DailyWildfireCAN: A Dataset and Benchmark for Next-Day Wildfire Spread Prediction in Canada
Organizations: Mila - Quebec AI Institute, 6666 Rue Saint-Urbain, Montreal, QC, H2S 3H1, Canada · Computer and Software Engineering Department, Polytechnique Montreal, 2500 Chemin de Polytechnique, Montreal,QC, H3T 1J4, Canada
Abstract
Wildfires represent a critical ecological and socio-economic challenge in Canada, where vast boreal forests and a warming climate have amplified the frequency and severity of extreme fire seasons. Predicting next-day wildfire spread can support fire management, but requires capturing complex interactions among weather, fuels, and terrain. Deep learning has emerged as a promising approach to learning these relationships from heterogeneous environmental data, yet its development and evaluation depend on representative datasets. Existing deep learning research on wildfire spread relies predominantly on US-focused datasets, leaving Canadian fire environments underrepresented. To address this gap, we present DailyWildfireCAN, a multi-modal, multi-temporal dataset and benchmark for predicting next-day wildfire spread across Canada. The dataset tracks continuous fire progression over multi-day events at a high spatial and temporal resolution of 90-m and 24-hour, respectively. This resolution preserves fine-scale information relevant to fire-edge geometry, local fuel variations, and terrain effects, supporting research toward improved daily wildfire spread forecasting. DailyWildfireCAN covers 2,440 wildfire events across eight fire seasons 2017-2024 and integrates 33 features across diverse modalities, including topography, meteorology, land cover, vegetation and forest structure, human influence indices, and satellite imagery. We establish a comprehensive benchmark comparing mono-temporal and multi-temporal prediction models, accompanied by a multi-dimensional evaluation framework assessing robustness across wildfire characteristics, geographic regions, and temporal conditions, spatial prediction quality, and sensitivity to reference label variations.
Figures & tables
| 2D Env. inputs | ||||||||||||
| Dataset a | Region | Period | # Fires | Grid | Task | Spectral inputs | W | T | V | F | D | H |
| Next Day Wildfire Spread | USA | 2012–2020 | — | 1 km | Active-fire prediction (1-day) | VIIRS index | ✓ | ✓ | ✓ | ✗ | ✗ | ✓ |
| WildfireSpreadTS | USA | 2018–2021 | 607 | 375 m | Active-fire prediction (1-day) | VIIRS bands, indices | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ |
| TS-SatFire (progression) | USA | 2017–2021 | 179 | 375 m | Burned-area growth (1-day) | VIIRS bands, indices | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ |
| CanadaFire2023 | Canada | 2023 | — | 10/30 m b | Burned-area mapping | Landsat-8/9, Sentinel-2 bands, indices | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ |
| WildfireCube | USA, Canada, Greece | 2017–2023 | 13 | 30 m | Fire progression (3-hour) | Sentinel-2 indices | ✓ | ✓ | ✓ | ✗ | ✗ | ✗ |
| Feature Group | Source | Features | Native Resolution | Temporal Frequency |
| Environmental Features | ||||
| Topography (elevation, slope, aspect) | ASTER DEM ( Abrams et al., 2020 ) | 3 | 30 m | Static |
| Weather (temperature, humidity, precipitation, wind (u, v)) | ERA5-Land ( Muñoz-Sabater et al., 2021 ) | 6 | 9 km | Hourly |
| Vegetation indices (NDVI, EVI) | VIIRS ( Didan, 2021 ) | 2 | 500 m | 16-day |
| Forest structure and composition | SCANFI ( Guindon et al., 2024 ) | 13 | 30 m | Static (5 y) |
| Land cover | CCRS ( Latifovic et al., 2017 ) | 1 | 30 m | Static (5 y) |
| Split | # Samples | # Fires | Time Span | Spatial Coverage | % |
| Train | 2017–2023 | 7 clusters | |||
| Validation | 2017–2023 | 1 cluster | |||
| (1) Test Easy | 2024 | 8 clusters | |||
| (2) Test Medium | 2017–2023 | 2 clusters | |||
| (3) Test Hard | 2024 | 2 clusters | |||
| Total | 2017–2024 | 10 clusters |
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
| Data Source | License |
| ASTER DEM ( Abrams et al., 2020 ) | CC0 |
| ERA5-Land ( Muñoz-Sabater et al., 2021 ) | CC BY 4.0 |
| VIIRS ( Didan, 2021 ) | CC0 |
| SCANFI ( Guindon et al., 2024 ) | Open Government Licence – Canada 2.0 |
| Land Cover ( Latifovic et al., 2017 ) | Open Government Licence – Canada 2.0 |
| CanLaD ( Perbet et al., 2025 ) | Open Government Licence – Canada 2.0 |
Wildfire Forecasting
Wildfire forecasting predicts where fires will start, how they will spread, and their likely intensity or impacts using environmental and fire observations.
See all 30 papers