Reprojection¶

AerEO can write data in its native projection or reproject it to a target CRS and resolution. The choice is controlled by the reproject and reproject_mode fields of ExtractionJob.

Reprojection modes¶

Mode	What it does	Best for
`raw`	Reproject the whole dataset once and write one file.	Small AOIs, single-CRS scenes, quick mosaics.
`grid`	Reproject each Major TOM cell to its local UTM geobox and write one file per cell.	Multi-sensor stacking, ML patches, large AOIs.
omitted	Write in the native projection and still intersect with the grid.	When you want the original sensor geometry.

Configuring reprojection¶

name: sentinel2_demo
grid_dist: 10000
output_uri: /tmp/aereo_demo
resolution: 10.0
margin: 0.0
reproject:
  _target_: aereo.builtins.reproject.reproject_odc
  crs: EPSG:32633
  resolution: 10.0
reproject_mode: raw
read:
  _target_: aereo.builtins.read.read_odc_stac
write:
  _target_: aereo.builtins.write.write_geotiff

In pure Python:

from aereo.builtins import reproject_odc
from aereo.pipeline import ExtractionJob

job = ExtractionJob(
    name="demo",
    grid_dist=10_000,
    output_uri="/tmp/demo",
    read=read_odc_stac,
    reproject=reproject_odc,
    reproject_mode="grid",
    write=write_geotiff,
    target_aoi=aoi,
)

Resolution and margin¶

resolution — target pixel size in metres. Used by the grid builder and by reprojectors that accept it.
margin / crop_buffer — extra buffer around cells or scenes to avoid edge effects.
grid_cells_margin — additional margin used when intersecting cells with the AOI.

Swath data¶

Sensors like VIIRS and Sentinel-3 are often stored as swaths (2-D lat/lon arrays). For these data you usually need the built-in reproject_swath helper, which uses pyresample under the hood. Install it with the swath extra:

uv add aereo[swath]
# or
pip install aereo[swath]

See the VIIRS and Sentinel-3 tutorials, and the Configuration reference for all reproject and reproject_mode fields.