Operational snow cover product from Sentinel-2 and Landsat-8 THEIA Level 2A products – README.md

README.md

Let-it-snow

Synopsis

This code implements the snow cover extent detection algorithm LIS (Let It Snow) for Sentinel-2, Landsat-8 and SPOT4-Take5 data.

The algorithm documentation with examples is available here:

  • DOI

Access to Theia Snow data collection:

  • DOI:10.24400/329360/f7q52mnk

To read more about the "Centre d'Expertise Scientifique surface enneigée" (in French):

The input files are Sentinel-2 or Landsat-8 level-2A products from the Theia Land Data Centre or SPOT-4/5 Take 5 level-2A products and a Digital Terrain Model (DTM) like SRTM for instance reprojected at the same resolution as the input image.

Usage

Run the python script run_snow_detector.py with a json configuration file as unique argument:

python run_snow_detector.py param.json

The snow detection is performed in the Python script run_snow_detector.py.

All the parameters of the algorithm, paths to input and output data are stored in the json file. See the provided example param_test_s2_template.json file for an example.

Moreover The JSON schema is available in the Algorithm Theoretical Basis Documentation and gives more information about the roles of these parameters.

NB: To build DEM data download the SRTM files corresponding to the study area and build the .vrt using gdalbuildvrt. Edit config.json file to activate preprocessing : Set "preprocessing" to true and set the vrt path.

Products format

  • COMPO: Raster image showing the outlines of the cloud (including cloud shadow) and snow masks drawn on the RGB composition of the L2A image (bands SWIR/Red/Green).
  • SNOW_ALL: Binary mask of snow and clouds.
    • 1st bit: Snow mask after pass1
    • 2nd bit: Snow mask after pass2
    • 3rd bit: Clouds detected at pass0
    • 4th bit: Clouds refined at pass0
    • 5th bit: Clouds initial (all_cloud)
    • 6th bit: Slope flag (optional 1: bad slope correction)

For example if you want to get the snow from pass1 and clouds detected from pass1 you need to do:

pixel_value & 00000101
  • SEB: Raster image of the snow mask and cloud mask.
    • 0: No-snow
    • 100: Snow
    • 205: Cloud including cloud shadow
    • 254: No data
  • SEB_VEC: Vector image of the snow mask and cloud mask. Two fields of information are embedded in this product. DN (for Data Neige) and type.
    • DN field :
      • 0: No-snow
      • 100: Snow
      • 205: Cloud including cloud shadow
      • 254: No data

Data set example

Sequence of snow maps produced from Sentinel-2 type of observations (SPOT-5 Take 5) over the Deux Alpes and Alpe d'Huez ski resorts are available on Zenodo.

Motivation

Code to generate the snow cover extent product on Theia platform.

Installation

LIS processing chain uses CMake (http://www.cmake.org) for building from source.

Dependencies

Following a summary of the required dependencies:

  • GDAL >=2.0
  • OTB >= 6.2
  • Python interpreter >= 2.7
  • Python libs >= 2.7
  • Python packages:
  • numpy
  • lxml
  • matplotlib

GDAL itself depends on a number of other libraries provided by most major operating systems and also depends on the non standard GEOS and Proj libraries. GDAL- Python bindings are also required

Python package dependencies:

  • sys
  • subprocess
  • glob
  • os
  • json
  • gdal

Optional dependencies:

  • gdal_trace_outline can be used alternatively to gdal_polygonize.py to generate the vector layer. It requires to install dans-gdal-scripts utilities.

Installing from the source distribution

General

In your build directory, use cmake to configure your build.

cmake -C config.cmake source_lis_path

In your config.cmake you need to set :

LIS_DATA_ROOT

For OTB superbuild users these cmake variables need to be set:

OTB_DIR
ITK_DIR
GDAL_INCLUDE_DIR
GDAL_LIBRARY

Run make in your build folder.

make

To install let-it-snow application and the s2snow python module. In your build folder:

make install

Add appropriate executable rights

chmod -R 755 ${install_dir}

The files will be installed by default into /usr/local and add to the python default modules. To overrsouride this behavior, the variable CMAKE_INSTALL_PREFIX must be configure before build step.

Update environment variables for LIS. Make sure that OTB and other dependencies directories are set in your environment variables:

export PATH=/your/install/directory/bin:/your/install/directory/app:$PATH
export LD_LIBRARY_PATH=/your/install/directory/lib:$LD_LIBRARY_PATH
export OTB_APPLICATION_PATH=/your/install/directory/lib:$OTB_APPLICATION_PATH
export PYTHONPATH=/your/install/directory/lib:/your/install/directory/lib/python2.7/site-packages:$PYTHONPATH

let-it-snow is now installed.

Tests

Enable tests with BUILD_TESTING cmake option. Use ctest command to run tests. Do not forget to clean your output test directory when you run a new set of tests.

Data (input and baseline) to run validation tests are available on Zenodo:

  • DOI

Download LIS-Data and extract the folder. It contains all the data needed to run tests. Set Data-LIS path var in cmake configuration files. Baseline : Baseline data folder. It contains output files of S2Snow that have been reviewed and validated. Data-Test : Test data folder needed to run tests. It contains Landsat, Take5 and SRTM data. Output-Test : Temporary output tests folder. Do not modify these folders.

Contributors

Manuel Grizonnet (CNES), Simon Gascoin (CNRS/CESBIO), Tristan Klempka (CNES), Germain Salgues (Magellium)

License

This is free software under the GNU Affero General Public License v3.0. See http://www.gnu.org/licenses/agpl.html for details.