Model options

Model options let a component change which variables or equations it creates without introducing a separate selectable submodule. Every component receives inputs in its get_constraints(m, inputs) function and reads options through inputs.config_value. Defaults are defined once in config_default.yaml and supplied by configuration validation.

Keep an option beside the scientific parameters for the component that uses it. For example, ACCREU adaptation settings belong in economics.damages.accreu, and the sea-level-rise projection belongs in sealevelrise.

Adding an option

For example, the ACCREU damage module can be constructed with no adaptation, combined adaptation, or separate adaptation by sector. Define the option under economics > damages > accreu in config_default.yaml:

mimosa/inputdata/config/config_default.yaml

economics:
  damages:
    accreu:
      adaptation:
        descr: How adaptation is represented in ACCREU
        type: enum
        values:
          - noadaptation
          - combined
          - separate
        default: separate

Defining each named option as a normal configuration entry gives it type checking, a default value and an entry in the generated parameter reference.

The component reads the option while adding its variables and equations:

def get_constraints(m, inputs):
    adaptation = inputs.config_value("economics.damages.accreu.adaptation")

    if adaptation == "sectoral":
        # Add separate adaptation variables and equations for each sector
        ...

Users can change the option before creating the model:

params = load_params()
params["economics"]["damages"]["accreu"]["adaptation"] = "sectoral"
model = MIMOSA(params)

Model option or Pyomo parameter?

Use a model option when the value changes which model objects are created. For example, an option can enable a sector or choose between a combined and sector-specific set of adaptation equations.

Use a Pyomo Param for a numerical or domain assumption within equations, such as an adaptation cost, effectiveness coefficient or start year. Parameters receive values when declared on the concrete model through explicit initialization. See Adding parameters and data.

Testing model options

At minimum, test that:

1. the configuration parser accepts every documented option value;
2. each option produces the intended variables and equations;
3. the documented default produces the normal model structure; and
4. unsupported option values give a clear configuration error.

Model construction with MIMOSA(params, prerun=False) is normally sufficient. Only behaviour that depends on an optimiser needs a full solver run.