Calls

An ordered list of stat operations. Each call is a dict with a func field naming a registered stat function, plus any function-specific parameters.

Common fields

Field	Type	Default	Description
func	string	—	Name of a registered stat function. Required.
data_vars	list or "all"	"all"	Variables to pass to the function. Use "all" for all dataset variables.
derived_vars	list	[]	Derived variables to compute before the stat runs. See Derived Variables.
dim	string	"time"	Dimension to reduce along.
group	string	null	Time component for groupby: month, season, year.
chunks	dict	{}	Rechunk the data before this call: {dim: size}.
tiles	dict	{}	Process in spatial tiles: {dim: tile_size}. Useful for memory-intensive stats like quantile or rpv.
use_dask_cluster	bool	true	Whether to use the Dask cluster for this call (if cluster.enabled: true).
use_flox	bool	true	Whether to use flox for groupby reductions. Efficient for most operations, but set to false for quantile on large grids — flox's quantile path uses ~2× the memory of the native xarray implementation.
nsector	int	null	If set, bin the data into this many directional sectors and apply the function to each.
dir_var	string	"dpm"	Directional variable used for sectorisation when nsector is set.
suffix	string	null	Override the auto-generated output variable suffix. Default: _{func} (plus _{group} and _direc if applicable).

Function-specific parameters

Any additional fields in a call dict are forwarded as keyword arguments to the stat function. See Operations for the parameters accepted by each built-in function.

Example

calls:
  - func: rpv
    data_vars: [hs]
    dim: time
    group: null
    chunks:
      time: -1        # full time axis in one chunk (required by rpv)
      latitude: 50
      longitude: 50
    tiles:
      latitude: 10
    use_dask_cluster: true
    use_flox: true
    nsector: 8
    dir_var: dpm
    suffix: _return_period
    # rpv-specific:
    return_periods: [10, 50, 100]
    percentile: 95
    distribution: gumbel_r
    duration: 24