pelagos_py.steps.quality_control.range_qc

class pelagos_py.steps.quality_control.range_qc.range_qc(data, **kwargs)

Bases: pelagos_py.steps.base_qc.BaseQC

Flag measurements by value range: a single replacement for the old separate “gross range” and “impossible range” tests.

Each {flag: bounds} entry describes one or more bands:

• ``[low, high, “outside”]`` — a good band; data outside it gets the flag (accepts outside/out/o).

• ``[low, high, “inside”]`` — an impossible band; data between the bounds inclusive gets the flag (accepts inside/in/i).

• A single scalar value — flags exact matches (e.g. a fill value 4: 0.0).

A flag may give a list of bands to cover several ranges. If the keyword is omitted the bound order decides: ascending [low, high] means outside, descending [high, low] means inside. Entries apply most-severe-flag-first, so the worse flag wins on overlap; anything checked but not flagged is marked good (1).

Target Variable: Any Flag Number: Any (user-defined) Variables Flagged: Any (the tested variables, minus any flag_instead sources, plus any also_flag/flag_instead companions)

Example

- name: "Apply QC"
  parameters:
    qc_settings:
      range qc:
        variable_ranges:
          PRES:
            3: [-2.4, -5, inside]   # impossible band: flag data INSIDE it
            4: [-5, -.inf, inside]
            9: 0.0                  # single scalar -> flag the exact fill value 0.0
          TEMP:
            3: [0, 30, outside]     # good band: flag data OUTSIDE it
            4: [-2.5, 40, outside]
          CNDC:
            # one flag, two bands: flag bad both inside [2, 3] and outside [0.1, 10]
            4: [[2, 3, inside], [0.1, 10, outside]]
        also_flag:
          CNDC: [PRES, TEMP]    # CNDC's flags propagate onto PRES & TEMP (worst wins)
        test_depth_range: [0, 100]    # OPTIONAL: only check this DEPTH window
  diagnostics: true             # plots every flagged variable, coloured by flag

Flagging one variable by another’s range (flag_instead) — here CHLA is judged by DEPTH (probably-bad in the quenched top 5 m, probably-good below it) while DEPTH itself is left unflagged:

- name: "Apply QC"
  parameters:
    qc_settings:
      range qc:
        variable_ranges:
          DEPTH:
            3: [0, 5, inside]     # top 5 m -> probably bad (correctable)
            2: [0, 5, outside]    # deeper   -> probably good
        flag_instead:
          DEPTH: [CHLA]           # DEPTH's flags go onto CHLA; no DEPTH_QC written

return_qc()

Representative of QC processing, to be overridden by subclasses.

Returns:

flags – Output QC flags for the data specific to the test.

Return type:

array-like

plot_diagnostics()

Representative of diagnostic plotting (optional).