Deep Learning Models for Forest Fire Prediction: Insights into Feature Selection for Climate- Resilient Forestry

Loading...
Thumbnail Image

Date

2025-11-25

Journal Title

Journal ISSN

Volume Title

Publisher

Journal of Sustainable Forestry,

Abstract

This scoping review investigates feature selection and integration practices in deep learning models for forest-fire prediction, emphasiz- ing their impact on model performance and sustainable forest man- agement. Following PRISMA guidelines, 36 peer-reviewed studies published between 2020 and 2025 were systematically synthesized. The analysis identifies commonly adopted features, topographical, climatic, vegetation, and anthropogenic, and examines how their selection influences predictive accuracy, interpretability, and contex- tual relevance. Results show that hybrid architectures such as CNN- LSTM and U-Net variants effectively capture spatial and temporal fire dynamics. Feature selection techniques, including correlation analysis, SHAP, Relief-F, and Information Gain, improved model robustness and transparency. Challenges such as class imbalance, high-dimensional data, and inconsistent temporal resolution were addressed using advanced preprocessing, dimensionality reduction, attention mechan- isms, and explainable AI. However, gaps persist, particularly in under- represented regions, regarding standardized, context-aware feature integration. The study highlights the importance of incorporating socio-environmental indicators, such as land use and human activity, to reflect real-world fire triggers. Multi-source data fusion emerged as critical for improving generalizability and operational deployment. This review emphasizes deep learning’s potential to advance sustain- able forest management, support climate mitigation, and foster resi- lient ecosystems and communities. It also emphasizes its alignment with SDG 13: Climate action and SDG 15: Life on land.

Sustainable Development Goals

SDG 13: Climate Action SDG 15: Life on Land

Keywords

Deep learning, forest fire prediction, feature selection, sustainable forestry, climate change adaptation

Citation