MetOfficeRelativeHumidityCorrection

There are differences in values of relative humidity H(x) produced by the unified model (UM) interface and the Met Office observation processing (OPS) system. The differences are caused by the order in which (1) RH is computed from specific humidity (2) temporal and spatial interpolation are performed. The computation of relative humidity from specific humidity is nonlinear, which can lead to differences in H(x) of up to 5%.

This ObsFunction computes two values of relative humidity H(x). The first reproduces what occurs in the UM interface, i.e. it vertically interpolates the relative humidity GeoVaLs at each location. The second reproduces what occurs in OPS, i.e. it computes relative humidity from GeoVaLs of specific humidity, temperature, and pressure and then performs vertical interpolation. The output of the ObsFunction is the difference between the two interpolated H(x) values.

The H(x) difference can be added to observed relative humidity (using an Arithmetic ObsFunction) prior to running QC filters that rely on relative humidity O-B such as the Background Check. After those filters have run, the H(x) difference can be subtracted back off.

Parameters

  • observed pressure: [required] Name of observed pressure.

  • output relative humidity units: [required] Desired units of the output correction, percentage or fraction are valid strings.

  • capsupersat: [optional, default false] Cap relative humidity 1 for fraction and 100% for percentage. Default false.

Example yaml

The following yaml block shows how the ObsFunction can be used.

- filter: Variable Assignment
  assignments:
  - name: HofXCorrection/relative_humidity
    type: float
    function:
      name: ObsFunction/MetOfficeRelativeHumidityCorrection
      options:
        observed pressure: MetaData/pressure
        output relative humidity units: percentage
        capsupersat: true

The correction variable, HofXCorrection/relative_humidity, can be added to the RH observation values prior to using other filters that use RH O-B. (Adding the correction to the observation values is equivalent to subtracting it from the background values.) The correction should be subtracted off afterwards.