# Hotspot disambiguation dataset

This repository contains the dataset assembled as part of a work **A Multimodal Supervised Machine Learning Approach for Satellite-based Wildfire Identification in Europe**. Tha paper has been presented at the International Geoscience and Remote Sensing Symposium (**IGARSS**) 2023. 
The full paper is available at https://arxiv.org/abs/2308.02508 . 

The data folder contains two files:
- `dataset.csv`: this file contains the full cross-referenced dataset, obtained by conducing a temporal and spatial data intersection between the EFFIS burned areas and the MODIS/VIIRS hotspots.
- `dataset_500.csv`: this file contains a subset of the previous dataset (~500k data points), subsampled to obtain a dataset stratified with respect to the spatial distribution, and with a positive-negative proportion of 10%-90%. In addition to MODIS/VIIRS data points, additional columns have been added to improve the models' performances. This file is the one used to obtain the results showed in the paper. 

## Code
The code and models used in this work are available at https://github.com/links-ads/hotspot-disambiguation .

## Contributions
- Angelica Urbanelli (angelica.urbanelli@linksfoundation.com)
- Luca Barco (luca.barco@linksfoundation.com)
- Edoardo Arnaudo (edoardo.arnaudo@polito.it | linksfoundation.com)
- Claudio Rossi (claudio.rossi@linksfoundation.com)

## BibTex

```
@inproceedings{urbanelli2023hotspot,
  title={A Multimodal Supervised Machine Learning Approach for Satellite-based Wildfire Identification in Europe},
  author={Urbanelli, Angelica and Barco, Luca and Arnaudo, Edoardo and Rossi, Claudio},
  booktitle={2023 IEEE International Geoscience and Remote Sensing Symposium, IGARSS},
  year={2023}
} 
```

## Licence
cc-by-4.0

## Acknowledgments
This work was carried out in the context of two H2020 projects: SAFERS (GA n.869353) and OVERWATCH (GA n.101082320), and presented at IGARSS 2023.