search

NGEET / fates

repository for the Functionally Assembled Terrestrial Ecosystem Simulator (FATES)
Other
106 stars 92 forks source link

Single cohort BGC mode #679

Open rosiealice opened 4 years ago

rosiealice commented 4 years ago

In light of our desire to a) figure out if and how FATES could be speeded up (as per @dlawrenncar 's tests in https://github.com/ESCOMP/CTSM/issues/1076) we thought it would be good to add another reduced-complexity mode which has

This is therefore as close as one can get FATES to looking like CLM/ELM BGC modes, and therefore would provide a way of assessing the cost of just the FATES physiology implementation, as well as providing a platform for people who would still like to do BGC simulations but in the absence of demographics (e.g. to rapidly test hypotheses e.g. of controls on NPP?)

The primary differences between FATES and CLM should thus be the multi-layer canopy in FATES as well as the different allocation scheme.

The major scientific hurdle to implementation would be figuring out what to do about 'n' for the single cohort in question. One could follow the logic in here: https://docs.google.com/document/d/1n72jKxuxvF4pxTKCpZ1Ot_w1y_aK24LpO37gumnFStA/edit assuming a full canopy layer. I have to think about this a little more.

ckoven commented 4 years ago

Hi @rosiealice, I was trying to think this through some more this morning. I think a possible solution to this is along the lines the following:

  1. turn off both recruitment and mortality.
  2. represent the effect of mortality via a turnover time of stemwood carbon, which we already have in place as the branch loss term in the allocation code. so set that as something like a 2%/year or similar loss of stemwood carbon to represent mortality
  3. at each daily timestep, update the carbon gains from NPP and carbon losses from turnover to prognose the biomass pools of the cohort as per usual.
  4. after updating the biomass, solve for new values of both DBH and n to satisfy the two conservation equations for crown area and stem biomass at the level of the ecosystem: n_new crown_area(DBH_new) = patch area n_new stem_biomass_allometry(DBH_new) = n_old * stem_biomass_allometry(DBH_old)

I think this would be a bit different from the solution that Levine et al 2016 arrived at in their implementation of a big-leaf version of ED2, which they describe in appendix S2 of (https://www.pnas.org/content/113/3/793.short), because FATES has the crown area conservation from PPA that is not in ED2, which via the equations above should allow us to prognose diameter and thus height(DBH) instead of fixing height at some value as they did in ED2-big leaf.

ckoven commented 4 years ago

I guess, to make this even more of a closer analogue to the full model, we could calculate all of the mortality rates as per normal but just use them in the sense of a stemwood turnover rate in step 2 above rather than an actual mortality rate?

aswann commented 4 years ago

I'm very much in favor of having this mode of FATES. I currently use CLM-BGC mode to have prognostic leaves & biomass without having to deal with changing plant distributions.

This approach sounds reasonable to me but good to explicitly recognize that it has a built in assumption that roughness length will change when biomass changes. @ckoven pointed out that if there was a big disturbance event the left over plants would all of a sudden become smaller. This would impact roughness which will impact turbulent fluxes with the atmosphere.

I think the ED-Big leaf version from the Levine et al 2016 paper had a different goal - which was to show how the cohort-based approach differed from other DGVM approaches, rather than from a static spatial distribution of plants.

dlawrenncar commented 4 years ago

Should note that this is effectively what happens in big-leaf model as well. Disturbance (fire, logging, etc), just removes some carbon and vegetation height is adjusted downward, so roughness is similarly affected, even if this is unrealistic way to handle disturbance.

On Wed, Aug 5, 2020 at 3:24 PM Abby Swann notifications@github.com wrote:

I'm very much in favor of having this mode of FATES. I currently use CLM-BGC mode to have prognostic leaves & biomass without having to deal with changing plant distributions.

This approach sounds reasonable to me but good to explicitly recognize that it has a built in assumption that roughness length will change when biomass changes. @ckoven https://github.com/ckoven pointed out that if there was a big disturbance event the left over plants would all of a sudden become smaller. This would impact roughness which will impact turbulent fluxes with the atmosphere.

I think the ED-Big leaf version from the Levine et al 2016 paper had a different goal - which was to show how the cohort-based approach differed from other DGVM approaches, rather than from a static spatial distribution of plants.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NGEET/fates/issues/679#issuecomment-669519726, or unsubscribe https://github.com/notifications/unsubscribe-auth/AFABYVDO6YESFHZKMFV73OLR7HE2JANCNFSM4PVHZN2A .

ckoven commented 4 years ago

Yeah I guess the slightly weird dynamic of the scheme I'm suggesting above is that, when something causing elevated mortality rates and/or low growth rates were to occur, the number density of plants would tend to actually increase, as I'm pretty sure that the solution to the two allometry equations (at least for the parameter values that I would think we'd use) would generally result in a larger number of smaller-statured trees to fill the available canopy area. I guess that's kind of how the real world works in a long-term steady-state sense, whereby high disturbance rates mean lots of small canopy trees and low disturbance rates mean fewer large canopy trees, but it does seem a bit odd that mortality would instantaneously increase the population size. But maybe that's just a fundamental artifact of a big-leaf view of the world.

We'd probably also have to use higher values of the leaf biomass allometry parameters to compensate for having only one canopy layer.

ckoven commented 4 years ago

I tried to work through the math of how to solve for n and DBH, given the dual constraints on crown area and biomass conservation. There is an analytic solution for the special case where both biomass and crown area allometries are power laws. Algebra here: https://github.com/ckoven/assorted_notebooks/blob/master/fates_bigleaf_equations.ipynb

I tried to work through the case with Chave et al allometry and the Michaelis-Menten height allometry from Martinez-Cano (which is needed since Chave has a height term in the biomass allometry equation), but arrived at a set of equations that didn't seem to have an obvious analytic solution. So probably we could use the equations above for the special case of power-law allometries but then put in a numerical solution for the more general case of arbitrary allometries.

wwieder commented 2 years ago

I'm not sure where to drop this question, but I wondered about the status of other reduced complexity modes from FATES? FATES-sp is supported, but what's the status and timeline of getting features of other modes running?

glemieux commented 2 years ago

We have a FATES board dedicated to this, although its been a while since it was updated: https://github.com/NGEET/fates/projects/5

jkshuman commented 2 years ago

This list of various complexity modes are functional and undergoing calibration and testing for various sites, region, and global applications across a range of projects.

Modes are set at case creation using namelist options as summarized below and here FATES runtime modes

ckoven commented 2 years ago

I gave up on this proposed configuration a while ago, because I'm not sure it's really needed or what the real use-case would be (other than maybe a one-off study). While the current fixed-biogeography/nocomp is not exactly parallel in assumptions to big-leaf BGC mode, it is close enough to be comparable. I think our current portfolio of configurations might be sufficient for the long run?

rosiealice commented 2 years ago

Thanks all.

Indeed the titular mode of this thread is the one @ckoven didn't end up pursuing, but all the modes @jkshuman describes are functional, plus twommore from earlier, so in total we have:

-FATES- SP

-fixed biogeography (without competition) -fixed biogeography (with competition) -No competition by itself *

-Full FATES

Plus: -Static stand structure (initialised from inventory) -Prescribed physiology (just runs stuff downstream of NPP and mortality)

These latter two are presently mostly run for site level applications.

Many of these have publications associated with them. Rutuja and Jessie have papers on the latter two, and Marius Lambert's papers on hydraulics and frost damage will utilise SP mode and no comp mode.

At the moment I am focusing on the bigger biases in SP mode and @JessicaNeedham is looking further down the chain (with iteration i between).

jkshuman commented 1 year ago

given comment from @ckoven "

I gave up on this proposed configuration a while ago, because I'm not sure it's really needed or what the real use-case would be (other than maybe a one-off study). While the current fixed-biogeography/nocomp is not exactly parallel in assumptions to big-leaf BGC mode, it is close enough to be comparable. I think our current portfolio of configurations might be sufficient for the long run?

should this issue be closed?