Open dataset parameters

This page is intended to provide level of support/applicability of the various parameters that can be passed to the open_dataset function when dealing with the different dataset layouts. Tabular and trajectories datasets are recent additions to the anemoi-datasets package and are still under development. The meaning of the emojis, applied per layout column, is as follows:

• ✅: Parameter has been tested and works for this layout.

• 🔁: Parameter will work, but has a slightly different meaning or behaviour for this layout.

• ❌: Parameter is not applicable to this layout.

• ❓: Parameter may work, but has not been tested with this layout.

• 🧪: Parameter should work, but has not been tested with this layout.

• ⚠️: Parameter may work, but the behaviour is not fully understood.

• 🛠️: Will work in the future. Not yet implemented or tested, but expected to work without major issues.

• 🗑️: Is obsolete and will be removed in the future. Should not be used for new code, but may still work for now. Will be removed in a future release.

Warning

Datasets of different layouts cannot be combined. So concat, join, and similar operations are not supported between layouts. The table below assumes that when combining datasets, they have the same layout. See Dataset layouts for more information.

Parameters for open_dataset

parameter	gridded	tabular	trajectories	comment
adjust	✅	⚠️	⚠️	Adjustment mode when combining datasets, e.g. select common dates, variables, etc. This needs testing and possibly decision on expected behaviour for tabular and trajectory datasets.
area	✅	❌	🧪	Spatial cropping area as a list [lon_min, lat_min, lon_max, lat_max].
base_end	❌	❌	✅	Trajectories only: filter the base-date axis directly (no valid-time envelope logic). See Trajectories-only options.
base_start	❌	❌	✅	Trajectories only: filter the base-date axis directly (no valid-time envelope logic). See Trajectories-only options.
chain	🗑️	🗑️	🗑️	Experimental chain operation. Same behaviour as concat, but does not check that the dates are continous. Will be removed in the future.
concat	✅	⚠️	🧪	Concatanate two or more datasets along the time dimension. That may work, but the behaviour of the windowing at the seam is not well defined. So it should be skipped for now.
complement	✅	❌	🧪	Complement/cutout configuration (used for creating complements).
cutout	✅	❌	🧪	List of datasets used as cutouts for complements/cutout operations.
drop	✅	🧪	🧪	Variables to drop (list).
end	✅	🔁	🔁	Set the end date for the opened dataset. For gridded datasets, the date must be present in the dataset. For tabular datasets, the date is used as-is, and any windows requested between the actual end of date of the data and that date will return empty arrays (See Tabular). For trajectories, the date applies to the validity-time envelope: a base date is kept iff [base + step_start, base + step_end] lies inside [start, end] (See Trajectories). Use base_end to filter the base-date axis directly.
ensemble	✅	❌	🧪	List of datasets forming an ensemble (e.g. ensemble=[d1, d2]).
fill_missing_dates	✅	❌	❌	Method to fill missing dates (“interpolate” or “closest”). Trajectories have no missing-date concept.
fill_missing_gaps	✅	❌	❌	Fill virtual datasets for gaps when concatenating.
frequency	✅	🔁	🔁	For gridded dataset, select the frequency of the return sample; it must be a multiple of the dataset frequency. For tabular datasets, it is used to create windows of the specified frequency (e.g. “1D” for daily windows) and is not connected to the dataset frequency (which is undefined) (See Tabular). For trajectories, frequency acts on the base-date axis; use step_frequency for the step axis (See Trajectories).
grids	✅	❌	🧪	List of grids/datasets to combine as multiple grid sources.
interpolate_frequency	✅	❌	🧪	Frequency used to interpolate a dataset to a higher temporal resolution.
interpolate_variables	✅	❌	🧪	Variables to interpolate spatially (with optional max_distance).
interpolation	✅	❌	🧪	Interpolation method (example: “nearest”).
join	✅	🧪	🧪	Join two or more datasets along the variable dimension.
max_distance	✅	❌	🧪	Maximum distance used by spatial interpolation (e.g. nearest-neighbour).
member, members	✅	❌	🧪	0-based member selection (see number for 1-based selection).
merge	✅	❌	🧪	Merge operation key to combine datasets by overlaying fields.
name	✅	✅	✅	Experimental. Optional name assigned to the resulting dataset subset that can be used to name masks that will be retrieved in inference.
number, numbers	✅	❌	🧪	1-based member selection (see member for 0-based selection).
reorder	✅	🧪	🧪	Reorder variables (list or mapping).
rename	✅	🧪	🧪	Rename variables mapping.
rescale	✅	🛠️	🧪	Rescaling mapping/tuples/units for variables.
select	✅	✅	🧪	Select variables (list, set or string).
set_missing_dates	✅	❌	❌	Debug option: list of dates to mark as missing.
shuffle	🗑️	🗑️	🗑️	Boolean to shuffle dataset indices when subsetting.
skip_missing_dates	✅	❌	❌	Boolean: skip missing dates when iterating (requires expected_access). Trajectories have no missing-date concept.
source	✅	❌	🧪	Source dataset name/path used in complement examples.
start	✅	🔁	🔁	Set the start date for the opened dataset. For gridded datasets, the date must be present in the dataset. For tabular datasets, the date is used as-is, and any windows requested between that date and the actual start of date of the data will return empty arrays (See Tabular). For trajectories, start applies to the validity-time envelope (see end). Use base_start to filter the base-date axis directly.
statistics	✅	🧪	🧪	Use the statistics of another dataset.
step	❌	❌	✅	Trajectories only: select a single forecast step. Returns a 4-D view (base_dates, variables, ensembles, cells) — shape-compatible with a gridded dataset at that lead time. See Trajectories-only options.
step_end	❌	❌	✅	Trajectories only: end of a step-axis range selection.
step_frequency	❌	❌	✅	Trajectories only: frequency for a step-axis range selection.
step_start	❌	❌	✅	Trajectories only: start of a step-axis range selection.
steps	❌	❌	✅	Trajectories only: list of forecast steps to retain. Keeps the 5-D shape with the step axis narrowed. See Trajectories-only options.
thinning	✅	✅	🧪	Thinning factor or proportion.
trim_edge	✅	❌	🧪	Tuple to trim edges of the grid (e.g. (1,2,3,4)).
window	❌	✅	❌	Window specification for tabular datasets. For gridded and trajectory datasets, that parameter is ignored. See Tabular for details.
x	🗑️	🗑️	🗑️	Experimental: x coordinate for xy selection.
xy	🗑️	🗑️	🗑️	Experimental xy selection mode.
y	🗑️	🗑️	🗑️	Experimental: y coordinate for xy selection.
zip	🗑️	🗑️	🗑️	Experimental zip mode to combine datasets.