Heuristics

assign_index() automatically detects the affine transform by trying each source in order, using the first one that succeeds:

Priority	Source	Location	Reference
1	GeoTransform	CF grid mapping variable (e.g. spatial_ref)	GDAL GeoTransform
2	proj:transform	DataArray .attrs	STAC Projection Extension
3	model_tiepoint + model_pixel_scale	DataArray .attrs	GeoTIFF spec
4	spatial:transform	DataArray, data variable, or Dataset .attrs	Zarr Spatial Convention
5	1D coordinate arrays	Coordinate variables for x/y dims	Common in NetCDF

Grid mapping variable lookup

For priority 1, the grid mapping variable is found following CF conventions:

• DataArray: looks for a grid_mapping attribute on the DataArray

• Dataset: uses the first grid_mapping attribute found across data variables

• Fallback: a coordinate variable named spatial_ref

Zarr spatial convention lookup

For priority 4, the convention must be registered in the zarr_conventions attribute (by name spatial: or by UUID); bare spatial: attributes are ignored otherwise. Attributes are looked up on the DataArray for DataArrays, and on each data variable then the Dataset itself for Datasets — array-level attributes take precedence over group-level ones, per the convention.

If a spatial:dimensions attribute is present, it is also used to auto-detect the x/y dimension names when x_dim/y_dim are not passed to assign_index(). The listed names are interpreted as [y, x], following the convention’s examples; spatial:transform uses standard Affine coefficient ordering [a, b, c, d, e, f], mapping (column, row) -> (x, y).

Coordinate array fallback

For priority 5, coordinate variables must be 1D with at least 2 values. Pixel spacing is computed as x[1] - x[0] and y[1] - y[0], and coordinates are assumed to be pixel-centered.

Logging

Enable trace-level logging to see which source is used:

import logging
from rasterix.lib import TRACE

logging.basicConfig(level=TRACE)
logging.getLogger("rasterix").setLevel(TRACE)