote:

Yes, that's likely correct. I'll make sure that's handled correctly in the final version. It won't matter if you only using clm5_1 for the PPE branch right now. I'm working on making sure the PPE branch matches your simulation results. So I'll be able to make sure this is all straightened out in the end version.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-778683679, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVD7FXVRQOVYW33MHQ3S63YTNANCNFSM4XFF5SSQ .

wwieder commented 3 years ago

Yeah, was it the LUNA bug fixes that increased LAI globally, or did another change come in made this shift from CLM5.0?

On Sat, Feb 13, 2021 at 5:48 PM David Lawrence notifications@github.com wrote:

LAI is now more biased, but that was pretty much true with the LMBIRCH simulation too. GPP is higher, but from global perspective is actually closer to the GBAF/FLUXNETMTE product. Don't think should place too much weight on that one way or other. Still looks good to me.

On Sat, Feb 13, 2021 at 2:57 PM Erik Kluzek notifications@github.com wrote:

Yes, that's likely correct. I'll make sure that's handled correctly in the final version. It won't matter if you only using clm5_1 for the PPE branch right now. I'm working on making sure the PPE branch matches your simulation results. So I'll be able to make sure this is all straightened out in the end version.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-778683679, or unsubscribe < https://github.com/notifications/unsubscribe-auth/AFABYVD7FXVRQOVYW33MHQ3S63YTNANCNFSM4XFF5SSQ

.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ESCOMP/CTSM/issues/1274#issuecomment-778700887, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB5IWJHMYSV2MKI32SO7C73S64MWHANCNFSM4XFF5SSQ .