Reducing respiration when carbon stores are low as an understory survival strategy

[ckoven]

We've been struggling for a while about what to do to help out understory plants. While we have at times found parameter sets that allow understory plants to survive for a little while, in most cases they are stuck with a negative NPP, which leads them to crash out their carbon stores and die quickly via termination mortality at the entire cohort level. There are a few issues with this: (1) it makes it really hard to achieve an understory at all, and (2) when we do it is super sensitive to parameters, (3) when things die, they tend to die all at once rather than spread out over a while, (4) there are some weird numerical consequences of that (e.g. this comment https://github.com/NGEET/fates/issues/296#issuecomment-345026550 ).

We talked a bit about a strategy to solve this, whereby, when plants are on the part of the carbon storage curve where they are dying from carbon starvation, they also reduce their respiration rates. The theoretical justification for this is that the reason they are dying is that they aren't able to maintain themselves properly or invest in defense compounds, so this should be reflected in a reduced rate of maintenance respiration. @xuchongang points out that there is empirical support for this too in Sanna Sevanto's paper ( http://onlinelibrary.wiley.com/doi/10.1111/pce.12141/abstract ) where the respiration rates drop roughly proportionally to carbon stores when the carbon stores are low. But right now, the way it works in FATES is that the plants suffer the cost of low carbon stores (that they die) but not the benefit (that they are saving their stores by deferring their maintenance). Also if we can shift away from termination mortality being a dominant thing without having to resort to deficit accounting (as discussed in #296 ) that would also be a good thing.

A wrote up a preliminary version of this in commit 987144a. Basically it makes a curve that has a single parameter that can be adjusted from a shape that assumes a linear reduction in respiration as stores dip into the mortality zone to one that assumes no reduction in respiration until reserves are completely gone (which is the current behavior).

I played around with it a bit and the approach basically works. If I plot a timeseries of understory mortality rates (where 1 = 100% mortality per yer) for a few cases, given the current default parameters at the BCI testbed site, it looks like this:

Black line is the default case (so mortality is mainly termination mortality due to negative carbon stores; it only looks smooth because I've applied a 10-year smoother, in practice it is really really spikey). The colored lines are with different values of the parameter, and in each of those cases the mortality is now all carbon starvation mortality rather than termination mortality.

The result of this is that the understory is now much more occupied. You can see this by plotting the number of canopy layers within a given patch (y-axis) as a function of the patch age (x-axis), with the same line colors as before. The default case barely ever has a second canopy layer because mortality rates are so high, whereas the cases with the reduced respiration have up to about a 50% full understory in the medium age patches.

Note that at the canopy level, things like GPP and Ra are not very strongly affected; they both go up when we add this function, and canopy-level CUE only changes by 2%. Biomass goes up too, but not very much.

Anyway, I'm starting this as an issue now to solicit opinions about (a) whether people think this is a reasonable approach or not, and (b) if so, whether the parameter that controls the degree of reduction in respiration ought to be thought of as a plant trait or a global parameter (fig. 5 in Sanna's paper suggests possibly a plant trait?). Thanks-

[rosiealice]

Did you ever decide to go with this fix? It's encouraging that it doesn't change ecosystem level CUE all that much.

On Nov 29, 2017 12:36 PM, "Charlie Koven" notifications@github.com wrote:

We've been struggling for a while about what to do to help out understory plants. While we have at times found parameter sets that allow understory plants to survive for a little while, in most cases they are stuck with a negative NPP, which leads them to crash out their carbon stores and die quickly via termination mortality at the entire cohort level. There are a few issues with this: (1) it makes it really hard to achieve an understory at all, and (2) when we do it is super sensitive to parameters, (3) when things die, they tend to die all at once rather than spread out over a while, (4) there are some weird numerical consequences of that (e.g. this comment #296 (comment) https://github.com/NGEET/fates/issues/296#issuecomment-345026550 ).

We talked a bit about a strategy to solve this, whereby, when plants are on the part of the carbon storage curve where they are dying from carbon starvation, they also reduce their respiration rates. The theoretical justification for this is that the reason they are dying is that they aren't able to maintain themselves properly or invest in defense compounds, so this should be reflected in a reduced rate of maintenance respiration. @xuchongang https://github.com/xuchongang points out that there is empirical support for this too in Sanna Sevanto's paper ( http://onlinelibrary.wiley.com/doi/10.1111/pce.12141/abstract ) where the respiration rates drop roughly proportionally to carbon stores when the carbon stores are low. But right now, the way it works in FATES is that the plants suffer the cost of low carbon stores (that they die) but not the benefit (that they are saving their stores by deferring their maintenance). Also if we can shift away from termination mortality being a dominant thing without having to resort to deficit accounting (as discussed in #296 https://github.com/NGEET/fates/issues/296 ) that would also be a good thing.

A wrote up a preliminary version of this in commit 987144a https://github.com/NGEET/fates/commit/987144a231e27549356c0d7ce0f8c43548b2f807. Basically it makes a curve that has a single parameter that can be adjusted from a shape that assumes a linear reduction in respiration as stores dip into the mortality zone to one that assumes no reduction in respiration until reserves are completely gone (which is the current behavior).

I played around with it a bit and the approach basically works. If I plot a timeseries of understory mortality rates (where 1 = 100% mortality per yer) for a few cases, given the current default parameters at the BCI testbed site, it looks like this:

[image: screen shot 2017-11-29 at 10 37 36 am] https://user-images.githubusercontent.com/10852790/33394439-305dc740-d4f7-11e7-9360-d0163131d340.png

Black line is the default case (so mortality is mainly termination mortality due to negative carbon stores; it only looks smooth because I've applied a 10-year smoother, in practice it is really really spikey). The colored lines are with different values of the parameter, and in each of those cases the mortality is now all carbon starvation mortality rather than termination mortality.

The result of this is that the understory is now much more occupied. You can see this by plotting the number of canopy layers within a given patch (y-axis) as a function of the patch age (x-axis), with the same line colors as before. The default case barely ever has a second canopy layer because mortality rates are so high, whereas the cases with the reduced respiration have up to about a 50% full understory in the medium age patches.

[image: screen shot 2017-11-29 at 11 24 53 am] https://user-images.githubusercontent.com/10852790/33394679-f3a9007a-d4f7-11e7-82e0-8fbac31ed5af.png

Note that at the canopy level, things like GPP and Ra are not very strongly affected; they both go up when we add this function, and canopy-level CUE only changes by 2%. Biomass goes up too, but not very much.

Anyway, I'm starting this as an issue now to solicit opinions about (a) whether people think this is a reasonable approach or not, and (b) if so, whether the parameter that controls the degree of reduction in respiration ought to be thought of as a plant trait or a global parameter (fig. 5 in Sanna's paper suggests possibly a plant trait?). Thanks-

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/NGEET/fates/issues/308, or mute the thread https://github.com/notifications/unsubscribe-auth/AMWsQwgEMdS5eKKgmMfutqPSz-EU311nks5s7bI0gaJpZM4Qvgmc .

[ckoven]

@rosiealice and all,

So, yes, I've continued to play with this on my branch. I think it is a useful hypothesis to be able to encode in FATES, and also provides a solution to our understory survival problem. That said, I think we also want to have the ability to turn this mechanism off completely. So what I was thinking of doing is to add two new parameters: the first is what the parameter maintresp_reduction_parameter is doing in 987144a, which sets the curvature of the curve relating the reduction in respiration as a function of carbon storage. The second parameter would be the y-intercept of that curve; so a value of 1 would mean that respiration would still be at 100% no matter how low the carbon stores got (i.e. the behavior of the current master branch, so turning this behavior off completely), and a value of 0 would be the behavior of this branch currently (i.e. respiration completely ceases at the point where carbon stores reach zero, so that death under that point happens at the rate set by the current global fates_stress_mort parameter and there is no possibility for negative carbon and associated termination mortality).

While at it, we may want to remove the fixed assumption that carbon starvation mortality begins when the storage pool equals the leaf carbon pool and add a third new parameter (currently assumed to be 1.0) relating the threshold level of carbon starvation as a multiple of the leaf carbon pool, and then make the all these terms a function of that rather than the leaf carbon as they currently are.

Together these new parameters would logically be grouped with the fates_stress_mort and the fates_cushion parameters to define the rates of mortality and respiration as a function of carbon storage. I think some discussion on which of these ought to be PFT-level vs global parameters is warranted though. It seems like we could pull all of these out as PFT parameters and make understory resistance to carbon starvation a set of plant traits; alternately we could make them global and make all the plants compete given the same set of constraints under limited carbon. So I'd still be interested to know if people have opinions about that.

[ckoven]

solved by #341.