S2BAVG - Sentinel-2 Burned Area Validation Grid

Description

The S2BAVG GeoPackage is a database designed for validating burned area (BA) products using Sentinel-2 images. It contains:

• A layer with the geometries of the sampling units and their permanent attributes.
• Annual tables with variable attributes since the availability of Sentinel-2 images.

Main Attributes

• Name: Sampling unit ID.
• orbits: Sentinel-2 orbits that fully cover the unit.
• tile_area: Total area in m².
• land_area: Land area in m².
• land_perc: Percentage of land area over the total area (land_area / tile_area).
• biome: The dominant biome in the unit.
• mcd64a1_area: Total burned area at least once throughout the year according to MCD64A1.
• mcd64a1_burned_perc: Percentage of mcd64a1_area over the land area. (mcd64a1_area/land_area).
• vnp64a1_area: Total burned area at least once throughout the year according to VNP64A1.
• vnp64a1_burned_perc: Percentage of vnp64a1_area over the land area. (vnp64a1_area/land_area).
• firecci51_area: Total burned area at least once throughout the year according to FireCCI51.
• firecci51_burned_perc: Percentage of firecci51_area over the land area. (firecci51_area/land_area).
• viirs_count_year: Total number of VIIRS (VNP14IMGTDL) detections in the year.
• image_days_year: Days with at least one Sentinel-2 image covering at least 99% of the unit.
• median_days_interval_year: Median interval of days between Sentinel-2 images.
• semester: Estimated semester as the main fire season of the year.
• date_pre_fireseason and date_post_fireseason: Estimated start and end dates of the fire season.
• viirs_count_fireseason: Number of VIIRS detections in the fire season.
• image_days_fireseason: Number of Sentinel-2 image days in the fire season.
• image_days_fireseason_nocloudy: Sentinel-2 image days with low cloud cover in the fire season.
• median_days_interval_fireseason_nocloudy: Median interval between cloud-free Sentinel-2 images.

---

Generating Sampling

The 'Sampling.py' script is used to generate the stratified sampling for annual burned area validation. To ensure smooth execution, it is recommended to use a conda environment.

1. Creating and Activating a Conda Environment

To avoid dependency conflicts, follow these steps to create a dedicated environment:

# Create a new conda environment named 's2bavg_env' with Python 3.9
conda create --name s2bavg_env python=3.9 -y

# Activate the environment
conda activate s2bavg_env

# Install required dependencies
conda install -c conda-forge geopandas pandas numpy rasterio rasterstats -y

2. Running the Script

Once the environment is set up, run the script using command-line arguments:

python sampling.py \
    --gpk_path "C:/path/to/S2BAVG.gpkg" \
    --output_sampling_path "C:/path/to/output/sampling \
    --year "2019" \ 
    --ba_data "firecci51" \  # Options: "firecci51", "mcd64a1" or "vnp64a1"
    --total_sample_size 100 \
    --land_perc_filter 50 \
    --nocloudy_interval_filter 10

3. Output Files

The script generates a GeoPackage output file containing three main layers:

1. stratification_layer: Stratification layer based on biome and fire activity.
2. filtered_layer: Total population after cloud cover and land area filtering.
3. sampling_layer: Final selection of sampled units.

An aditional csv file is generated storing the population size (Nh) and the sample size (nh) per stratum. This file is needed for the statistical inference.

The output file will be stored at the location specified in output_sampling_path.

4. Deactivating the Conda Environment

After execution, you can deactivate the environment with:

conda deactivate

---

Additional Notes

• Customization: You can adjust the filtering criteria (land_perc_filter, nocloudy_interval_filter) and total_sample_size as needed.
• Compatibility: The script is designed for Python 3.9 and later versions.
• Validation: You can open the sampling.gpkg file in GIS software like QGIS or ArcGIS to visualize results.

---

Citation

If you use this repository, please cite the following article:

Gonzalez-Ibarzabal, J., Bastarrika, A., Stehman, S. V., Stroppiana, D., and Franquesa, M. (2025).
S2BAVG: A global Sentinel-2 grid for burned area product validation.
International Journal of Applied Earth Observation and Geoinformation, 144, 104889.
https://doi.org/10.1016/j.jag.2025.104889