Source

Defines the single input dataset. The type field is required and selects the loader.

type: xarray

Load from any file or URL supported by xarray (local NetCDF, Zarr, cloud storage, etc.).

Field	Type	Default	Description
type	"xarray"	—	Selects the xarray loader. Required.
urlpath	string	—	Path or URI to the dataset. Accepts local paths, gs://, s3://, https://, etc. Required.
engine	string	"zarr"	xarray engine: zarr, netcdf4, h5netcdf, scipy, cfgrib, …
open_kwargs	dict	{}	Extra keyword arguments forwarded verbatim to xarray.open_dataset.
mapping	dict	{}	Rename variables on load: {old_name: new_name}.
sel	dict	{}	Label-based selection applied after load (see sel / isel below).
isel	dict	{}	Index-based selection applied after load (see sel / isel below).
chunks	dict	{}	Dask chunk sizes applied on open: {dim: size}.

source:
  type: xarray
  urlpath: /data/wave_hindcast.zarr
  engine: zarr
  chunks:
    time: 100
    latitude: 50
    longitude: 50
  mapping:
    Hs: hs
    Tp: tp
  sel:
    latitude: {start: -50, stop: -30}
    time: {start: "2000-01-01", stop: "2020-12-31"}

type: intake

Load from an intake-forecast catalog. Requires the intake-forecast package.

Field	Type	Default	Description
type	"intake"	—	Selects the intake loader. Required.
catalog	string	—	Path or URI to the intake catalog YAML file. Required.
dataset_id	string	—	Entry name within the catalog. Required.
mapping	dict	{}	Rename variables on load.
sel	dict	{}	Label-based selection applied after load.
isel	dict	{}	Index-based selection applied after load.
chunks	dict	{}	Dask chunk sizes.

source:
  type: intake
  catalog: /catalogs/oceanum.yaml
  dataset_id: wave_global_era5
  chunks:
    time: 50

sel / isel

sel selects by coordinate label, isel by integer index — exactly as in xr.Dataset.sel and xr.Dataset.isel.

Each dimension value can be:

Value type	Behaviour	Example
Scalar	Select a single coordinate value	depth: 0.0
List	Select those specific coordinate values	level: [500, 850, 1000]
{start, stop} dict	Select a range — converted to slice(start, stop)	latitude: {start: -50, stop: -30}

Range dicts vs. lists

A {start, stop} dict is the only way to specify a range. A list like [-50, -30] selects exactly those two coordinate values, not all values between them — consistent with xarray's own behaviour. Either key may be omitted: {stop: -30} → slice(None, -30).

source:
  sel:
    latitude: {start: -50, stop: -30}         # range: all latitudes in [-50, -30]
    time: {start: "2000-01-01", stop: "2020-12-31"}
    depth: 0.0                                 # single value
    level: [500, 850, 1000]                    # exact list of values
  isel:
    member: 0                                  # first ensemble member by index
    x: {start: 100, stop: 200}                # index range