Skip to content

darts_acquisition.load_s2_masks

Load the valid and quality data masks from a Sentinel 2 scene.

Parameters:

  • fpath (str | pathlib.Path) –

    The path to the directory containing the TIFF files.

  • reference_geobox (odc.geo.geobox.GeoBox) –

    The reference geobox to reproject, resample and crop the masks data to.

Returns:

  • xarray.Dataset –

    xr.Dataset: A merged xarray Dataset containing two data masks: - 'valid_data_mask': A mask indicating valid (1) and no data (0). - 'quality_data_mask': A mask indicating high quality (1) and low quality (0).

Source code in darts-acquisition/src/darts_acquisition/s2.py 
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
@stopuhr.funkuhr("Loading Sentinel 2 masks", printer=logger.debug, print_kwargs=["fpath"])
def load_s2_masks(fpath: str | Path, reference_geobox: GeoBox) -> xr.Dataset:
    """Load the valid and quality data masks from a Sentinel 2 scene.

    Args:
        fpath (str | Path): The path to the directory containing the TIFF files.
        reference_geobox (GeoBox): The reference geobox to reproject, resample and crop the masks data to.


    Returns:
        xr.Dataset: A merged xarray Dataset containing two data masks:
            - 'valid_data_mask': A mask indicating valid (1) and no data (0).
            - 'quality_data_mask': A mask indicating high quality (1) and low quality (0).

    """
    # Convert to Path object if a string is provided
    fpath = fpath if isinstance(fpath, Path) else Path(fpath)

    logger.debug(f"Loading data masks from {fpath.resolve()}")

    # TODO: SCL band in SR file
    try:
        scl_path = next(fpath.glob("*_SCL*.tif"))
    except StopIteration:
        logger.warning("Found no data quality mask (SCL). No masking will occur.")
        valid_data_mask = (odc.geo.xr.xr_zeros(reference_geobox, dtype="uint8") + 1).to_dataset(name="valid_data_mask")
        valid_data_mask.attrs = {"data_source": "s2", "long_name": "Valid Data Mask"}
        quality_data_mask = odc.geo.xr.xr_zeros(reference_geobox, dtype="uint8").to_dataset(name="quality_data_mask")
        quality_data_mask.attrs = {"data_source": "s2", "long_name": "Quality Data Mask"}
        qa_ds = xr.merge([valid_data_mask, quality_data_mask])
        return qa_ds

    # See scene classes here: https://custom-scripts.sentinel-hub.com/custom-scripts/sentinel-2/scene-classification/
    da_scl = xr.open_dataarray(scl_path)

    da_scl = da_scl.odc.reproject(reference_geobox, sampling="nearest")

    # Match crs
    da_scl = da_scl.rio.write_crs(reference_geobox.crs)

    da_scl = xr.Dataset({"scl": da_scl.sel(band=1).fillna(0).drop_vars("band").astype("uint8")})
    da_scl = convert_masks(da_scl)

    return da_scl