Reconcile parameters and scales driving leaf phenology and mortality

[mpaiao]

Deciduousness can be thought as a defence mechanism against environmental conditions. For example, drought deciduous should make plants resistant to hydraulic failure mortality, because plants abscise leaves during the dry season.

However, the parameters that drive leaf phenological cycles are completely independent from those that impact mortality. In fact, they do not even use the same environmental driver to predict mortality and deciduousness: drought deciduous is driven by soil matric potential, whereas hydraulic failure mortality is driven by either BTRAN or fraction of maximum conductivity. There is a mismatch in time scale too: phenology uses 10-day averages of environmental drivers, whereas hydraulic failure mortality seems to use instantaneous values (please correct me if I am wrong). This set of inconsistencies creates the risk of deciduous PFTs dying of the very mechanism deciduousness should protect them from.

Below are some initial thoughts about how to address this for drought deciduous in an upcoming pull request that I plan to submit. However, I am not entirely sure how to go about this, so would be quite appreciated!

1. Use a single environmental driver for both mortality and phenology. I am agnostic to it, but maybe we can use what is currently used for hydraulic failure mortality.
2. Add consistency tests in subroutine PRTCheckParams to prevent simulations to start whenever the environmental threshold for phenology would be less than the threshold for mortality. Alternatively, we can drop the lower bound for phenology parameter altogether and use the hydraulic failure parameters for both hydraulic failure and mortality, though maybe we want to keep some "safety margin" between them.
3. Ensure that the time scale mismatch is no longer an issue. I can see two approaches:
   • If we keep both phenology and hydraulic failure thresholds, impose leaf abscission whenever the 10-day average falls below the phenology threshold OR the instantaneous value falls below the hydraulic failure mortality. For leaf flushing, this may still lead to risks of failure.
   • Start using 10-day averages for hydraulic failure mortality. This would effectively eliminate inconsistencies and simplify the number of logical tests and make flushing safer. However, this depends on whether hydraulic failure mortality is nearly instantaneous or if it requires multiple dry days to effectively kill the plant. This would also likely impact existing calibrations for both evergreen and deciduous PFTs, though hopefully not by a lot.

I am focussing on drought deciduous here, but I can see a similar case for cold deciduous too: we use of growing degree days and chilling days for phenology, and average temperature for cold-related mortality.

[mpaiao]

Following the FATES software engineering meeting discussion last week, we agreed with the following plan:

• Update the drought deciduous phenology code, to ensure that the variable driving phenology is the same variable that drives hydraulic failure mortality.
• For now, we will keep the 10-day averaging for drought deciduous phenology, but we will not change the hydraulic failure mortality, which uses instantaneous data.

This approach should not have any impact on ongoing calibrations that involve the hydraulic failure mortality parameters, at least for evergreen and cold-deciduous PFTs. We will see how this goes, and decide if this solution is sufficient or if it requires changes in the hydraulic failure mortality module.