Co-limitation of photosynthesis #472
Comments
|
A very simple fix would be to increase both the theta and beta (Collatz et al., 1991) parameter values from 0.95 and 0.98 to 0.999 for both of them. This preserves the smoothing which can benefit the current solver and is a very simple fix. A value of 0.999 for both parameters would give a maximum reduction in gross assimilation of 3.1 % compared to selecting the minimum rate. The current parameter values give a maximum reduction of ~25%, so if we go ahead with this we can expect quite a jump in GPP, a rough guess would be ~10% increase but it's difficult to say. Once we've progressed the semi-analytical solver to a point where it's faster than the current solver in FATES we'll have a greater overhaul of the photosynthesis code and we can decide to keep a small amount of smoothing or delete it completely. A small amount of smoothing is probably realistic, but it comes at the cost of two quadratic solves for every calculation of photosynthesis. |
|
I concur with Anthony’s suggestion.
From: walkeranthonyp <notifications@github.com>
Sent: Tuesday, February 12, 2019 3:44 PM
To: NGEET/fates <fates@noreply.github.com>
Cc: Subscribed <subscribed@noreply.github.com>
Subject: Re: [NGEET/fates] Co-limitation of photosynthesis (#472)
A very simple fix would be to increase both the theta and beta (Collatz et al., 1991) parameter values from 0.95 and 0.98 to 0.999 for both of them. This preserves the smoothing which can benefit the current solver and is a very simple fix. A value of 0.999 for both parameters would give a maximum reduction in gross assimilation of 3.1 % compared to selecting the minimum rate. The current parameter values give a maximum reduction of ~25%, so if we go ahead with this we can expect quite a jump in GPP, a rough guess would be ~10% increase but it's difficult to say.
Once we've progressed the semi-analytical solver to a point where it's faster than the current solver in FATES we'll have a greater overhaul of the photosynthesis code and we can decide to keep a small amount of smoothing or delete it completely. A small amount of smoothing is probably realistic, but it comes at the cost of two quadratic solves for every calculation of photosynthesis.
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub<#472 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ANTfrc2OE0CJQQPloy5KYmPy4zu3atmrks5vMyeEgaJpZM4a1IJm>.
|
|
So, it looks like these numbers are hard wired in here: and used here-ish: fates/biogeophys/FatesPlantRespPhotosynthMod.F90 Line 1052 in 08bb140 There appear to be three relevant thetas, theta_cj, which is 0.8(!) for C3 and 0.98 for C4, and theta_ip (0.95) for adding in the TPU colimitation. I'm not sure what 'beta' maps on to in this instance @walkeranthonyp. Is that the theta_ip? Do we think these should be in the parameter file @rgknox and @ckoven? My feeling is that that might make people think they need to go out and measure these things, which don't really exist. On the other hand, having them in the file facilitates the type of sensitivity analysis that Anthony just did with MAAT which would have been impossible with them hard-wired in! I'll set off a simulation with all three numbers at 0.999 now and see what happens... |
|
0.8! JULES has a 0.9 theta (theata_cj) and a 0.83 Beta (theta_ip) and had a 38% reduction in leaf level assimilation due to the smoothing artifact. I’m guessing the FATES 0.8 theta_cj will have a similar impact. Brace yourself for a big GPP increase.
Alistair
From: Rosie Fisher <notifications@github.com>
Sent: Thursday, February 14, 2019 10:33 AM
To: NGEET/fates <fates@noreply.github.com>
Cc: Rogers, Alistair <arogers@bnl.gov>; Comment <comment@noreply.github.com>
Subject: Re: [NGEET/fates] Co-limitation of photosynthesis (#472)
So, it looks like these numbers are hard wired in here:
https://github.com/NGEET/fates/blob/08bb1406bf4de99252cc14faa08a7d2eb651f140/biogeophys/FatesPlantRespPhotosynthMod.F90#L929
and used here:
https://github.com/NGEET/fates/blob/08bb1406bf4de99252cc14faa08a7d2eb651f140/biogeophys/FatesPlantRespPhotosynthMod.F90#L1052
There appear to be three relevant thetas, theta_cj, which is 0.8(!) for C3 and 0.98 for C4, and theta_ip (0.95) for adding in the TPU colimitation. I'm not sure what 'beta' maps on to in this instance @walkeranthonyp<https://github.com/walkeranthonyp>. Is that the theta_ip?
Do we think these should be in the parameter file @rgknox<https://github.com/rgknox> and @ckoven<https://github.com/ckoven>? My feeling is that that might make people think they need to go out and measure these things, which don't really exist. On the other hand, having them in the file facilitates the type of sensitivity analysis that Anthony just did with MAAT which would have been impossible with them hard-wired in!
I'll set off a simulation with all three numbers at 0.999 now and see what happens...
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub<#472 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ANTfrVhFXjPxZgPXrPUdrVkoHHDBWNjcks5vNYGpgaJpZM4a1IJm>.
|
|
I set off the run. Beginning to suspect this might be behind the CLM5
Amazon problem too...
…On Thu, Feb 14, 2019, 7:08 PM alistairrogers ***@***.*** wrote:
0.8! JULES has a 0.9 theta (theata_cj) and a 0.83 Beta (theta_ip) and had
a 38% reduction in leaf level assimilation due to the smoothing artifact.
I’m guessing the FATES 0.8 theta_cj will have a similar impact. Brace
yourself for a big GPP increase.
Alistair
From: Rosie Fisher ***@***.***>
Sent: Thursday, February 14, 2019 10:33 AM
To: NGEET/fates ***@***.***>
Cc: Rogers, Alistair ***@***.***>; Comment <
***@***.***>
Subject: Re: [NGEET/fates] Co-limitation of photosynthesis (#472)
So, it looks like these numbers are hard wired in here:
https://github.com/NGEET/fates/blob/08bb1406bf4de99252cc14faa08a7d2eb651f140/biogeophys/FatesPlantRespPhotosynthMod.F90#L929
and used here:
https://github.com/NGEET/fates/blob/08bb1406bf4de99252cc14faa08a7d2eb651f140/biogeophys/FatesPlantRespPhotosynthMod.F90#L1052
There appear to be three relevant thetas, theta_cj, which is 0.8(!) for C3
and 0.98 for C4, and theta_ip (0.95) for adding in the TPU colimitation.
I'm not sure what 'beta' maps on to in this instance @walkeranthonyp<
https://github.com/walkeranthonyp>. Is that the theta_ip?
Do we think these should be in the parameter file @rgknox<
https://github.com/rgknox> and @ckoven<https://github.com/ckoven>? My
feeling is that that might make people think they need to go out and
measure these things, which don't really exist. On the other hand, having
them in the file facilitates the type of sensitivity analysis that Anthony
just did with MAAT which would have been impossible with them hard-wired
in!
I'll set off a simulation with all three numbers at 0.999 now and see what
happens...
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub<
#472 (comment)>, or
mute the thread<
https://github.com/notifications/unsubscribe-auth/ANTfrVhFXjPxZgPXrPUdrVkoHHDBWNjcks5vNYGpgaJpZM4a1IJm>.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#472 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AMWsQy9ZCEMF9atcwFdH_SYNP-RHfnyYks5vNaYngaJpZM4a1IJm>
.
|
|
@rosiealice , I think the indexing on those is reversed. ! photosynthetic pathway: 0. = c4, 1. = c3 For c3 we are using theta_cj = 0.98, can you confirm? real(r8),parameter,dimension(0:1) :: theta_cj = [0.80_r8,0.98_r8] |
|
Ah right. Yes, they are indeed the other way around.
Le jeu. 14 févr. 2019 à 19:50, Ryan Knox <notifications@github.com> a
écrit :
… @rosiealice <https://github.com/rosiealice> , I think the indexing on
those is reversed.
! photosynthetic pathway: 0. = c4, 1. = c3
For c3 we are using 0.98, can you confirm?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#472 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AMWsQ3xN42uPI3XDEJqZ-D7Rs3fGLKEFks5vNa_8gaJpZM4a1IJm>
.
--
------------------------------------------------------------
Dr Rosie A. Fisher
Staff Scientist
Terrestrial Sciences Section
Climate and Global Dynamics
National Center for Atmospheric Research
1850 Table Mesa Drive
Boulder, Colorado, 80305, USA
and
Visitor @ C.E.R.F.A.C.S
Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique
42 Avenue Gaspard Coriolis 31057, Toulouse, France
http://www.cgd.ucar.edu/staff/rfisher/
|
|
So, this does seem to have a large effect. This is what happens after 6 years and 9 months (it's nearly bedtime here) with our default evergreen broadleaf notionally tropical trees. These runs aren't equilibrated in any sense yet, but you can see that the Amazon is looking quite a lot healthier.
|
|
Cool. It’d also be interesting to see the effect of ditching smoothing on the understory survival – I suspect you might not kill off your young cohorts quiet as aggressively.
A
From: Rosie Fisher <notifications@github.com>
Sent: Thursday, February 14, 2019 4:53 PM
To: NGEET/fates <fates@noreply.github.com>
Cc: Rogers, Alistair <arogers@bnl.gov>; Comment <comment@noreply.github.com>
Subject: Re: [NGEET/fates] Co-limitation of photosynthesis (#472)
So, this does seem to have a large effect. This is what happens after 6 years and 9 months (it's nearly bedtime here) with our default evergreen broadleaf notionally tropical trees. These runs aren't equilibrated in any sense yet, but you can see that the Amazon is looking quite a lot healthier.
[lai]<https://user-images.githubusercontent.com/12954691/52819716-516a3800-30aa-11e9-8953-08d8ce893dd6.png>
[gpp]<https://user-images.githubusercontent.com/12954691/52819719-53cc9200-30aa-11e9-8dd9-7f35fe5b799f.png>
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub<#472 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ANTfrXXNm_ZKn7Alj9uwu0PbX0isVwpFks5vNdqwgaJpZM4a1IJm>.
|
|
Very cool! ... and mad that this has such an effect. Nice to see the Amazon looking healthier. |
|
@rosiealice I forget which way round theta and beta are in the original paper. I wouldn't worry about it though if we just set them both to 0.999. I'm not sure what to do about C4, it will obviously have an influence, probably too much. I'm less familiar with the theory tho. The C4 a-ci curves I have seen have a pretty strong inflection point but I don't know about the a-par curves. @alistairrogers thoughts? Lastly I would caution against putting these in the parameter file @rosiealice, @rgknox, @ckoven. This would just encourage people to vary them when really the reason we're leaving them in the model is for numerical expediency with a longer view to chucking them out, unless we have any good data. @alistairrogers looking at you for them high-res aci curves ;) |
|
unfortunately Collatz et al has theta and beta listed one way in the text and the opposite in the table :-/
C4 photosynthesis should be CO2 saturated at current CO2 concentration so for ACi it shouldn’t make a difference - I hope! I’ll have a think about AQ curves.
A
Alistair’s Mobile
…________________________________
From: walkeranthonyp <notifications@github.com>
Sent: Thursday, February 14, 2019 6:25 PM
To: NGEET/fates
Cc: Rogers, Alistair; Mention
Subject: Re: [NGEET/fates] Co-limitation of photosynthesis (#472)
@rosiealice<https://github.com/rosiealice> I forget which way round theta and beta are in the original paper. I wouldn't worry about it though if we just set them both to 0.999.
I'm not sure what to do about C4, it will obviously have an influence, probably too much. I'm less familiar with the theory tho. The C4 a-ci curves I have seen have a pretty strong inflection point but I don't know about the a-par curves. @alistairrogers<https://github.com/alistairrogers> thoughts?
Lastly I would caution against putting these in the parameter file @rosiealice<https://github.com/rosiealice>, @rgknox<https://github.com/rgknox>, @ckoven<https://github.com/ckoven>. This would just encourage people to vary them when really the reason we're leaving them in the model is for numerical expediency with a longer view to chucking them out, unless we have any good data. @alistairrogers<https://github.com/alistairrogers> looking at you for them high-res aci curves ;)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub<#472 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ANTfrU529efe5918WaPR9wi4Upa0pklRks5vNfBNgaJpZM4a1IJm>.
|
|
There weren't actually any C4's in my simulation... Could test it though, just to check. We have been, both in CLM and FATES, a bit concerned about over-productivity of the C4 algorithm in general, but haven't had any time or expertise to go through it in enough detail... |
|
Here's some quick figures (apologies for their sketchiness) from the 10th year (I just did 10 years, will run them out longer now). Indeed, as suspected/predicted from Anthony's results, the understory does much better here...
|
|
My guess would be quantum yield. The quantum yield for ET in C3 plants in CLM4.5 is close to the theoretical maximum i.e. QY for ET on an absorbed light basis =(1-f)/2 where f = 0.15 in CLM. In CLM f is markedly lower than most other TBMs resulting in a high QY. I’m guessing this is similar for C4s which would definitely juke your photosynthesis.
A
From: Rosie Fisher <notifications@github.com>
Sent: Friday, February 15, 2019 4:13 AM
To: NGEET/fates <fates@noreply.github.com>
Cc: Rogers, Alistair <arogers@bnl.gov>; Mention <mention@noreply.github.com>
Subject: Re: [NGEET/fates] Co-limitation of photosynthesis (#472)
There weren't actually any C4's in my simulation... Could test it though, just to check. We have been, both in CLM and FATES, a bit concerned about over-productivity of the C4 algorithm in general, but haven't had any time or expertise to go through it in enough detail...
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub<#472 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ANTfrVobgULB65Ry3be4UMCMmnyWDkr5ks5vNnoagaJpZM4a1IJm>.
|
|
Thanks @alistairrogers. Guess I should have asked you this (about C4's) a while ago! |
|
@alistairrogers I have a set of C4 simulations, and would like to update them according to this thread. Will follow up with you on the details this week. As @rosiealice indicated the C4 is highly competitive. |
|
Should we start a new thread for C4 issues wrt Alistair's point?
…On Mon, Feb 18, 2019, 4:55 PM jkshuman ***@***.*** wrote:
@alistairrogers <https://github.com/alistairrogers> I have a set of C4
simulations, and would like to update them according to this thread. Will
follow up with you on the details this week. As @rosiealice
<https://github.com/rosiealice> indicated the C4 is highly competitive.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#472 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AMWsQwOABkNRf6LNn9u59x3nHZmeEJPjks5vOsz1gaJpZM4a1IJm>
.
|
|
Good idea @rosiealice I will create a separate C4 issue. There may already be a C4 issue floating around, so will confirm that as well. |
|
Some better equilibrated figures (57 years and counting) from the changes to the smoothing parameter global 4x5 runs. Trivially, GPP and LAI are higher. They don't completely eradicate the low central/Eastern Amazon LAI issue though. My first thought was that this was to do with water limitation, but BTRAN is not associated with the low LAI values at all, so maybe it's low humidity or high temperature that's driving that. @ckoven I've been trying to get the vertical profile of GPP. There's a net assimilation output that's stratified like that, but it's complicated by the issue of partially filled layers and it's annual counting mandate, so I'm going to try and add GPP by leaf layer directly. I'll try and make some figures that more usefully highlight the differences, rather than y'all having to squint.
|
|
Finally posting figures for this issue. These are sets of 2 PFT (Evergreen Tropical Tree and C3 grass, and Evergreen Tropical Tree and C4 grass) runs with and without theta set to 0.99 in tag: sci.1.21.0_api.7.0.0 (from mid January) Results for annual average for understory biomass, GPP and LAI for years 52 and 53 of simulation. C3 is on the top of the panel, C4 bottom. Theta=0.99 on the left and default on the right.
|
|
This could technically be closed now, with PR #482, right? |
|
I think so, nice to tidy up the "open issues" list when possible |
@walkeranthonyp has a very compelling argument from MAAT that co-limitation of photosynthesis should either be removed, or the parameters controlling it should be increased. Adding this as a placeholder to remember to do this. Can you add some detail to what you would advocate Anthony?
The text was updated successfully, but these errors were encountered: