Accessing previous values of diagnostic variables

[BeatrizArellano]

Hi all,

I am working on a biogeochemical model and have a variable currently implemented as a diagnostic. For numerical reasons, I would like to use its value from the previous timestep as an input (as an initial guess in an iterative solver).

I understand that state variables naturally provide this access to the last value, but in this case the variable should not be treated as a state variable (i.e. it should not be transported or time-integrated by the host).

My questions are:

• Is there any mechanism in FABM to access the previous value of a diagnostic variable?

• Does setting act_as_state_variable=.true. for a diagnostic variable can be used to retrieve its previous value, or is it only used for internal coupling and source-sink handling?

Many thanks in advance for your help!

[jornbr]

Hi! You'll need a workaround to access the (previous) value of the diagnostic. In the module that implements the diagnostic, you can register a dependency with the same name. This dependency is then implicitly coupled to the diagnostic and will provide read access to it. In the routine that updates the original diagnostic, the dependency will be equal to the diagnostic's previous value.

act_as_state_variable=.true. is typically used for more exotic functionality. It can be used by module A to create a diagnostic that look like a state variable to another module B, which then contributes source terms (or bottom/surface fluxes) to it. Module A can then retrieve those sources and do something useful with them - without B having been coded explicitly to expose its source terms. This sounds pretty abstract, so here are some real examples:

• Module A creates a tracer-like "total alkalinity" but redirects all its source terms (from modules B) to a "biological contribution to alkalinity"; the value of total alkalinity is calculated by A by summing a salinity-derived alkalinity estimate and the biological contribution. (This is an option in ERSEM)
• Module A creates a 3D predator concentration (diagnostic with act_as_state_variable=.true. ), calculated from a 2D depth-integrated biomass (a state variable) + assumed vertical distribution, thereby allowing all modules B to contribute sources and sinks to the 3D field. Module A then collects those sources/sinks and depth-integrates them to calculate sources/sinks for the depth-integrated biomass.

[BeatrizArellano]

Thanks for your help @jornbr and for the clear explanation! The workaround worked pretty well.

[jornbr]

I have now ported some of the documentation describing these features to from its not-easy-to-find PDF to the wiki

[BeatrizArellano]

Thanks @jornbr, that is very helpful! I was not aware of that document.