slv.inversion.covariances.build_mdm_error#

slv.inversion.covariances.build_mdm_error(name, obs_index, std, correlated=True, scale=None, interday=False, time_dim='obs_time', site_config=None, obs_sites=None, **kwargs)[source]#

Unified builder for Model-Data Mismatch covariance components.

Returns a covariance model, to build on the observations’ space (whose labels must be obs_index, in that order: a per-observation std is read in that order).

Parameters:

  • name (str) – Component name (used in error messages)

  • obs_index (pd.MultiIndex) – Observation index with ‘obs_location’ and ‘obs_time’ levels

  • std (float or dict) – Absolute standard deviation in ppm. If dict with nested dicts: std[site][season] structure (site/season-specific) If dict with scalar values: std[organization] structure (organization-specific)

  • correlated (bool) – Whether errors are correlated

  • scale (str, optional) – Time scale for correlation decay

  • interday (bool) – Whether correlations extend across midnight

  • time_dim (str) – Name of the time dimension

  • site_config (pd.DataFrame, optional) – Site configuration with ‘organization’ column, required for organization-based std

  • obs_sites (sequence of str, optional) – The site of each observation, aligned with obs_index, for site- or organization-keyed std. Defaults to the obs_location level, which is the site for a tower but a PYSTILT location_id for a mobile receptor.

  • **kwargs – Additional unused parameters (for compatibility)

Return type:

CovarianceModel