Methods and attributes

For a complete list of methods and attributes available for the objects returned by open_dataset, please see Dataset. Here, a subset of stable and commonly used methods and attributes are described in more detail.

Warning

All methods and attributes will take into account any subsetting, selecting, or combining used to construct the final dataset, with the exception of statistics, which will return the values of the first dataset encountered. See Statistics for more details.

Methods

__len__()

The number of rows (dates) in the dataset.

__getitem__(key)

Access the dataset’s data values. With a few exceptions, the package supports the same indexing and slicing as NumPy. The following examples are valid:

ds[0]
ds[-1]
ds[0:10]
ds[0:10:2]
ds[0, 1, :]

The data returned is a NumPy array. Please note that Zarr will load the entire dataset into memory if you use a syntax like ds[:].

metadata()

Returns the dataset’s metadata.

provenance()

Returns the dataset’s provenance information.

source(index)

For debugging. Given the index of a variable, this will return from which Zarr store it will be loaded. This is useful for debugging combining datasets with join.

tree()

For debugging. Returns the dataset’s internal tree structure.

statistics_tendencies(delta):

Statistics tendencies are statistics of the difference between two dates, typically over 1h, 3h, 6h, 12h and 24h, depending on the underlying frequency of the dataset. The delta is given in hours, or as a datetime.timedelta. See statistics: below for the format of the returned information.

Attributes

shape:

A tuple of the dataset’s dimensions.

field_shape:

The original shape of a single field, either 1D or 2D. When building datasets, the fields are flattened to 1D.

dtype:

The dataset’s NumPy data type.

dates:

The dataset’s dates, as a NumPy vector of datetime64 objects.

frequency:

The dataset’s frequency (i.e., the delta between two consecutive dates) in hours.

latitudes:

The dataset’s latitudes as a NumPy vector.

longitudes:

The dataset’s longitudes as a NumPy vector.

statistics:

The dataset’s statistics. This is a dictionary with the following entries:

{
    "mean": ...,
    "stdev": ...,
    "minimum": ...,
    "maximum": ...,
}

Each entry is a NumPy vector with the same length as the number of variables, each element corresponding to a variable. You can therefore use it like:

values = ds[0]
normalised = (values - dataset.statistics["mean"]) / dataset.statistics["stdev"]

Use the name_to_index attribute to map variable names to indices.

resolution:

The dataset’s resolution.

name_to_index:

A dictionary mapping variable names to their indices.

print(dataset.name_to_index["2t"])

variables:

A list of the dataset’s variable names, in the order they appear in the dataset.

missing:

The set of indices of the missing dates.

grids:

A tuple of the number of grid points for each dataset that is combined with the grids method.