Design Evaluation of Machine Learning and Sensors for Detection of Wildfires

There is urgency to address the planetary emergency of wildfires. With impacts on ecology, health, water, and infrastructure, wildfires threaten ecosystems, societies, economies, and communities. Early detection and effective monitoring are vital for improving response efforts and mitigating risks. The latest variety of detection and monitoring technologies calls for a comprehensive analysis approach to provide insights into the capabilities of detection and monitoring systems while recognizing their complex, uncertain operating contexts. This paper describes an application of machine learning classification techniques for wildfire detection to improve the understanding of systems that could be deployed for wildfire risk mitigation. Specifically, camera network, weather station, and satellite observation data are processed, and three models are developed using SmokeyNet architectures that leverage a convolutional neural network, long short-term memory model, and vision transformer. The contributions of this work are the incorporation of satellite data sources and the inclusion of additional weather features in the Multimodal SmokeyNet architecture, as well as the addition of approximately three and a half years of data, increasing the dataset by approximately 36%. Analysis of machine learning model performance provides insights for system design, including how technologies can be leveraged together to improve detection capabilities, such as for accuracy, precision, and time to detection. The incorporation of weather data demonstrates the potential to decrease false positives because the model exhibits higher mean precision (84.3%) and accuracy on nonfire observations (84.1%) compared to the baseline model using only camera images. The detection system that also leverages satellite observations demonstrates enhancements in early detection with a mean average time to detection of approximately 2.25 min and reduced variability in detection times. Overall, the development and analysis of SmokeyNet models illustrates an approach for the initial evaluation of systems of wildfire detection technologies.

Authors

Institutions

Publication Details

Journal
ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems Part A Civil Engineering
Published
2026-11-21
DOI
https://doi.org/10.1061/ajrua6.rueng-1958
Primary Topic
Fire Detection and Safety Systems
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Design Evaluation of Machine Learning and Sensors for Detection of Wildfires

Marco Piras, John F. Organek, James H. Lambert, R. Ranger Dorn et al.
ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems Part A Civil Engineering
Fire Detection and Safety Systems
article

Design Evaluation of Machine Learning and Sensors for Detection of Wildfires

Marco Piras, John F. Organek, James H. Lambert, R. Ranger Dorn, William A. Barletta, Davis C. Loose, Bilal M. Ayyub, Megan E. Gunn, S. Fabrizio Zichichi, Matthew Gunn
article en

Abstract

There is urgency to address the planetary emergency of wildfires. With impacts on ecology, health, water, and infrastructure, wildfires threaten ecosystems, societies, economies, and communities. Early detection and effective monitoring are vital for improving response efforts and mitigating risks. The latest variety of detection and monitoring technologies calls for a comprehensive analysis approach to provide insights into the capabilities of detection and monitoring systems while recognizing their complex, uncertain operating contexts. This paper describes an application of machine learning classification techniques for wildfire detection to improve the understanding of systems that could be deployed for wildfire risk mitigation. Specifically, camera network, weather station, and satellite observation data are processed, and three models are developed using SmokeyNet architectures that leverage a convolutional neural network, long short-term memory model, and vision transformer. The contributions of this work are the incorporation of satellite data sources and the inclusion of additional weather features in the Multimodal SmokeyNet architecture, as well as the addition of approximately three and a half years of data, increasing the dataset by approximately 36%. Analysis of machine learning model performance provides insights for system design, including how technologies can be leveraged together to improve detection capabilities, such as for accuracy, precision, and time to detection. The incorporation of weather data demonstrates the potential to decrease false positives because the model exhibits higher mean precision (84.3%) and accuracy on nonfire observations (84.1%) compared to the baseline model using only camera images. The detection system that also leverages satellite observations demonstrates enhancements in early detection with a mean average time to detection of approximately 2.25 min and reduced variability in detection times. Overall, the development and analysis of SmokeyNet models illustrates an approach for the initial evaluation of systems of wildfire detection technologies.

ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems Part A Civil EngineeringVol. 13(1)
Tongji University (CN), Politecnico di Torino (IT), Electric Infrastructure Security Council (GB), Imperial College London (GB), University of Virginia (US), Massachusetts Institute of Technology (US)
Fire Detection and Safety Systems
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon