All Smoke and Some Flame: A Review of Artificial Neural Networks in Wildfire Prediction and Detection 54 views

Wildfires are becoming more frequent and severe due to climate change, unsustainable land-use practices, and expanding human development, generating demand for accurate prediction and detection technologies that support the needs of forest management. Artificial neural networks (ANNs) have emerged as promising tools for wildfire applications, but their effectiveness and practical management value remain weakly connected in the literature. This systematic review evaluated the performance, strengths, limitations, and management applications of ANN architectures used for wildfire prediction and detection. A systematic literature search of Web of Science identified 50 studies published between 2023 and 2025 that met predefined inclusion criteria. The quantitative analysis compared model performance across ANN architectures, spatial scales, and wildfire applications, while the qualitative synthesis reviewed reported management uses and reported limitations. Convolutional neural networks (CNNs) were the most commonly used architecture and achieved the highest mean reported accuracy (90.16%). Model performance varied moderately across spatial scales, with local studies reporting the highest mean accuracy. Across the reviewed literature, ANN-based detection systems showed the greatest potential for operational implementation. However, ANN models did not consistently outperform traditional machine-learning methods such as Random Forest or Support Vector Machine. In addition, limitations related to data quality, model transferability across geographic regions, and inconsistent evaluation metric reporting were consistently identified. These findings suggest that the value of ANN models depend on their suitability for specific wildfire management applications as opposed to predictive accuracy alone.

MA (Master of Arts)

Artificial Neural Network; Forest Management; ANN; Convolutional Neural Network; Neural Network; Wildfire Prediction; Wildfire Detection; Wildfire

English

2026-09-26