Abstract
This study evaluates the impact of the implementation of Machine Learning (ML) integration into early warning systems (EWS) wildfire detection. We provide a longitudinal evaluation of how detection and containment of wildfires has changed with the advent of AI, using data from the California Department of Forestry and Fire Protection, focusing on 2013–2023 as the period when the advent, promotion, and choice of ML tools entered local government disaster preparedness policy design. We include a control number of 5 years before the first adoption of the ALERTCalifornia detection model to provide categorizations for the use of AI in wildfire detection times and acres burnt, as proxies for successful containment. AI utilization in detection and containment is based on a combination of AI models, grouped by University of California, San Diego’s ALERTCalifornia tool, which we categorize into traditional (pre-AI), hybrid, and AI-assisted systems. This classification facilitates the evaluation of AI’s measurable impacts on detection efficiency and containment success. The results show that fairly more sophisticated use of AI significantly reduced detection times. However, we also find that since the deployment of AI, actual acres burnt have significantly increased. AI-aided early detection did not help contain fires faster. We posit that this fact is evidence of climate change. The severity and frequency of wildfire spread in recent years has increased, which has coincided with the recency of AI innovations and adoption in disaster management.