Operations overview¶

An operation (or "op") is a function that reduces an xr.Dataset along one or more dimensions and returns a result dataset. Operations are registered by name in the registry and referenced from config via the func field.

Built-in operations¶

`func`	Description
`mean`	Arithmetic mean
`max`	Maximum value
`min`	Minimum value
`std`	Standard deviation
`count`	Count of non-NaN values
`quantile`	Quantiles at specified levels
`pcount`	Percentage of non-NaN values
`exceedance`	Probability that values exceed a threshold
`nonexceedance`	Probability that values stay below a threshold
`rpv`	Return period values via extreme-value analysis
`distribution3`	3-D joint histogram (Hs × Tp × Dir)
`distribution2`	2-D joint histogram (speed × direction)
`distribution3_timestep`	Memory-efficient 3-D histogram (accumulated in time chunks)
`statdir`	Apply multiple functions over directional sectors
`modal_direction`	Per-cell modal direction from a weighted circular histogram
`hmo`	Frequency-domain Hs in spectral bands
`range_probability`	Probability of values falling within specified ranges
`winpow`	Wind turbine power from wind speed

Output variable naming¶

Each operation's output variables are renamed with a suffix before merging into the output dataset. The default suffix is:

{variable}_{func}[_{group}][_direc]

Examples:

Call	Input var	Output var
`func: mean`	`hs`	`hs_mean`
`func: mean, group: month`	`hs`	`hs_mean_month`
`func: rpv, nsector: 8`	`hs`	`hs_rpv_direc`
`func: rpv, suffix: _100yr`	`hs`	`hs_100yr`

Grouping¶

All operations that accept a group parameter will compute the statistic separately for each value of a time component. Supported groups:

`group`	Output dimension	Values
`month`	`month`	1–12
`season`	`season`	DJF, MAM, JJA, SON
`year`	`year`	actual years present in the data

Directional sectorisation¶

Setting nsector on any call bins the data by direction before applying the function. For example, nsector: 8 creates 8 × 45° sectors. The output gains a direction dimension with sector-centre values (0°, 45°, …, 315°).

The dir_var field (default: dpm) names the directional variable used for binning. It must be present in the dataset alongside the data_vars.

Derived variables¶

Derived variables are pre-computed quantities added to the dataset before any stat function runs. They transform raw input variables (e.g. u/v components) into derived quantities (e.g. wind speed) that can then be used as data_vars.

See the Derived Variables section for the full function reference.

Custom operations¶

See Custom Plugins for how to register your own stat functions.