V0 compatability with fabm_get_light_extinction

[janko-om]

Context: I run SCHISM-FABM-ECOSMO and recently switched to FABMv2. The switch entailed changing the used light model to gotm/light. I however cannot reproduce my pre-FABMv2 results with this setup because light attenuation due to phytoplankton self-shading and POM /DOM behaves completely different. Tuning sofar didn't get me anywhere. I now want to (temporarily) call the old get_light_extinction() subroutine to calculate light in FABMv2 (instead of gotm/light), so I can produce some results while I figure out what the problem with gotm/light is.

I however run into following problem in the fabm_schism.F90 driver during compilation:

1031 |       call fabm_get_light_extinction(fs%model,1,nvrt,nel,localext)
      |                                                                  1
Error: Type mismatch in argument 'self' at (1); passed TYPE(type_fabm_model) to CLASS(__class_fabm_v0_compatibility_Type_model_t)
/home/g/g260156/sandbox/schism_dev5.11/src/Fabm/fabm_schism.F90:1066:44:

 1066 |     call fabm_get_light(fs%model,1,nvrt,nel)
      |                                            1
Error: Type mismatch in argument 'self' at (1); passed TYPE(type_fabm_model) to CLASS(__class_fabm_v0_compatibility_Type_model_t)

Whats the correct way of calling fabm_get_light_extinction() and fabm_get_light() in FABMv2?

[jornbr]

A mix of FABM v1/v2 and and FABM v0 compatibility APIs won't work, as you found. To get light extinction from FABM v1 and higher, you'd do:

call model%request_variable(fabm_standard_variables%attenuation_coefficient_of_photosynthetic_radiative_flux)

before calling model%start(), and then once, after:

ext => model%get_data(model%get_interior_variable_id(fabm_standard_variables%attenuation_coefficient_of_photosynthetic_radiative_flux))

with ext being declared as a full interior array with the pointer attribute.

The first call ensures that the extinction will be computed, even if no FABM model has an interest in it, and the second retrieves a pointer to the extinction values that will be updated throughout the simulation. It may be slightly more verbose than in v0, but it ensures we can handle any feedback to physics without having to introduce new variable-specific APIs.

fabm_get_light is no longer needed in FABM v1 and higher; in the v0 compatibility layer it is there but does nothing.

[janko-om]

No, fabm_schism didn't call model%finalize_outputs, and adding it did the trick! Thanks a lot!

[janko-om]

Short side question: I now use the get_light subroutine from FABMV0 to calculate PAR attenuation. But unfortunately the standard parameter settings produce far too strong light attenuation. I for example need to choose BioC(5) (--> phyto self-shading [units m2/mgC]) two orders of magnitude smaller, and EXDet/ EXDOM need to be one order of magnitude smaller.

I use BioC(5) [units m2/mgC] as the scale factor when registering the contribution to light attenuation:
call self%add_to_aggregate_variable(standard_variables%attenuation_coefficient_of_photosynthetic_radiative_flux, self%id_dia, & scale_factor=self%BioC(5))

I assumed that results in: diatom concentration [mgC/m3] * scale_factor [m2/mgC] = [1/m]. Is that correct? Or is there another plausible explanation for the substantial difference in the parameter settings?

[jornbr]

Your interpretation is fine, in the sense that it is identical to what FABM would do. But since you say you needed to makes the specific attentuation coefficients smaller: what was the change that caused this? E.g., a change in the biogeochemical code, the FABM version, the host coupler, or?...

[janko-om]

To my understanding the specific attenuation coefficients should be the same, and I don't understand why they need to be changed. The essential change was the change of the FABM version from v0 to v2.1.3. The biogeochemical code was not changed, except for the modifications necessary to make it compatible with FABMv1. This for example included removing the fabm_get_light_extinction subroutine and replacing it with calling self%add_to_aggregate_variable(). The SCHISM-FABM driver underwent substantial changes when upgraded to FABMV1, but as far as I can see the versions are equivalent for everything related to PAR. Especially now that I basically use the old get_light() subroutine that I used with FABMv0. Using only (physical) background extinction gives identical results to the FABMV0 version. So my best guess was that I missed something regarding the scaling of the specific light attenuation coefficients in the call of add_to_aggregate_variable(). I can work with the new parameter settings for now, but I'm still puzzled...

[jornbr]

That does sound puzzling. One thing to check is that you're not double-counting attenuation by having both add_to_aggregate_variable and get_light_extinction implemented (the latter is deprecated - add_to_aggregate_variable should be used instead). But even if you did have both, it would explain overestiamtion of extinction by a factor 2, not by an order of magnitude...

[janko-om]

Thanks for the quick reply! I still register the contributors to attenuation as:

call self%add_to_aggregate_variable(standard_variables%attenuation_coefficient_of_photosynthetic_radiative_flux, self%id_dia, & scale_factor=self%BioC(5))

Is that correct? And this didn't work:

call model%request_variable(fabm_standard_variables%attenuation_coefficient_of_photosynthetic_radiative_flux)

I did some digging and found that this works:

call fabm_request_variable(fabm_standard_variables%attenuation_coefficient_of_photosynthetic_radiative_flux)

Just wanted to double check that this is the right way initializing the extinction?

[jornbr]

Ah apologies - that should have been model%require_data rather than model%request_variable

The syntax to register contributions to attenuation from the biogeochemical model (call self%add_to_aggregate_variable(...)) is fine.