General Information

Interfaces

All submodules have both an xarray and a numpy interface. The numpy interface wraps the underlying C-extensions closely. The xarray interface wraps the numpy interface, handles coordinates automatically and is dask-compatible. It is highly recommended to use the xarray-based functions.

Coordinates and Units

• Input data must be provided on a regular longitude-latitude grid. Grid spacing is generally allowed to differ between the longitude and latitude dimensions but must be equally spaced within each dimension. E.g., a grid with resolution 1°×2° should work, but a Gaussian grid will not.

• Latitude coordinates must start in the north. The North Pole is located at +90°, the South Pole at -90°.

• Vertical coordinates must be specified top-down if the vertical coordinate is a core dimension. For a pressure coordinate that means values must be increasing (ascending order), for an isentropic coordinate values must be decreasing (descending order).

The following names for data coordinates are recognized by the xarray interface:

Name	Description	Unit	Notes	Override
latitude	latitude	°	N to S	lat
longitude	longitude	°		lon
level	pressure (isobaric levels)	hPa	ascending if core dim	isob
isentrope	potential temperature (isentropic levels)	K	descending if core dim	isen

The following names for data fields are recognized by the xarray interface:

Name	Description	Unit	Notes	Override
u	zonal wind	m/s	positive towards east	u
v	meridional wind	m/s	positive towards north	v
t	temperature	K		t
pres	pressure	hPa		pres
pt	potential temperature	K		pt
sg	isentropic density	kg/K/m²		sg
vo	relative vorticity	1/s		vo
av	absolute vorticity	1/s		av
pv	Ertel potential vorticity	PVU		pv
ks	stationary wavenumber	unitless		ks

Overrides can be provided for nonconforming data in any function of the xarray interface that accepts a names argument. E.g., if a pressure coordinate is called isobaricInhPa instead of level, use an override (based on the override name from the tables above):

>>> rwguide.xarray.pvgradient.potential_temperature_isob(..., names={ "isob": "isobaricInhPa" })

Vectorization

Many functions offer to use the vectorization of xarray.apply_ufunc(). This vectorization mode can be slower than the vectorization built into the C-extensions but often uses much less memory.