xrspatial.reproject.merge#

xrspatial.reproject.merge(rasters, *, target_crs=None, resolution=None, bounds=None, resampling='bilinear', nodata=None, strategy='first', chunk_size=None, transform_precision=16, name=None, bounds_policy='auto')[source]#

Merge multiple rasters into a single mosaic.

Each input is reprojected to the target CRS (if needed) and placed into a unified output grid. Overlapping regions are resolved using the selected strategy.

Parameters:

  • rasters (list of xr.DataArray) – Input rasters to merge.

  • target_crs (optional) – Target CRS. Defaults to the CRS of the first raster.

  • resolution (float or (float, float) or None) – Output resolution in target CRS units.

  • bounds ((left, bottom, right, top) or None) – Explicit output extent.

  • resampling (str) – Interpolation method: ‘nearest’, ‘bilinear’, ‘cubic’.

  • nodata (float or None) – Nodata value for the output.

  • strategy (str) – Merge strategy: ‘first’, ‘last’, ‘mean’, ‘max’, ‘min’.

  • chunk_size (int or (int, int) or None) – Output chunk size for dask. If None, defaults to 512 for the standard dask path and 2048 for the in-memory streaming and dask+cupy paths (chosen to amortize kernel launch overhead).

  • transform_precision (int) – Control-grid subdivisions for the coordinate transform (default 16). Higher values increase accuracy at the cost of more pyproj calls. Set to 0 for exact per-pixel transforms matching GDAL/rasterio.

  • name (str or None) – Name for the output DataArray.

  • bounds_policy ({"auto", "raw", "clamp", "percentile"}, default "auto") – How to derive the unified output extent from the input rasters when bounds is not supplied. See reproject() for the full description of each option. Ignored when bounds is given.

Returns:

The output y coordinate is always emitted in descending order (top-down, north-up) regardless of the input direction. This matches the standard raster convention and the output of common GIS libraries.

Non-spatial coords from the first raster (such as a scalar time coord) are carried through to the output. Coords aligned to the spatial dims are dropped because their values do not apply to the merged grid.

Return type:

xr.DataArray

Notes

There is no streaming in-memory path; for very large output mosaics, pass dask-backed inputs (or rely on the automatic promotion to the dask path) so that each output chunk is computed independently.

Examples

>>> import xarray as xr
>>> import numpy as np
>>> from xrspatial.reproject import merge
>>> tile_a = xr.DataArray(
...     np.full((32, 32), 1.0), dims=['y', 'x'],
...     coords={'y': np.linspace(50, 45, 32),
...             'x': np.linspace(-5, 0, 32)},
...     attrs={'crs': 'EPSG:4326'},
... )
>>> tile_b = xr.DataArray(
...     np.full((32, 32), 2.0), dims=['y', 'x'],
...     coords={'y': np.linspace(50, 45, 32),
...             'x': np.linspace(0, 5, 32)},
...     attrs={'crs': 'EPSG:4326'},
... )
>>> mosaic = merge([tile_a, tile_b], resolution=0.5)
>>> mosaic.shape[1] >= 32
True
On this page
Show Source