PyroAdapt: Adapting Wildfire Prediction under Spatial Heterogeneity and Temporal Shift
PyroAdapt adapts wildfire models by retrieving similar sites and ranking fire-prone cells under a daily budget.
PyroAdapt adapts pretrained wildfire models to new spatial and temporal conditions by retrieving historically similar locations and fine-tuning through risk ranking. On a California grid of 666 cells, ranking objectives lift daily average precision from 21.62% under continued focal fine-tuning to 24.35–24.57%, and Top 5% recall from 18.70% to 22.11–22.79%. Under a fixed daily budget of 34 cells, selective ranking captures 344 additional positive cell-days and improves recall on the most extreme fires, with gains holding in rolling Yosemite tests.
- Pretrain-retrieve-rank adaptation for rare wildfire events under spatial and temporal shift.
- California grid of 666 cells at 0.25-degree resolution.
- Daily average precision rises from 21.62% to about 24.4–24.6%.
- Selective ranking finds 344 more positive cell-days under a 34-cell daily budget.
- Gains persist in rolling Yosemite evaluations under temporal shift.
From the original article
1 min read at huggingface.co
Prediction of wildfire occurrence is a rare-event problem compounded by spatial heterogeneity and temporal distribution shift, as fire occurrences are vastly outnumbered by non-occurrences, and predictor--fire relationship varies across space and time. Models trained on historical fire data may perform poorly under new conditions and require adaptation to the target distribution before operational use.
By Enyi Jiang, Wu Sun · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.