CNES Research and Innovation Project

The main goal is to create an "impact map" containing information about wildfires, such as the affected area (crops, forests, etc.) and the impact of the ensuing smoke using the wind's direction.

---

Introduction

This project was our M.Eng's Innovation and Research Project at INSA Toulouse, in partnership with the French space agency CNES (National Centre for Space Studies, in French).

This project consisted in detecting, quantifying, and studying the impact of wildfires. To do this, we proceeded in three main steps:

1. First, we retrieved images from the affected area using Sentinel's mission's images and with using its Python API (details here and here).
2. Then, using an index called NDVI¹, which "measures the density of green on an area of land" (Weier and Herring, 2000), we were able to highlight the affected area by measuring the difference in NDVI from before and after the wildfire.
3. The next step consisted in finding out what type of land cover was burned using different datasets, such as Copernicus CORINE Land Cover, or MODIS Land Cover.
4. Finally, by retrieving the wind data from the day of the wildfire, we wanted to analyze the areas that were affected by the ensuing smoke.

The final objective was to automate this process as much as possible, where we only have some information of the wildfire: a pair of coordinates from where it happened, the date, and the true area that burned (our results regarding the computation of the burnt area are shown in the program).

¹Normalized difference vegetation index

---

Obtaining the code

Downloading the code

In order to execute the code in this repository, you need to install Anaconda and git.

Next, clone this project by typing the following commands in a terminal:

> git clone https://github.com/jayalabanda/projet-CNES.git
> cd projet-CNES

Alternatively, you can download the repository as a ZIP file by clicking here, then unzipping the files into the directory of your choice.

Then, run these commands to create a new environment called "cnes":

> conda env create -f environment.yml
> conda activate cnes

Note: if you want to install the packages manually, it is highly recommended to install them using Anaconda's conda command and conda-forge channel instead of pip, since using both package managers can produce many errors.

Executing the code

There are two ways of executing the code. You can type:

> python main.py --fire_name <name_of_fire>

in a console, which will launch the application on the command line. The command line argument is optional, but it speeds up the process. For more information, refer to the next section ("Wildfire information).

You can also open the notebook main.ipynb in Jupyter Notebook by executing the command:

> jupyter notebook

and navigating to the folder in the tab that opens in your browser.

---

Setup

Wildfire information

Some previous information is needed to execute the code in the notebook or scripts. This information includes:

• the date of the wildfire in format YYYY-MM-DD,
• a pair of coordinates latitude and longitude inside the wildfire, such as $44.5, 4.0$ (more information in INSTRUCTIONS.md),
• the true area that burned in hectares. This value is easily retrievable in the news.

This information must be saved in a JSON file saved as info_NAME.json inside data/info_fires/:

{
    "wildifire_date": "YYYY-MM-DD",
    "longitude": XX.XX,
    "latitude": YY.YY,
    "true_area": ZZ.ZZ
}

where the latitude, longitude, and true area are floats (without the ""). An example is:

{
    "name": "Var",
    "wildfire_date": "2021-08-17",
    "latitude": 43.314677,
    "longitude": 6.451894,
    "true_area": 7100.0
}

Note: the name field is not necessary but is given for clarification, and the date must be in the same format.

A good place to start is the Prométhée list of wildfires website, which gives detailed information about these events, up until the end of 2020 and in the French mediterranean area. If you want global data for 2021, you can also visit the European Forest Fire Information System (EFFIS) website and go to "Current Situation Viewer".

Sentinel API

Moreover, you will need a free Copernicus Open Access Hub account. More details on the Sentinel API are available here. Then save your credentials in a JSON file inside the secrets folder named sentinel_api_credentials.json as follows ("" included):

{
    "username": "YOUR_USERNAME",
    "password": "YOUR_PASSWORD"
}

Google Earth Engine API

In order to use Google's Earth Engine API, you must create a (free) account at Earth Engine API and then authenticate once. Your token will be saved so you will not need to do it again anymore.

To install it, use the command

> conda install -c conda-forge earthengine-api

Climate Data Store API

You will also need a free CDS account to retrieve wind data. Instructions are given here: How to use the CDS API. Remember to install the CDS API client using conda:

> conda install -c conda-forge cdsapi

instead of using pip if it was not installed during the creation of the conda environment.

---

File structure

The file structure of the repository is, ideally, as follows:

+---data
|   +---coordinates_files
|   +---geojson_files
|   +---info_fires
|   +---nc_files
+---output
+---scripts
+---secrets
+---utils

where:

• the downloaded images and land cover dataframes are in data,
• the produced TIFF images, maps, and plots will be stored inside output/,
• various Python scripts are inside the folder of the same name scripts,
• the utility functions are located inside utils,
• wind data files stored as NetCDF (.nc) are located in the nc_files folder,
• and your access credentials for the Sentinel API are inside secrets.

Downloaded images

The first step is obtaining the images. The downloaded folders can vary in size (up to 1.2 GB each), and the produced TIFF files are also around 1 GB each, so please keep this in mind if you have low storage left in your drive. Although, if you are only interested in using the NDVI, you may delete the folders R20m and R60m containing the JP2 image files.

Moreover, as the downloaded data are large, the first step of the process may take a long time, depending on your download speed and the server's upload speed.

Coordinates files

This folder will contain any CSV files created during the execution of the program. These files will contain pairs of latitude and longitude coordinates of the affected areas which you can usein QGIS, for example. This website provides a good explanation on how to obtain similar results using QGIS.

GeoJSON files

The GeoJSON file format is a special type of file storing geospatial data, in a similar format to JSON files. More in-depth information is available at ArcGIS Online.

A step-by-step guide to creating a GeoJSON file for an area of interest is given inside the instructions guide.

---

Contact details

If you have any questions or encounter any problems, please contact me at my email: [email protected], or open a pull request.

---

To-do list:

• [ ] create convex hull figures from fire coordinates
• continue working on main.py
• clean and test new scripts inside utils folder
• make the process as automatic as possible