sits is a high-level Python package designed to simplify the extraction and processing of Satellite Image Time Series (SITS) referenced in STAC catalogs. For any given point or polygon, it efficiently handles data retrieval and, leveraging spyndex, can calculate a wide array of spectral indices. The processed results can be exported in various formats, including image files, CSV tables, or dynamic animated GIFs, with customizable dimensions suitable for applications such as deep learning. In addition to its core functionalities, the package includes an experimental analysis module that integrates forecasting methods from the sktime library. This module enables users to apply time series models to satellite-derived data, opening possibilities for predictive analytics and temporal pattern exploration.

GitHub: https://github.com/kenoz/SITS_utils

Documentation: https://sits.readthedocs.io/

PyPI: https://pypi.org/project/sits/

Tutorials: https://sits.readthedocs.io/en/latest/tutorials.html

Use the package manager pip to install sits.

pip install sits

Here is a basic Python script example. For more details, read the documentation here.

from sits import sits

# loads csv table with geographic coordinates into GeoDataFrame object
csv_file = 'my_file.csv'

# instantiates a SITS.Csv2gdf object
sits_df = sits.Csv2gdf(csv_file, 'lon', 'lat', 4326)

# converts coordinates of sits_df into EPSG:3035 
sits_df.set_gdf(3035)

# calculates buffer with a radius of 100 m for each feature.
sits_df.set_buffer('gdf', 100)

# calculates the boundiug box for each buffered feature.
sits_df.set_bbox('buffer')

# exports geometries as a GIS vector file
sits_df.to_vector('bbox', 'output/my_file_bbox.geojson', driver='GeoJSON')

# gets Sentinel-2 time-series from STAC catalog

# requests STAC catalog for each geometries of sits_df.bbox
for index, row in sits_df.bbox.iterrows():
    gid = sits_df.bbox.loc[index, 'gid']
    
    row_geom = sits_df.bbox.loc[index, 'geometry']
    row_geom_4326 = sits_df.bbox.to_crs(4326).loc[index, 'geometry']
    
    aoi_bounds = list(row_geom.bounds)
    aoi_bounds_4326 = list(row_geom_4326.bounds)

    # opens access to a STAC provider (by default Microsoft Planetary)
    imgs = sits.StacAttack()
    # searches items based on bbox coordinates and time interval criteria
    imgs.searchItems(aoi_bounds_4326, 
                     date_start=datetime(2016, 1, 1), 
                     date_end=datetime(2019, 12, 31))
    
    # extracts Sentinel-2 metadata and writes in csv file.
    imgs.items_prop["station_id"] = gid
    imgs.items_prop.to_csv(f'output/id_{gid}_s2_metadata.csv')
    
    # loads time-series images in EPSG:3035
    imgs.loadCube(aoi_bounds, arrtype='image', crs_out=3035)
    
    # exports time-series into csv file and netCDF file
    imgs.to_csv(out_dir, gid)
    imgs.to_nc(out_dir, gid)

If you want to explore the different ways to use the sits package, we recommend running the following Jupyter notebooks, in Google Colab for instance:

• Example 01: explain the basics for retrieving a satellite image time series according to a polygon feature.
• Example 02: explain how to export a satellite time series into an animated GIF or video file.
• Example 03: explain how to compute spectral indices with sits and spyndex packages.
• Example 04: explain how to parallelize processing tasks in case of multiple vector features.
• Example 05: explain how to automatically detect forest clear-cuts by using the analysis module.

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

Please make sure to update tests as appropriate.

To ensure the correct behaviour of the module, you can run:

# Or any of the supported python versions
uv run --python 3.10 --group dev pytest tests/

If you want to test all the supported python versions, you can run ./test_all_versions.sh.

If you want an isolated environment, you can use the Dockerfile provided.

To run the matrix test in the isolated environment you can do:

# Build the image
docker build -t sits-gdal .

# Then run the tests with persistent caching
docker run --rm \
  -v uv-cache:/root/.cache/uv \
  sits-test

Or if you want to test live changes, you can Bind Mount your working directory as:

# Run this AFTER building the image
docker run --rm \
  -v uv-cache:/root/.cache/uv \
  -v $(pwd):/app \
  sits-test

Remember that if you make changes to the Dockerfile, you will need to build your image again.

GNU GPL v.3.0