CFSI - Cloud-Free Satellite Imagery

CFSI is a set of Python scripts and tools designed to automatically generate cloudless mosaic images from Sentinel 2 data. It relies heavily on OpenDataCube (ODC) to index Sentinel 2 imagery from Amazon S3 and generate cloudless mosaic images using cloud masks. Cloud detection and masking is done using S2Cloudless. The mosaics are indexed into ODC and can be served over WMS/WCS using datacube-ows.

Configuration

CFSI can be configured using a config.yaml file. A sample configuration is provided in the repository with comments explaining most options. The configuration file syntax is subject to change without notice until a possible stable release in the future.

Deployment

Docker

The easiest way to setup CFSI locally is to use Docker and Docker-Compose. The compose file includes both a backend PostGIS database service and a ODC Python environment.

Dependencies

• Docker
• Docker-Compose
• Python 3.7+ (for using the CLI)

Configuration and setup

First, edit the variables in the .env file to match your system, this applies mainly to the variables ending with _HOST.

Then, start CFSI with:

docker-compose up -d

AWS

A Terraform configuration is also provided to automatically launch and manage CFSI on AWS or other cloud platforms supported by TF.

Dependencies

• Terraform 0.14+

Configuration and setup

Set local environment variables AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY in your profile. To deploy CFSI on AWS, run:

terraform init
terraform apply

Local

The easiest way to set up CFSI locally without containers is to use Conda. Create a new Conda environment using the provided environment.yml file. You also need to set up a PostGIS database.

Dependencies

For a conda-based setup (recommended) using a local database backend:

• Anaconda/Miniconda
• PostgreSQL 12+
• PostGIS 3+ (optional for datacube-ows)

Configuration and setup

Follow the ODC installation guide (using the environment.yml file from this repository) and the database setup guide.

To-Do

The project is still under active development and things may and will change without notice. The to-do list before a possible future stable release includes:

• CLI for common operations such as generating mosaics and indexing new S2 imagery to ODC
• Support for other cloud masking methods besides s2cloudless
• Migrate write operations from GDAL to rasterio
• User testing with different areas and configurations
• Automated test framework and unit tests