Pseudo-model¶

Pseudo-model is a generic model implementation for model interfaces that read their data from file rather than connecting to a forecast model. PseudoModel can be used for any MODEL which has implemented basic components such as the Geometry and State.

The state configuration depends upon the implementation in the particular model in use, however, for compatibility with the Pseudo-model, it must include a specification of the validity datetime of that state, and each state in the state list must be seperated by a single time step of size tstep.

Here is an example configuration for a 4D H(x) application (HofX4D):

#.
#.
#.
forecast length: P3D
time window:
  begin: 2023-06-01T00:00:00Z
  length: P4D
initial condition:
  date: 2023-06-02T00:00:00Z
  state variables: [ sea_surface_temperature ]
model:
  name: PseudoModel
  tstep: P1D
  states:
    - date: 2023-06-02T00:00:00Z
      state variables: [ sea_surface_temperature ]
      #.
      #.
      #.
    - date: 2023-06-03T00:00:00Z
      state variables: [ sea_surface_temperature ]
      #.
      #.
      #.
#.
#.
#.

To specify many states and to reduce redundant information, the state can also be specified using a template:

#.
#.
#.
model:
  name: PseudoModel
  tstep: P1D
  states from template:
    start datetime: 2023-06-02T00:00:00Z
    number of states: 2
    pattern: "@numerical_counter@"
    start: 1
    except: [2, 3]
    zpad: 1
    template:
      state variables: [ sea_surface_temperature ]
      data field: "special-number-@numerical_counter@"
      #.
      #.
      #.
#.
#.
#.

Each state is generated from the template section. The date is constructed from start datetime and model:tstep to create a list of number of states state configuration entries. There is also the facility to substitute a pattern into the state template. The pattern is specified using:

start: for the starting value
zpad: the number of zeros to left-pad the pattern
except: values of the pattern to skip over
pattern: the string to replace with the pattern value