Overview
Operationalizing and institutionalizing innovative technologies for using geospatial data is key to producing accurate, transparent information critical for reversing climate change, reducing deforestation and degradation, and catalysing restoration and conservation. The first phase of SEPAL was successfully carried out from 2016 to 2021, implemented by the Forestry Division of the Food and Agriculture Organization of the United Nations (FAO) and financed by an initial contribution from Norway’s International Climate and Forest Initiative (NICFI). The second phase – Forest and Land Monitoring for Climate Action – is a four-year project that aims to improve climate change mitigation plans and create better informed land-use policies.
It is connected to modern geospatial data infrastructures, such as Google Earth Engine (GEE), further driving the generation of high-integrity forest and land-use data that enable countries to attract public and private carbon finance for forest-related climate action. SEPAL caters to specific monitoring and reporting needs through dedicated modules, such as SE.PLAN for spatially explicit forest restoration planning, SEPAFE to receive and validate fire alerts, and BFAST-GPU for running dense, large-area, time-series analyses.
SEPAL’s goals are to support countries in their efforts to halt tropical deforestation and degradation, and to enhance reforestation and restoration by:
- providing a free, easily accessible, powerful platform for efficient geospatial data access and customized processing;
- enhancing global capacity to operationally apply high-resolution satellite images to critical forest and land monitoring issues; and
- building country capacity in high-resolution satellite imagery applications for critical forest and land monitoring issues in response to domestic policy needs.
The project will directly benefit countries through different capacity-building modalities toward autonomous use of the data for key forest and land monitoring needs. In addition, some countries will receive targeted support and commitment, leading to the institutionalization of data use at the national level, thus strengthening institutional arrangements and linkages to policy formulation, implementation and evaluation through capacity development at the national level
Key Features
Extensive Data Access: Provides access to vast archives of satellite imagery (Landsat, Sentinel-2, etc.).- Integrated Geospatial Tools: Combines various open-source geospatial libraries and tools (e.g., GDAL, ORFEO Toolbox, SNAP Toolkit) within a unified environment.
- Multiple Scripting Environments: Supports geoprocessing via Jupyter Notebooks, JavaScript, Python, and R.
- User-Friendly Interface: Offers both code-based and graphical user interface (GUI) options for various workflows (e.g., mosaic generation, image classification, change detection).
- Capacity Building: Designed to support technical capacity development in land monitoring and remote sensing.
- Collaboration: Facilitates collaborative work and sharing of scripts and results.
SEPAL Architecture
& Worklflow
- Data Discovery: Users log into SEPAL and can browse a vast catalog of satellite imagery (Landsat, Sentinel) and other geospatial datasets available directly within the platform.
- Area of Interest (AOI) Definition: Users define their specific area of interest (e.g., country, region, custom polygon) on a map, which guides subsequent data access and processing.
- Time Series & Sensor Selection: Users select the desired time period for their analysis and choose specific satellite sensors, allowing for multi-temporal studies like change detection.
- Mosaic Generation: Users can generate cloud-free, composite mosaics from selected satellite imagery. This often involves atmospheric correction and stitching multiple scenes to create a seamless image for the AOI.
- Integrated Development Environments (IDEs): SEPAL provides hosted Jupyter Notebooks (for Python) and RStudio Server (for R). Users can write and execute custom code scripts leveraging powerful geospatial libraries to perform complex analyses.
- Pre-built Modules/Recipes: For users with less coding experience, SEPAL offers a suite of pre-built "Recipes" or "Modules." These are automated workflows (e.g., land use/land cover classification using machine learning, deforestation detection, biomass estimation) that can be executed through a user-friendly graphical interface.
- Integration with Google Earth Engine (GEE): For computationally intensive tasks involving large data volumes or global coverage, SEPAL transparently offloads processing to GEE, leveraging its massive parallel computing capabilities.
- Execution of Open-Source Tools: SEPAL provides access to a range of command-line geospatial tools (e.g., GDAL, ORFEO Toolbox, SNAP) that can be executed within its cloud environment, allowing for specialized processing not directly covered by GEE or standard libraries.
- Visualization: Processed outputs (e.g., classified maps, change maps, statistical charts) can be visualized directly within the SEPAL platform's interactive dashboards.
- Data Export: Users can download their analysis results in various standard geospatial and tabular formats (e.g., GeoTIFF for raster images, Shapefile/GeoJSON for vector data, CSV for tabular data) for further use in external GIS software or reporting tools.
- Reporting: Derived products and statistics from SEPAL analyses contribute directly to national reports for international frameworks like the UNFCCC (e.g., REDD+ reporting), as well as for national land-use planning and policy development.
- Collaboration & Sharing: Users can share their analysis scripts, input data, and processed results with other SEPAL users, fostering collaborative research and monitoring efforts among teams or across institutions.
Technology Stack
-
Frontend Technologies:
-
Web Interface: Primarily developed using JavaScript. The interactive dashboard elements often utilize ipyvuetify (a Python wrapper for the Vue.js-based Vuetify component framework), for a modern, component-based UI.
-
Interactive Environments: Jupyter Notebooks (for Python development) and RStudio Server (for R development) provide the integrated coding environments, accessed through the web browser.
-
-
Backend Technologies:
-
Core Logic: The main SEPAL backend (
openforis/sepalrepository) is largely built with JavaScript (likely Node.js) and extensively uses Python for scripting and integrating with geospatial libraries and services. -
Cloud Infrastructure: Hosted by FAO, it leverages various cloud-based services. Given the Google Earth Engine integration, Google Cloud Platform (GCP) services are heavily implied for underlying infrastructure, storage, and possibly serverless functions.
-
Processing Engines:
-
Google Earth Engine (GEE): The core engine for global-scale geospatial computation.
-
Python Libraries: A vast array of Python libraries for geospatial data processing and scientific computing, including
GDAL,Rasterio,Fiona,numpy,scipy,scikit-learn,earthengine-api,xarray, and more. -
R Libraries: Numerous R packages for statistical analysis and geospatial tasks within RStudio Server.
-
Specialized Geospatial Tools: Integration of command-line tools like GDAL, ORFEO Toolbox (OTB), and SNAP Toolkit (for SAR data processing).
-
Shiny: Used for developing interactive web applications (R Shiny apps) within SEPAL, often for specific analysis workflows or dashboards.
-
-
Containerization: Container technologies (e.g., Docker) to manage and deploy the various processing tools and environments consistently and efficiently.
-
Data Storage: Cloud-based storage solutions for vast satellite imagery archives (accessed via GEE or directly) and user-specific data/scripts.
-
Key Integration Features
Google Earth Engine Access: Users can connect their Google accounts to access Earth Engine's multi-petabyte satellite imagery catalog and global-scale analysis capabilities without coding.
Real-time Collaboration: The system uses WebSocket communication for real-time updates on file and asset changes, enabling collaborative workflows .
Multi-language Support: The interface supports 9 languages including English, French, Spanish, Chinese, Russian, Arabic, Swedish, Portuguese, and Italian .
