Skip to content

Matching with available Sentinel-2 data

This step aims at matching observation data with available Sentinel-2 tiles or polygons to Sentinel-2 tiles so it can be used to extract the matching Sentinel-2 data using the extraction function. If polygons are used, they are converted to grid points located at the centroid of 10m Sentinel-2 pixels.

If points or polygons intersect several Sentinel-2 tiles, the resulting points are duplicated for each of them. Observation polygons which are not contained in a Sentinel-2 tile, or are too small to contain a pixel centroid are removed and their IDs are printed to the console.

INPUTS

The input parameters are :

  • obs_path : Path to a vector file containing observation points or polygons, must have an ID column corresponding to name_column parameter.
  • sentinel_dir : Path of the directory containing Sentinel-2 data.
  • export_path : Path used to write resulting vector file, with added epsg, area_name and id_pixel columns.
  • name_column (optional) : Name of the ID column. The default is id.
  • list_tiles (optional) : A list of names of Sentinel-2 directories. If this parameter is used, extraction is limited to those directories.
  • overwrite (optional) : If True, allows overwriting of file at obs_path. The default is False.

OUTPUT

The output is a vector file at export_path containing points for each 10m Sentinel-2 pixel of available tiles at the location of points, or contained in the polygons in the vector file at obs_path. The following attributes are added :

  • area_name : The name of the matching Sentinel-2 tile, parsed from its directory name in sentinel_dir.
  • epsg : The CRS of the matching Sentinel-2 tile
  • id_pixel : The ID of the pixel whose centroid is at the point location in 10m Sentinel data in the corresponding epsg. id_pixel goes from 1 to the number of pixels in the observation.

NOTE : Since observation polygons can be in overlapping Sentinel-2 tiles with different CRS. One needs to use name_column and id_pixel as well as epsg to identify a unique point location (and therefore Sentinel-2 pixel).

How to use

From a script

from fordead.validation.obs_to_s2_grid import obs_to_s2_grid

obs_to_s2_grid(obs_path = <obs_path>, export_path = <export_path>, name_column = <name_column>)

From the command line

fordead obs_to_s2_grid [OPTIONS]

See detailed documentation on the site

How it works

Importing the vector file containing observations

The vector file at obs_path is imported using the geopandas package.

Creating a Sentinel-2 tiles extent vector from existing Sentinel-2 data

  • Available Sentinel-2 tiles in the sentinel_dir directory are listed
  • The extent of each listed Sentinel-2 tiles is extracted and converted to a GeoDataFrame
  • Each GeoDataFrame is given the attributes area_name and epsg, corresponding to the name of the directory containing the tile data, and the epsg of the tile.
  • All GeoDataFrames are concatenated

    Function used: get_polygons_from_sentinel_dirs()

If vector file at obs_path contains points :

  • Observation points are intersected with the Sentinel-2 tiles extent vector, transferring the attributes area_name and epsg
  • An id_pixel column is added and filled with 0 so the resulting vector can be used in the export_reflectance function.
  • Points outside of available Sentinel-2 tiles are detected and their IDs are printed.

    Function used: get_sen_intersection_points()

If vector file at obs_path contains polygons :

Matching observation polygons with Sentinel-2 tiles

  • Observation polygons are overlaid with the Sentinel-2 tiles extent vector, transferring the 'area_name' and 'epsg' columns corresponding to the name of the tile, and the projection system respectively
  • Observation polygons which are not contained in a Sentinel-2 tile are removed, their IDs are printed to the console.

    Function used: get_sen_intersection()

Generating points for pixels inside the polygons

  • For each polygon, points are generated in a grid corresponding to the centroids of Sentinel-2 pixels inside the polygon.
  • They are given the attributes
  • Polygons with no pixels centroids inside of them have their IDs printed to the console.

    Function used: polygons_to_grid_points()

Exporting the resulting the vector file

The resulting points are exported to export_path.