Incorperate leaf age demographics into the FATES model

[serbinsh]

Summary of Issue:

Incorporating leaf age demographics into the FATES model to enable leaf age /phenological control on physiology and RT

Expected behavior and actual behavior:

NA

Steps to reproduce the problem (should include create_newcase or create_test command along with any user_nl or xml changes):

NA

What is the changeset ID of the code, and the machine you are using:

NA

have you modified the code? If so, it must be committed and available for testing:

NA

Screen output or output files showing the error message and context:

NA

[serbinsh]

We would like to introduce a representation of leaf age cohorts/demographics into FATES in order to improve the modeling of C, water, and energy cycling and storage for tropical evergreen forests (TEF) . Enabling the ability to track leaf age will allow us to prescribe or model the annual changes in photosynthetic capacity and radiative transfer (among other properties) of the canopy and potentially better track satellite and EC tower observations of TEF seasonality, particularly during the dry season (where applicable).

[serbinsh]

Note from @rosiealice regarding potential issues that need to be addressed and though about regarding leaf age demographics:

"On the leaf age study, the FATES model doesn't track leaf age cohorts at the moment, and unless we implement a method for doing so then it'll be hard to implement those findings. I had this conversation in another email thread with Jin and Loren, which I can't recall if you were on. The concern is that implementing leaf age cohorts would basically increase the expense and memory needs of the model linearly with respect to the number of cohorts we chose, so, unless we though that was the primary dynamic we wanted to capture and we were OK with the cost, we'd likely have to have some other means of representing the impacts of, say, how mean leaf age varies through the growing season, or perhaps use LUNA to represent how the photosynthetic parameters of the whole canopy change with environmental conditions. Does ED2 have leaf age cohorts? If so I'd be interested in how they include those parsimoniously."

[rgknox]

Hi @serbinsh and @rosiealice. I'm a little unclear on what an "age cohort" is. A cohort in model terms, defines a group of plants that share similar functional traits, life-history, structure and biophysical/biochemical states. Of course there is variability within the cohort regarding these status' that we don't capture.

Are you saying that there should be a whole extra level to the ED scaling framework, where cohorts are further partitioned into groups with different leaf age-distributions tendencies? Or, are you saying that for any given cohort, there is a dynamic distribution of leaf ages that should be captured and tracked? Or, are you saying that the variability in leaf ages is connected to PFTs, and its not about changing the model structure, but more about incorporating these different PFTs into the model and allowing cohorts to manifest from the different available pfts?

[ckoven]

@rgknox I think what they mean are cohorts of leaves that sit within cohorts of plants (your option 2?)

An issue is that we still don't even break leaves down by cohorts of plants -- the code is still lumping all the cohorts of a given PFT in a patch together for leaf calculations. Disaggregating leaves by plant cohort has been on our list to do for a while but keeps getting pushed back due to software concerns etc. And doing so will necessarily already slow things down. So leaf age cohorts probably ought to be a second step beyond just disaggregating leaves by plant cohorts.

[rosiealice]

@serbinsh - can you & Jin make a science case for why age cohorts need to happen, as opposed to say, canopy scaling changes or using LUNA to predict vertical profiles of leaf properties? (given that age cohorts are by some large measure the most expensive of those options). You mentioned you were writing a manuscript about it, so presumably you already have the bones of an argument about why it is necessary?

[serbinsh]

@rosiealice @ckoven @rgknox

Yes, we can definitely explain and expand on the discussion a bit to clarify our through process and rationale. In general I think there is a bit of confusion revolving around terminology and what exactly we are getting at so I will try to better explain this as well.

@rosiealice I am not sure I fully understand your comments but in general the idea is that photosynthetic capacity, optical properties, traits, etc change with leaf age or development stage. In addition, for some canopies such as TEF that are leaf exchangers this drives the seasonality we see in GPP, etc. Moreover, where the majority of leaf turnover (replacing old leaves with young leaves) matters in terms of overall canopy photosynthesis; assuming all leaves in sun and shade have the same turnover results in an incomplete representation of seasonality and does not match observations, which show that more turnover happens in the upper canopy. So what we are looking at is a way to capture phenology and scale photo capacity (or other properties) into the canopy profile (e.g. using a Lloyd approach) based on, say, young mature and old leaf cohorts. Canopy-scale Vcmax, for example, is the aggregate of the Y, M, O cohorts by PFT and LAI and that is used in the fast flux calculations.

This is a simplification for now as I don't have time right now to prepare a full justification but we will work on that. Moreover, we aren't suggesting the tracking of individual leaves or cohorts per se but thinking about whether we can come up with a means for prescribing the fractions within a canopy through time. It may be more clear when I can share the manuscript.....

Perhaps LUNA could capture the essence (or even be used as the way to capture this effect) of this by predicting changes in leaves over time but I think we would need to explore that.

[rosiealice]

Cool. It would indeed be interesting to build, say, a test case in FATES where we could maybe look at alternative representations of the aging issues (I guess including options that might ultimately be expensive, to see whether they have fundamentally different dynamics)...

[serbinsh]

@xuchongang FYI - here is a discussion I created on this in 2016. I think we should set a coordinated development plan on this where we can utilize your implementation and then also build in what we are working on with MAAT that we expect to build into FATES over the next year or so. Tagging @walkeranthonyp @wujin1985 @alistairrogers here

[ckoven]

@serbinsh thanks for re-opening this thread. To me a really key question is what would be the differences between the simple approach in @xuchongang's #432 implementation — where the photosynthesis eqns are calculated using a mean vcmax and jmax across the leaf age cohorts — versus a more expensive solution where we calculate the photosynthesis eqns on each leaf age cohort and take the mean leaf-level fluxes across the age cohorts. This seems like a question that MAAT is ideally suited to answer. @walkeranthonyp and I discussed this last week as a possible way to approach reconciling this issue?

[xuchongang]

@serbinsh , thank you very much for sharing of the discussion. It is very useful. I agree with @ckoven to compare the two approaches for differences in MAAT but has the option of two approaches of implementation for computational cost purposes.

[serbinsh]

@ckoven @walkeranthonyp @xuchongang thanks for the quick replies. Indeed this is exactly the idea(s) we had/have with MAAT. The idea we laid out previously here and in calls is to guide the necessary implementation and complexity to best represent the intra- and inter-annual C, H2O leaf fluxes and the leaf and canopy scales. I think its valuable to have multiple hypotheses and implementation on how this can be done, i.e. more continuous leaf age classes and detailed leaf turnover distributions in the canopy, based on a demographic model, as developed in canopy MAAT, versus discrete means of the distributions by age. I would argue that it isn't just age, however, but the distribution of age classes and where they occur in the canopy the regulates fluxes. Moreover, given the direction we are going with Hydro and PARTEH I would say that having the ability to directly link leaf cohorts and multiple processes is valuable and we can compare detailed simulations to those implemented with less internal details by dynamically modify model internal complexity, photosynthesis, and scaling assumptions. It may be that depending on the question or system, it makes sense to increase/reduce the complexity.

Its again a question about how much trait/parameter variability we want to implicitly/explicitly capture in FATES. As with the MAAT simulations we have presented WRT ENSO, this too is important to understand the impact of age, age by species/PFT, seasonality, and vertical distribution on fluxes at the site, but then ultimately how this impacts global sims.

[serbinsh]

...also just to be clear, I am not advocating for one approach over another. Just explaining why we have been focused on a specific approach that we had plan to use with MAAT first to test and evaluate, based on the earlier work in GCB. I personally think we should have options, especially since we want to have predictive ability, yes, but also have a chance to ask questions and play with different assumptions/hypotheses in FATES to see what impacts they have on model simulations. I would say that this is a great example of how we can facilitate new approaches to modeling seasonality to explore the impacts on our ability to model sites/regions, etc. Thats what we are interested in going, but use MAAT first allows us to more quickly experiment

[ckoven]

this is a thing as of #462; closing.