Search for: All records

FLAME 3 is the third dataset in the FLAME series of aerial UAV-collected side-by-side multi-spectral wildlands fire imagery (see FLAME 1 and FLAME 2). This set contains a single-burn subset of the larger FLAME 3 dataset focusing specifically on Computer Vision tasks such as fire detection and segmentation. Included are 622 image quartets labeled Fire and 116 image quartets labeled No Fire. The No Fire images are of the surrounding forestry of the prescribed burn plot. Each image quartet is composed of four images - a raw RGB image, a raw thermal image, a corrected FOV RGB image, and a thermal TIFF. Each of the four data types are detailed in the below Table 1. More information on data collection methods, data processing procedures, and data labeling can be found in https://arxiv.org/abs/2412.02831. This dataset also contains a NADIR Thermal Fire set, providing georeferenced overhead thermal imagery, captured by UAV every 3-5 seconds, focusing on monitoring fire progression and burn behaviors over time. This data, when processed, enables centimeter-grade measurements of fire spread and energy release over time. Pre, post, and during burn imagery are included, along with ground control point (GCP) data.  This dataset is based on the research conducted in the paper: FLAME 3 Dataset: Unleashing the Power of Radiometric Thermal UAV Imagery for Wildfire Management. It provides detailed insights and analysis related to forest fire monitoring and modeling.If you use this dataset in your research or projects, please cite the original paper as follows: APA: Hopkins, B., ONeill, L., Marinaccio, M., Rowell, E., Parsons, R., Flanary, S., Nazim I, Seielstad C, Afghah, F. (2024). FLAME 3 Dataset: Unleashing the Power of Radiometric Thermal UAV Imagery for Wildfire Management. arXiv preprint arXiv:2412.02831.BibTeX: @misc{hopkins2024flame3datasetunleashing, title={FLAME 3 Dataset: Unleashing the Power of Radiometric Thermal UAV Imagery for Wildfire Management}, author={Bryce Hopkins and Leo ONeill and Michael Marinaccio and Eric Rowell and Russell Parsons and Sarah Flanary and Irtija Nazim and Carl Seielstad and Fatemeh Afghah}, year={2024}, eprint={2412.02831}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2412.02831}, } 

Terrestrial LiDAR scans (TLS) offer a rich data source for high-fidelity vegetation characterization, addressing the limitations of traditional fuel sampling methods by capturing spatially explicit distributions that have a significant impact on fire behavior. However, large volumes of complex, high resolution data are difficult to use directly in wildland fire models. In this study, we introduce a novel method that employs a voxelization technique to convert high-resolution TLS data into fine-grained reference voxels, which are subsequently aggregated into lower-fidelity fuel cells for integration into physics-based fire models. This methodology aims to transform the complexity of TLS data into a format amenable for integration into wildland fire models, while retaining essential information about the spatial distribution of vegetation. We evaluate our approach by comparing a range of aggregate geometries in simulated burns to laboratory measurements. The results show insensitivity to fuel cell geometry at fine resolutions (2–8 cm), but we observe deviations in model behavior at the coarsest resolutions considered (16 cm). Our findings highlight the importance of capturing the fine scale spatial continuity present in heterogeneous tree canopies in order to accurately simulate fire behavior in coupled fire-atmosphere models. To the best of our knowledge, this is the first study to examine the use of TLS data to inform fuel inputs to a physics based model at a laboratory scale. 

ABSTRACT Long-term terrestrial ecosystem monitoring is a critical component of documenting outcomes of land management actions, assessing progress towards management objectives, and guiding realistic long-term ecological goals, all through repeated observation and measurement. Traditional monitoring methods have evolved for specific applications in forestry, ecology, and fire and fuels management. While successful monitoring programs have clear goals, trained expertise, and rigorous sampling protocols, new advances in technology and data management can help overcome the most common pitfalls in data quality and repeatability. This paper presents Terrestrial Laser Scanning (TLS), a specific form of LiDAR (Light Detection and Ranging), as an emerging sampling method that can complement and enhance existing monitoring methods. TLS captures in high resolution the 3D structure of a terrestrial ecosystem (forest, grassland, etc.), and is increasingly efficient and affordable (<$30,000). Integrating TLS into ecosystem monitoring can standardize data collection, improve efficiency, and reduce bias and error. Streamlined data processing pipelines can rigorously analyze TLS data and incorporate constant improvements to inform management decisions and planning. The approach described in this paper utilizes portable, push-button TLS equipment that, when calibrated with initial transect sampling, captures detailed forestry, fuels, and ecological features in less than 5 minutes per plot. We also introduce an interagency automated processing pipeline and dashboard viewer for instant, user-friendly analysis, and data retrieval of hundreds of metrics. Forest metrics and inventories produced with these methods offer effective decision-support data for managers to quantify landscape-scale conditions and respond with efficient action. This protocol further supports interagency compatibility for efficient natural resource monitoring across jurisdictional boundaries with uniform data, language, methods, and data analysis. With continued improvement of scanner capabilities and affordability, these data will shape the future of terrestrial ecosystem monitoring as an important means to address the increasingly fast pace of ecological change facing natural resource managers.