Model failed with varying direct logging mortality by PFT size class

[XiulinGao]

Describe the issue

Hi folks! I am working on one of my local branches where I mainly modified the logging model to do the following things:

1. Apply different logging direct mortality rates given the current cohort DBH with a capped rate for each defined size class (e.g. max. 75% direct mort for cohort with DBH between 5-25 cm, 20% for cohort with DBH > 75cm)
2. with this size-based selective logging approach, I eventually aim to reach a target tree basal area (e.g. 34 m2/ha) at the patch level. That being said, if this target basal area is approximately approached (allowing for 1 m2/ha error), all direct mortality will be set to 0 even DBH thresholds are met for logging. To do this, I calculate delta basal area (current patch total basal area - target basal area) and update this after each logging that happened at cohort level (only account for trees that are directly logged).
3. I manually changed the logging relevant parameters to make any mortality that results from the logging side effects (e.g. collateral, fracture) to be 0, but allow for degradation.
4. I used area based logging approach. And just to keep the externally read-in harvest_rate to be still used during any logging event and to avoid any potential differences in total disturbance rate resulted from logging, I made the external harvest_rate to be 0.75, which is the maximum direct logging rate that can be reached in 1) for any size class. This is to make sure the summed disturbance rate (direct + collateral + infrastructure + degradation) is always equal to 0.75 if it ever happens.

Then I ran into model failure , of course :) . See the log output below for details for the error message. In brief, it said that demotions don't add up and model terminated by line 625 in EDCanopyStructureMod. Honestly, I don't quite understood what it does in that Canopy structure mod, but found another posted issue that might be related #714

Strangely, if I made all direct logging mortality to be exact the externally read-in harvest_rate, which also skips the target basal area check (so direct logging mort will apply as long as it's within the DBH ranges for logging to happen), then the error goes away. So my current guess is that it either has something to do with 1) not apply direct logging mortality when it is supposed to log or 2) the different direct mort rate applied to different size group or a combined effects of 1) and 2).

Any insights would be appreciated! Thanks!

Relevant log output

demotions dont add up
 sum demotions:   2.084800163685286E+044
 area needed to be demoted:   2.084800163685285E+044
 excess:   1.188422437713965E+029
 ENDRUN:
 ERROR in EDCanopyStructureMod.F90 at line 625

FATES tag

sci.1.71.0_api.33.0 with local code change for logging module

Host land model tag

ctsm5.1.dev164

Machine

derecho

Other supported machine name

No response

Additional context

No response

[XiulinGao]

Update: this error does not exist when running with FATES API31 with the exact same code changes to logging disturbance. So this might be something specific for API33 and the newer versions, but I haven't been able to track down to the root of bug.