Skip to main content

Wildfire events in California and around the world have resulted in significant social and economic losses. In response, there is strong focus on modeling wildfire behavior in the wildland to aid in developing strategies for fuel reduction and fire suppression. Post-fire observations of several historic fires indicate that within affected communities, some ignitable structures tend to survive even when most in close vicinity have been destroyed. This apparent spatial randomness in damage patterns raises an important question: Can the likelihood of survival of individual buildings be predicted for wildfire events?

What is needed is a more holistic approach that integrates the characteristics of the wildland with the features of the built environment to better predict damage and building loss. 

To advance this important modeling goal, Technosylva led a 2022 peer-reviewed study, published in the academic journal, Scientific Reports, that utilized integrated concepts from graph theory to create a relative vulnerability metric that can quantify the survival likelihood of individual buildings within a wildfire-affected region. By emulating the damage observed in historic wildfires the analysis was able to test and refine this framework, resulting in findings showing that both formulations modeling loss at the mid and extreme ranges based on graph centralities were effective in evaluating the vulnerability of buildings relative to each other.

We wanted to create models that could help to understand, in an operational way, the complexity of the building-to-building risk of fire propagation and try to understand the potential of urban conflagration. This kind of modeling is really complex and we wanted to make the actual models better when we apply fire simulations and ignition impact forecasting to operations. This means moving from empirical models to a little more physics-based models in a way that it could be applied to nation scale or even after global scale. To do that we have to go to the basics of fire and fire behavior and build back out.

Joaquin Ramirez, Ph.DCo-Author - President & CTO

Insight from the Research

The study presents an innovative approach to predict the survival of individual buildings during wildfire events by utilizing concepts from graph theory. It addresses a crucial question in the field of fire management: why do some structures within fire-affected communities manage to survive while neighboring buildings succumb to the flames? While traditional wildfire behavior models excel in natural environments, they often fall short in explaining the complex dynamics within built-up areas. This study seeks to bridge that gap by introducing a novel methodology that leverages graph theory to model wildfire propagation within communities.

Proposed relative vulnerability framework

Cite: Integrated graph measures reveal survival likelihood for buildings in wildfire events

Defining the Modeling in the Research

The research begins by introducing a wildfire graph model, which serves as the foundation of the study. The model represents the intricate interactions between buildings and vegetation in areas affected by wildfires. To enhance the model’s accuracy, the researchers include nodes to account for both wildland vegetation and urban vegetation. These two types of vegetation are recognized as influential factors in determining fire intensity and rate of spread. Additionally, the model incorporates various building features, such as deck type, eaves, roof type, vent type, fence, and window pane, as these factors significantly affect a structure’s vulnerability to wildfires.

Centrality Measures and Building Vulnerability

The core of the study involves evaluating different centrality measures from graph theory to assess the vulnerability of individual buildings within the wildfire-affected community. The traditional centrality measures considered include:

  1. Closeness Centrality: Measures how close a node is to all other nodes in the network.
  2. Eigenvector Centrality: Evaluates a node’s influence based on its connections to other influential nodes.
  3. Clustering Coefficient: Assesses the extent to which a node’s neighbors are connected to each other.
  4. Gravity Centrality: Determines the importance of nodes based on their spatial distribution within the network.
  5. Degree Centrality: Measures the number of connections a node has.
  6. Betweenness Centrality: Identifies nodes that act as bridges between different parts of the network.

The traditional centrality measures are widely accepted in network analysis but were found to have limited accuracy when it comes to predicting building damage in wildfire scenarios.

Novel Formulations for Building Vulnerability

To address the shortcomings of traditional centrality measures, the paper introduces two novel formulations for assessing building vulnerability:

  1. Modified Degree Formulation: This approach focuses on the influence of nearby nodes. It calculates the relative vulnerability of individual buildings by considering the mean of incoming edge weights from neighboring nodes. Furthermore, it introduces a mechanism to remove low-impact connections, thereby enhancing accuracy.
  2. Modified Random Walk Formulation: This formulation utilizes random walks to assess the spreading power of nodes within the network. It calculates the relative vulnerability based on the concepts of random walk and information entropy.
Research Findings

The study rigorously tests these novel formulations alongside traditional centrality measures on two historic wildfires, namely the 2018 Camp Fire and the 2020 Glass Fire. The results indicate that the modified degree and random walk formulations outperform traditional centrality measures in predicting building damage. While none of the formulations achieved exceptionally high accuracy, they provided valuable insights into general damage patterns within different regions affected by wildfires.

Next Steps from the Research

The study acknowledges the inherent challenges in modeling wildfire behavior, including aleatory uncertainties (randomness) and limited data availability at the community level. It emphasizes the need for improved models that can account for both aleatoric and epistemic uncertainties, which include factors like local weather effects and the extent of fire spread. While the proposed formulations represent a significant step forward, the authors highlight the complexity of the research problem and the need for further refinement and data collection.

The study offers a promising avenue for better understanding and mitigating the impacts of wildfires on communities. It introduces innovative concepts from graph theory to enhance our ability to predict building survivability in wildfire events, providing valuable tools for fire management and community planning.

Learn more about how this science is put into practice.