Global review of wildfire prediction using spatio artificial intelligence models

Engineering notes

The study also outlines the need for standard evaluation framework, uncertainty quantification and validation and suggest potential paths for further research, such as benchmark datasets and model adaptability in this domain.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Recent wildfires have raised concerns about the surging incidence of fires on a global scale that poses a significant threat to both humans and biodiversity. The ability of susceptibility and predict such disasters is of utmost importance to effectively address and minimize the associated risks. Several technologies have been suggested for wildfire susceptibility, but the integration of artificial intelligence (AI) is increasing to automate the susceptibility of wildfire instances. Here, we have given a comprehensive systematic evaluation of global applications of AI in wildfire susceptibility and prediction. We systematically explored the contributions of AI-based methods in wildfire susceptibility to date, and 143 scientific research articles have been selected from the Web of Science database. The primary aim of the study is to identify the research gap and analyse recently published research articles using AI methods to enhance a more profound understanding of wildfire susceptibility research; however, previous studies have either focused narrowly on specific algorithms or lacked conceptual clarity. Our review highlights the dominance of tree-based methods (e.g., random forest and boosting), the increasing role of deep learning (e.g., CNNs and LSTMs) models, and the strong performance of hybrid and ensemble approaches. The review also highlights that the evaluation practices rely mainly on metrics such as precision, accuracy, and AUC metrics, a feature that has not received enough attention in previous reviews, as well as pronounced geographic imbalance in AI driven wildfire research is analysed. The study also outlines the need for standard evaluation framework, uncertainty quantification and validation and suggest potential paths for further research, such as benchmark datasets and model adaptability in this domain.

Links and sources

• PDF from original source

Comments