eoio.readers.gistda_theos.data_io module

eoio.readers.gistda_theos.data_io module#

eoio.readers.gistda_theos.data_io - data reading functionality for THEOS meas vars.

exception eoio.readers.gistda_theos.data_io.GISTDATHEOSDataIOError[source]#

Bases: ValueError

eoio.readers.gistda_theos.data_io.georeference_from_gcps(img: DataArray) → tuple[DataArray, Any, str][source]#

Attach a real affine transform + CRS to img from its own embedded ground control points (GISTDA THEOS ships no GDAL geotransform – only GCPs – so rxr.open_rasterio otherwise leaves img’s x/y as bare pixel indices with no CRS at all: verified directly against a real product’s IMAGERY.TIF).

Fits one affine transform across every GCP (rasterio.transform.from_gcps(), a least-squares fit from >= 3 points) – appropriate for a geometrically-corrected Level-1 scene like this, where a single overall rotation/scale (THEOS’s own SCENE_ORIENTATION in METADATA.DIM) explains the corners, not a per-pixel nonlinear distortion; THEOS provides exactly the 4 scene corners, so the fit is over-determined by 2 equations and therefore self-consistency-checked.

Parameters:

img – DataArray from rxr.open_rasterio – GCPs read via rioxarray’s own .rio.get_gcps()/.rio.crs (which read a GeoTIFF’s embedded GDAL GCP tags directly), not reader-specific parsing.

Returns:

(img, transform, gcp_crs) – img with CRS/transform written in place (so rioxarray.raster_array.RasterArray.clip(), and any other rioxarray spatial op, can use them), the fitted affine transform and its CRS (for lonlat_grid_from_gcps() to reuse without refitting).

Raises:

GISTDATHEOSDataIOError – img has no usable ground control points.

eoio.readers.gistda_theos.data_io.get_var_geometry() → str[source]#

Get the variable geometry

Returns:

string of geometry

eoio.readers.gistda_theos.data_io.get_var_resolution() → str[source]#

Get the variable resolution

Returns:

string of resolution

eoio.readers.gistda_theos.data_io.lonlat_grid_from_gcps(transform: Any, *, width: int, height: int) → tuple[ndarray, ndarray][source]#

Full-resolution (height, width) longitude/latitude grids from transform (see georeference_from_gcps()) – a real 2-D (row, col) -> (lon, lat) mapping.

Deliberately not row/col treated as independent 1-D axes (wrong whenever the transform has any rotation term, which GISTDA THEOS scenes generally do – see SCENE_ORIENTATION in METADATA.DIM: verified on a real product with ~10 deg of rotation that independent axes mislabel most pixels by ~0.2 deg / ~20 km).

Parameters:

  • transform – affine transform from georeference_from_gcps().

  • width – image width in pixels.

  • height – image height in pixels.

Returns:

(lon, lat), each shape (height, width).

eoio.readers.gistda_theos.data_io.read_bands_into_dataset(*, ds: Dataset, layout: GISTDATHEOSLayout, meas: list[str], subset: ResolvedROISubset | None, mtd: GISTDATHEOSMetadataExtractor, chunks: dict[str, int] | None = None, use_chunks: bool = False) → Dataset[source]#

Read THEOS image bands into an xarray dataset.