Robust Wildfire Forecasting under Partial Observability: From Reconstruction to Prediction

TL;DR

This work forms wildfire forecasting under partial observability using a two-stage probabilistic framework that decouples observation recovery from spatiotemporal prediction, and inserts the reconstruction stage before forecasting significantly mitigates the domain gap, restoring next-day prediction accuracy to near-clean-input levels even under severe information loss.

Abstract

Satellite-derived fire observations are the primary input for learning-based wildfire spread prediction, yet they are inherently incomplete due to cloud cover, smoke obscuration, and sensor artifacts. This partial observability introduces a domain gap between the clean data used to train forecasting models and the degraded inputs encountered during deployment, often leading to unreliable predictions. To address this challenge, we formulate wildfire forecasting under partial observability using a two-stage probabilistic framework that decouples observation recovery from spatiotemporal prediction. Stage-I reconstructs plausible fire maps from corrupted observations via conditional inpainting, while Stage-II models wildfire dynamics on the recovered sequences using a spatiotemporal forecasting network. We consider four network architectures for the reconstruction module-a Residual U-Net (MaskUNet), a Conditional VAE (MaskCVAE), a cross-attention Vision Transformer (MaskViT), and a discrete diffusion model (MaskD3PM)-spanning CNN-based, latent-variable, attention-based, and diffusion-based approaches. We evaluate the performance of the two-stage approach on the WildfireSpreadTS (WSTS) dataset under various settings, including pixel-wise and block-wise masking, eight corruption levels (10%-80%), four fire scenarios, and leave-one-year-out cross-validation. Results show that all learning-based recovery models substantially outperform non-learning baselines, with MaskCVAE and MaskUNet achieving the strongest overall performance. Importantly, inserting the reconstruction stage before forecasting significantly mitigates the domain gap, restoring next-day prediction accuracy to near-clean-input levels even under severe information loss.

Figures (7)

• Figure 1: Overview of the proposed two-stage framework for wildfire forecasting under partial observability. Given multi-modal environmental observations together with partially observed fire maps, Stage-I performs morphological reconstruction to recover a plausible complete fire history from corrupted inputs, explicitly addressing the domain gap introduced by missing satellite observations. The reconstruction module can be implemented with various different approaches, including CNN-based, latent generative, transformer, or discrete diffusion networks. The recovered sequence is then fed into Stage-II, a spatiotemporal prediction model that learns wildfire dynamics and produces the future fire map $\hat{\mathbf{F}}_t$. By decoupling observation recovery from temporal forecasting, the framework enables robust prediction under severe observation degradation.
• Figure 2: Visualization of the 23-channel input features for a single daily observation of the Dolan Fire event in Monterey County, California, from Aug 14 to Aug 31, 2020. Grey regions indicate missing or undetected pixels (NaNs). Each subplot employs an independent color scale, identical colors across panels may represent different physical magnitudes.
• Figure 3: Daily active fire detections for the Dolan Fire (Aug 16--31, 2020), where colored pixels indicate detection time and black denotes missing or undetected (NaN).
• Figure 4: Definition of the four evaluation scenarios in wsts. Each row shows a binary 5-day historical fire sequence together with the prediction target, where white regions denote active fire pixels.
• Figure 5: Comparison of Stage-I recovery performance across varying masking difficulty levels ($\eta \in [10\%, 80\%]$). The performance is evaluated using the DICE coefficient, which is particularly effective for capturing the structural integrity of sparse flame patterns. In both (a) and (b), Random and Dilation represent the baseline methods, whereas the other four curves denote different learning-based models. The results highlight the superior recovery performance of learning-based approaches under high-sparsity conditions.