Method for checking whether a bbox in a Sentinel 2 scene from the Planetary Computer is cloudy or not

check_if_scene_is_cloudy.py

def check_if_scene_is_cloudy_at_box(item: pystac.Item, box: shapely.geometry.Polygon):
    """Uses the S2 Scene Classification Layer (SCL) to determine if a S2 L2A scene is cloudy at a given bbox.

    Args:
        item (pystac.Item): The S2 L2A item to check
        box (shapely.geometry.Polygon): The geometry to check (should be EPSG:4326)

    Returns:
        float: The fraction of the box that is classified as "Cloud medium probability" + "Cloud high probability"
    """
    scl_url = item.assets["SCL"].href

    with rasterio.open(scl_url) as f:
        geom = shapely.geometry.mapping(box)
        warped_geom = rasterio.warp.transform_geom("EPSG:4326", f.crs, geom)
        out_mask, _ = rasterio.mask.mask(f, [warped_geom], crop=True)
        out_mask = out_mask.squeeze()

    out_mask_pixels = out_mask[out_mask != 0]
    vals, counts = np.unique(out_mask_pixels, return_counts=True)
    N = counts.sum()
    val_counts = dict(zip(vals, counts))
    # Number of "Cloud medium probability" + "Cloud high probability"
    num_cloudy = val_counts.get(8, 0) + val_counts.get(9, 0)

    if N == 0:
        raise WindowEmpty()
    else:
        fraction_bad = num_cloudy / N
        return fraction_bad