Skip to content
FireSat-AI: Wildfire Risk Forecasting
Aug 2026

FireSat-AI: Wildfire Risk Forecasting

CNN-LSTM + attention model forecasting wildfire risk from satellite imagery

01. The Challenge

Wildfire risk models are frequently opaque and rarely disclose where they actually fail, making it hard to trust them for real decisions.

02. The Solution

Built an end-to-end pipeline fusing Sentinel-1/2, Landsat, and MODIS satellite imagery with ERA5 reanalysis weather through a CNN + attention encoder feeding a BiLSTM, producing interpretable 1/3/6-month wildfire risk classifications for two fire-active Alaska regions.


Overview

FireSat-AI is a GSoC-style MVP for the "Alaska Wildfire Prediction Using Satellite Imagery" proposal: a hybrid CNN-LSTM + attention model that fuses multi-source satellite imagery with weather reanalysis data to forecast wildfire risk at 1, 3, and 6 month horizons.


Architecture

Monthly satellite feature stacks pass through a ResNet-style CNN with squeeze-excite channel attention, producing a spatial embedding per month. These are concatenated with ERA5 weather features and fed through a bidirectional LSTM over a 24-month lookback, with additive temporal attention producing the final context vector for three multi-horizon classification heads.


Interpretability by Design

Both the channel attention and temporal attention are surfaced directly in the API and dashboard, alongside gradient x input saliency maps, so a predicted risk level always comes with an explanation of which regions and months drove it.


Honesty Notes

The shipped demo ships with a physically-motivated synthetic dataset (real Alaska climate normals, seasonal vegetation cycles) rather than live satellite pulls, and the evaluation report plainly states where the demo checkpoint does and doesn't beat a naive majority-class baseline — reported rather than hidden.


System Architecture

ResNet-style CNN encoder with squeeze-excite channel attention over monthly satellite feature stacks (NDVI, NBR, SAR, fuel moisture)
Bidirectional LSTM over a 24-month lookback, fused with ERA5 weather features
Additive temporal attention feeding three multi-horizon classification heads (No Risk / Moderate / High)
FastAPI backend + Leaflet dashboard exposing risk, attention, and historical trend data
Real acquisition clients for Earth Engine, ERA5, NASA FIRMS, and Alaska Fire Service, plus a physically-motivated synthetic generator for offline development

Project Links

Technologies

PyTorchFastAPICNNBiLSTMAttentionDockerGitHub Actions

Key Impact

49/49 tests passing across feature indices, model, dataset, training, and API layers
Gradient x input saliency and attention visualizations for model interpretability
Evaluation report honestly discloses where the model does and doesn't beat a majority-class baseline