termination threshold due to small cohort number is too high

[ckoven]

One thing I've noticed in looking at the termination statistics and understory dynamics is that a large fraction of demoted cohorts appear to be getting terminated en route between the canopy and understory when they are carved off from their donor cohorts but before they can be fused into larger understory cohorts of similar size. To address this, one solution is to reduce the minimum number of cohorts before they are terminated. I've pushed a version that includes this change to 3a7b068 on my fork for testing and comparison. These changes qualitatively change the size distribution of understory trees to allow for larger ones to exist, as they appear to allow for successful demotion of these cohorts. This doesn't solve our understory survival problem, as they pretty quickly succumb to carbon starvation anyway, but it may be a necessary step to allow us to address the understory survival issues.

[rosiealice]

Is it possible that this is an order of operations problem caused by the call to terminate_cohorts inside canopy_structure? The other calls, in patch_dynamics, and ecosystem_dynamics, both happen immediately after a fuse_cohorts. The was probably a good reason for the terminate_cohorts in canopy_structure, but maybe it's now redundant?

…

On 22 February 2017 at 11:44, Charlie Koven ***@***.***> wrote: One thing I've noticed in looking at the termination statistics and understory dynamics is that a large fraction of demoted cohorts appear to be getting terminated en route between the canopy and understory when they are carved off from their donor cohorts but before they can be fused into larger understory cohorts of similar size. To address this, one solution is to reduce the minimum number of cohorts before they are terminated. I've pushed a version that includes this change to 3a7b068 <3a7b068> on my fork for testing and comparison. These changes qualitatively change the size distribution of understory trees to allow for larger ones to exist, as they appear to allow for successful demotion of these cohorts. This doesn't solve our understory survival problem, as they pretty quickly succumb to carbon starvation anyway, but it may be a necessary step to allow us to address the understory survival issues. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#187>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMWsQ6MvOcal1TtJg2jpC9jT4Q3zBqjTks5rfIH6gaJpZM4MJCD1> .

-- ------------------------------------------------------------------------ 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. +1 303-497-1706 http://www.cgd.ucar.edu/staff/rfisher/

[ckoven]

very possibly it is an order of operations thing. i can try without the terminate_cohorts in canopy structure to see if that has qualitatively similar effects.

[ckoven]

trying that test now in 4157e60. i left in the call to terminate cohorts that was after the call to fuse them. it is plausible that things still won't survive if they are systematically terminated such that there is never an appropriate cohort to fuse into first, but we'll see if that happens in practice.

[ckoven]

@rosiealice the change in 4157e60 doesn't seem to fix the issue, whereas the one in 3a7b068 does. so it doesn't seem to be primarily an order of operations thing. on the other hand, there's nothing to say that we shouldn't do both sets of changes.

[ckoven]

was wondering if the changes in 4157e60 has any effect at all beyond those in 3a7b068. so i merged them together (37ba5ce) and did a run. below the change in understory biomass from that relative to 3a7b068. clearly there is a change, but it doesn't seem to be very large. will plan on keeping both sets of changes as it seems better to not terminate without fusing first if possible.

[ckoven]

OK, I tried another thing, which was to lower the threshold by 3 more orders of magnitude (e34e368). differencing that from the prior (37ba5ce) looks similar to the prior comparison. its not bit for bit the same, but the changes that emerge (in understory biomass, total number of cohorts, or understory size distributions) do not look qualitatively different from the control. so that suggests that the initial change in the threshold is enough to capture the dynamics that we were missing before. I don't see any benefit though to lowering it to that next level as we do want to eventually terminate cohorts whose number densities are sufficiently small (and also the further reduced threshold seems like it is starting to get close to unsafe math territory). as such, I am going to stick with the changes as of 37ba5ce and issue a PR on that.