Skip to main content
eScholarship
Open Access Publications from the University of California

UC San Diego

UC San Diego Electronic Theses and Dissertations bannerUC San Diego

Data Conditioning and Data Assimilation for Wildfire and Prescribed Fire Modeling

Abstract

Wildfire is an unplanned fire that can happen at any time in a natural area with combustible materials. Normally, there are two main reasons for the occurrence of a wildfire. One is due to human activity and the other one is owing to natural phenomenon, such as lighting, volcanic eruption, or even falling meteors. Although some naturally occurring wildfires may be beneficial for the ecological balance, most of the wildfires are destructive. They can lead to air pollution and may bring death and destruction. As the climate gets warmer and drier, the occurrence of wildfire is more frequent. Consequently, the estimation and prediction of the spread of wildfire is significant. Prescribed fires are the planned fires that burn under specified conditions to achieve specific objectives. Compared to wildfires, prescribed fires are implemented and controlled in a safer way to balance ecosystems. Meanwhile, prescribed fire can also be used as a tool to reduce fuel build-up and avoid the occurrence of wildfires. Hence, how to design a safe and effective burn plan for the prescribed fire is a good topic for research. This dissertation proposes multiple algorithms for estimation and prediction of wildfire spread, detection of the wildfire perimeter, and safety evaluation of the prescribed fire. First, this dissertation shows how to improve the prediction capability of a fire model by estimating the wind conditions. Two errors, an uncertainty-weighted least-squares error and an uncertainty-weighted surface area error, are established and computed between the predicted and measured fire perimeters, and an optimization is applied to find the optimal wind conditions based on iterative refined gridding of the wind conditions. To better characterize the wildfire for the estimation and prediction of the fire spread, three approaches are introduced to detect the wildfire perimeter in different situations. The first two approaches, quadriculation algorithm and iterative minimum distance algorithm, are used to establish a closed polygon of the wildfire perimeter for a well-defined fire image, and the third one, iterative trimming method, is based on the Delaunay triangulation and designed for the situation when the pixels with high infrared values in a thermal infrared image of the wildfire are disconnected or sparse. In addition to managing wildfire directly, the prescribed fire can be utilized to reduce and prevent the wildfire. An algorithm of automatically labeling the safety of a prescribed fire is proposed in this dissertation to avoid a labor-intensive process of manual labeling. All the proposed algorithms are illustrated on real data of a wildfire or simulations from modern fire simulation tools.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View