How can AI and computer vision help improve a property’s defense against wildfire? Wildfire continues to impact homeowners, communities, and the insurance industry. Climate change, persistent drought, hotter temperatures, increased migration into the wildland-urban interface (WUI), and stakeholders lacking strong insight into wildfire risk add up to a challenging situation. Legacy tools and data sources used by stakeholders, like fire agencies and insurers, to understand and quantify risk have proven to be inaccurate or out-of-date. New solutions that use aerial and satellite imagery can produce quantifiable insights for every home in the United States by leveraging AI and machine learning algorithms to augment existing regional-focused hazard information with property-level vulnerability characteristics. Due to industry and academic research highlighting their importance, computer vision and machine learning technologies can focus on the critical vulnerability aspects of a residential property like its vegetation clearance, roof construction, and proximity to surrounding structures. Using AI at scale then allows stakeholders to measure—for the first time—the exact impact of these property characteristics on risk and the likelihood of sustaining damage with precise statistical rigor. Understanding a specific property’s vulnerability profile to mitigate risk, especially for homes and businesses across the wildfire-prone Western United States, could transform wildfire risk assessment. Insights generated with AI can be used to optimize fire agency programs and give insurers new opportunities to engage homeowners on effective risk mitigation practices that can help keep properties safe from the increasing threat of widespread wildfires. In this presentation, learn about the practical deployment of this technology, its current capabilities and its limits, and how fire agencies and insurers can harness it to understand wildfire risk factors and to drive proactive conversations about mitigation with home and business owners.

• PPT Slides

A Probabilistic Wildfire Evacuation Planning Framework and Software Platform

Marilia Ramos, University of California, Los Angeles

Ensuring an efficient evacuation of Wildland Urban Interface (WUI) communities in case of wildfire is a pressing challenge. Wildfire evacuation modeling can be characterized by three main aspects: fire model, human decision-making, and traffic models. Efficient evacuation planning needs a comprehensive understanding of these aspects and their mutual interactions. A number of methods have been proposed for wildfire evacuation assessment, focusing on each of these components, but few address the issue considering all these layers. This presentation will present the Wildfire Safe Egress (WISE) framework for probabilistic evacuation planning in the case of wildfires. The WISE framework integrates a human decision model, a traffic model, and wildfire dynamics model for estimating the probability that a community safely evacuates when in danger by a wildfire. The likelihood of safe evacuation is calculated by probabilistically comparing two competing parameters: (i) the Available Safe Egress Time (ASET), which determines the total amount of time before the fire reaches a community's borders; and (ii) the Required Safe Egress Time (RSET), that determines the time a community needs to evacuate safely. These variables are modeled through a Bayesian Belief Network (BBN). WISE adopts an agent-based approach for estimating the time for community members to evacuate, based on their socio-demographic profile.. WISE is implemented as a web-based software platform, allowing users to have a practical egress assessment in a visual GIS-based environment.

• PPT Slides

Multimodal Deep Learning-based Remote Sensing for Large-Scale Wildfire Fuel Mapping

Riyaaz Shaik, University of California, Los Angeles

Mapping the distribution of vegetation and biomass available as fuel to wildfires is important for fire behavior simulations. The majority of existing approaches have mapped vegetation types and fuels using machine learning based on remote sensing data for regions of interest spanning smaller than a few hundred square kilometers. In such cases, numerous small-scale site-specific models are required to cover the landscape at the state level, which can result in increased latency and limit the real-time availability of up-to-date mapped fuels data for fire spread simulations. Considering this limitation, a deep-learning-based large-scale fuel identification procedure is developed by integrating and fusing data from different sources (high-resolution optical and multispectral imagery, synthetic aperture radar (SAR) data, vegetation indices, and biophysical climate and terrain data). The convolutional neural network technique is used to extract visual features of 40 categories of fuels from the imagery, whereas multi-layer neural networks are used with spectral and biophysical data. This model integrates both object-based and pixel-based classifications and uses the Monte-Carlo dropout mechanism to create a stochastic ensemble of many models that can estimate model uncertainty to boost the prediction performance. Ground truth labels from survey-based fuel plots—complemented with synthetic label data based on a random sampling of existing maps—are used to train the model. The results obtained using this approach show great potential for creating accurate large-scale based fuel maps with accuracy performance on par with available smaller-scale fuel identification methods. It is also shown that the uses of high-resolution imagery and a stochastic ensemble approach have improved the performance, while the latter also allows for the generation of a measure of prediction uncertainty.

• PPT Slides

A WUI-based Plan Integration for Resilience Scorecard™

Co-presented by William (Bill) Siembieda, Cal Poly WUI Fire Institute and Matthew Malecha, Texas A&M University

The problem at hand. Between 2003 and 2021, the top 10 costliest wildland fires in the United States all occurred in California. One in four Californians live in an area considered high risk for wildfires. Given the impact of catastrophic fires in western states, the need exists for increasing Wildland Urban Interface (WUI) fire resilience through assessing and managing local land use, mitigation and adaptation plans. Most California jurisdictions produce and follow many different types of plans (e.g., General Plan with required Safety Element; Hazard Mitigation Plan, Community Wildfire Protection Plan), each with its own set of policies and implementation scheme; generally, these are “siloed” and lack integration. This is especially true with regard to hazards, and in particularly the WUI fire hazard that is addressed by several agencies and plans, but without sufficient collaboration and spatial understanding of the heterogenous effects of policy across a community.
A proposed solution. To address the many-plan-little-integration dilemma, this project focuses on applying the Plan Integration for Resilience Scorecard™ (PIRS) method to California jurisdictions that are subject to the Wildland Urban Interface (WUI) fire hazard. This a method that spatially evaluates networks of plans to strengthen a jurisdiction’s resilience and reduce vulnerability to hazards. It provides a pathway to systematically evaluate and then adjust multiple policies to improve the focus and coordination of plans on building resilience in the most vulnerable locations. The process aims to harmonize a jurisdiction’s network of plans to support community priorities that lower risk from hazards.
With support from the Department of Homeland Security (DHS) and an initial focus on flood hazards, the PIRS™ method was first applied in several East and Gulf Coast cities and in the Netherlands. FEMA later used the process to conduct vulnerability analysis in the Southern U.S., and the American Planning Association (APA) has adopted it and developed an online course to teach PIRS™ to its members. NOAA recently provided support for urban heat island hazard analysis using the PIRS™ method for a pilot in several cities.
WUI fire presents a special challenge in that the hazard itself stems from the dynamics of fuel variables (natural and built) interacting with climate and human variables. Through building a scorecard supported by a spatial framework, policies that support risk reduction are made visible, while that those are in conflict with a risk reduction strategy are called out.
The project team will work with four California jurisdictions over a two-year period, to implement PIRS’s three part process and make adjustments to the challenges that the WUI fire hazard presents. We use a multidisciplinary team of faculty associated with the Cal Poly-San Luis Obispo WUI Fire Institute and the Texas A&M Hazard Reduction and Recovery Center. The Gordon and Betty Moore Foundation provides support for this effort.

• PPT Slides

Keynote #3: NIST Camp Fire Research with Brief CAL FIRE Overview

Keynote Speakers: Alexander Maranghides, NIST and Steve Hawks, CAL FIRE

• PPT Slides

Structural brick as a viable alternative to CMU for non-combustible construction

Steven Judd, Interstate Brick

This presentation will provide an introduction to the fire-resistive benefits of clay masonry in general, and structural clay masonry specifically for fire-resistive construction alternatives. Single wythe clay masonry meets the California State Energy Code for maximum U-factor values listed in Section 1207 - Mandatory Insulation Requirements for opaque portions of walls that separate conditioned spaces from unconditioned spaces or ambient air. The presentation will show how to determine the fire-resistive rating for different clay masonry wall types. The presentation will show examples of different building types that use structural clay masonry for exterior wall systems as simple veneers, and as reinforced structural bearing and shear walls, as well as reinforced curtainwalls, which is unique and not well known. The presentation will broaden the knowledge base of this structural clay brick and provide unique solutions to fire-resistive construction.

• PPT Slides

Increasing Community Resilience to Wildfire: Wildfire Risk Reduction Buffers

Deborah Glaser, The Nature Conservancy

In 2018, the town of Paradise, California, made international news when a devastating wildfire killed eighty-six people, burned 95% of the town’s structures to the ground, and caused billions of dollars in damage. Like Paradise, many communities throughout California and the western US are facing wildfire risk that makes their economic viability uncertain. New approaches are needed to provide more sustainable paths forward.
Building on previous work conducted by The Nature Conservancy, the Conservation Biology Institute, and the Paradise Recreation and Park District, this research applies a U.S. wildfire catastrophe model to quantify the financial benefits of ignition reductions in Paradise, including the potential impacts of spotting embers, in terms of avoided property loss and lower expected insurance premium costs for a range of wildland buffering strategies.
Results from the research indicate that land use strategies like fire buffers—areas of reduced fuels such as green spaces or parklands—can be as effective at reducing urban ignition risk as the enforcement of updated fire resilient building codes. When taken together, fire buffering strategies and code updates can substantially reduce loss risk.
The research also demonstrates the economic value of a community-level focus on catastrophe resilience and the potential for Community-Based Catastrophe Insurance (CBCI) to capture the financial benefits for residents through reduced insurance premiums.

• PPT Slides

Engaging Residents as Stakeholders in Becoming a Fire Adapted Community

Marco Mack, Retired-Central Fire District of Santa Cruz Co

“How do we reduce wildfire risks in our existing and built environment?”
Increasing building code and fire code regulations for new developments makes sense but making mandatory requirements for our existing structures will create tremendous push-back from some residents. The current regulatory process is evaluating new code requirements for hardening existing homes and removing combustible items and vegetation around structures; when implemented, this process could be on a range between pass-fail regulatory enforcement or a minimum standard with increased requirements placed on individual parcels that will mitigate specific hazards.
This presentation provides lessons learned from Aptos Fire & Central Fire in Santa Cruz County and a test-drive Wildfire Risk Reduction Outreach where we engaged residents as Stakeholders with, “How do we become a Fire Adapted Community without an ordinance?” This program created an alternative Defensible Space Inspection Program for portions of the Fire District where our operational capability would be challenged under normal weather conditions.
The take-away from this presentation is applicable to Fire Departments, Emergency Managers, Fire Safe Councils, Planning/Development, Resource Conservation Districts, Water Districts, and political leaders. Questions addressed, include: “What do residents need on the Urban side vs. the Rural side of the WUI?”; “How will this program scale to our entire Fire District?”; “How can we rank individual parcels so that we spend more time where we identify operational challenges?”; “We need a Risk Mitigation Scale for Residents who choose to do more than the code minimum”; “How do we create pathways for residents to replace missing neighborhood infrastructure (access-egress-emergency water-neighboring parcel fuel modification)?”; “How do we coordinate/integrate our efforts with local government agencies?”; and, “How do we motivate residents to stay engaged?

• PPT Slides

Jim McDougald, Matt Damon, and Frank Bigelow, CAL FIRE

Today's wildfire environment has changed, allowing for longer fire seasons with more intense and destructive wildfires. Climate change, land use, fire suppression, population growth, and other factors have contributed to putting millions of homes and hundreds of communities at extreme risk of wildfire. During this presentation, best management practices for living and building in wildfire-prone areas will be discussed. The 2019 California Building Code: Chapter 7A requirements for exterior protection of structures and the 2019 California Fire Code: Chapter 49 defensible space requirements intended to mitigate wildfire hazards will be discussed. Fire, building, and enforcement professionals interested in gaining knowledge on identifying your community's wildfire hazards, risks, and mitigation strategies are encouraged to attend.

• PPT Slides

Home Hardening Program

Frank Bigelow, CAL FIRE

Today's wildfire environment has changed, allowing for longer fire seasons with more intense and destructive wildfires. Climate change, land use, fire suppression, population growth, and other factors have contributed to putting millions of homes and hundreds of communities at extreme risk of wildfire. During this presentation, best management practices for living and building in wildfire-prone areas will be discussed. The 2019 California Building Code: Chapter 7A requirements for exterior protection of structures and the 2019 California Fire Code: Chapter 49 defensible space requirements intended to mitigate wildfire hazards will be discussed. Fire, building, and enforcement professionals interested in gaining knowledge on identifying your community's wildfire hazards, risks, and mitigation strategies are encouraged to attend.

• PPT Slides

Building WUI Water Distribution Resilience; Social vulnerability, Response, and Recovery to the 2018 Camp Fire and 2017 Tubbs Fire

Eliza Amstutz and Jenna Tilt, Oregon State University

Recent wildfires in California and Oregon caused significant damage to water distribution systems resulting in water that was contaminated with benzene and other volatile organic compounds (VOCs). To understand the risk of exposure, we developed a machine-learning model based on neural networks that accurately (85%) predicted water contamination after wildfire using publicly available spatial predictors. We then developed the Wildfire Vulnerability Explorer, a user-friendly interface that combines the water contamination predictive model with a place based social vulnerability assessment utilizing census, parcel, and land use data. Using fuzzy-logic Environmental Evaluation Modeling System (EEMS), the Wildfire Vulnerability Explorer provides an interactive visualization of areas most vulnerable to post-fire water contamination. The Explorer allows the user to visualize how social and environmental factors interact to more accurately reflect on-the-ground vulnerability. Utilization can help facilitate community conversations regarding innovations in water infrastructure and community response/adaptation to wildfire events. Stakeholders from both Santa Rosa and Paradise, CA have provided key suggestions and validation of the explorer for their respective communities through focus group workshops and qualitative thematic analysis. This research aims to engage with WUI communities to better understand community vulnerability and identify feasible adaptation strategies.

• PPT Slides

Wildfire impacts to water infrastructure

Amy Metz, Oregon State University

Wildfires have damaged civil infrastructure throughout wildland urban interface communities, including water distribution systems. Post-wildfire, communities have detected volatile organic compounds (VOCs) within drinking water systems. Previous studies have suggested that thermal degradation of plastic pipe materials is a potential source of this contamination; however, testing methodologies for pipelines at elevated temperatures is not standardized, non-plastic materials have not been tested, and the testing results have not been linked back to common infrastructure or wildfire characteristics (e.g. surface heat flux, burn duration, burial depth). To fill these gaps in knowledge, this study conducted tests to evaluate pipe materials at various elevated temperatures. Post-heating treatments of the pipes were considered (spirals, small pieces, and whole sections) to examine the influence of different testing methodologies on the contamination levels found in water. Water was analysed and regression curves were fitted to the results to quantify the threshold temperatures at which pipe materials may release VOCs in excess of federal limits. Practical metrics, including wildfire conditions and pipe burial depths, were varied to investigate under what circumstances buried service laterals may exceed threshold temperatures. Results show that small lengths of pipe were required to be heated to breach the federal limits.

• PPT Slides

Post-fire decision making for community recovery

Megan Ellery and Noah Gershon, University of Colorado Boulder

Over 1000 suburban homes in three jurisdictions with varying priorities and locally adopted building codes were destroyed by the Marshall Fire. This disaster presents a unique research opportunity considering both the large scope of the destruction and the multiple jurisdictional boundaries it encompasses. Dozens of interviews were held with homeowners and public officials regarding post-fire rebuilding decisions to characterize decisions, tradeoffs, rationale, and information needed. We focus specifically on cost, speed, sustainability, and fire resilience for home reconstruction.
Impacted homeowners face numerous challenges on the long road to recovery. They must find and make sense of substantial amounts of technical information and decide about if and how to rebuild. In response, community groups have formed to aid homeowners by providing platforms for discussion, information sharing, and collective action. At the same time, public officials must balance supporting community rebuilding efforts while also upholding other voter-initiated priorities. These decisions include tradeoffs around energy and fire code adoption, permitting, taxes, and resource allocation. At both the jurisdiction and household level, we found tensions between community and individually held values around sustainability, resilience, and costs and time of rebuilding. There is also substantial uncertainty and misinformation about the benefits and costs of various building practices.

• PPT Slides

What can we learn from the single-family homes that survived the 2018 Camp Fire in Paradise, California?

Yana Valachovic, UC Cooperative Extension

What are the best strategies to incentivize adaptation in the built environment?
The 2018 Camp Fire, which destroyed 18,804 structures, including most of the town of Paradise, provided an opportunity to investigate vegetation and housing factors associated with home loss and determine whether California's 2008 adoption of exterior building codes for homes in the wildland-urban-interface (WUI) improved survival. This presentation will answer two questions: 1) what was the impact of the 2007/2008 building code update on home survival, and 2) what factors were most strongly associated with home loss? This presentation will demonstrate the importance of continued defensible space actions along with essential home hardening actions.
Homes built before 1997, fared poorly, with only 11.5% surviving, compared with 38.5% for homes built in 1997 and after. The difference in survival percentage for homes built immediately before and after adopting Chapter 7A in the California Building Code (37% and 43%, respectively) was not statistically significant. Distance to nearest destroyed structure, number of structures destroyed within 100 m, and overstory canopy cover within 100 m of the home were the strongest predictors of survival. Most fire damage to surviving homes resulted from radiant heat from nearby burning structures or flame impingement from the ignition of near-home combustible materials, the latter highlighting the value of eliminating combustible items in the area closest to the home (0-1.5 m).

• PPT Slides

Opportunities and Challenges in Building Resilient Communities in Sonoma County California

(CANCELLED) Coupled Wildland Fire-Atmosphere Forecasting using WRF-Fire Informed by Real-World Fire Observations

Kasra Shamsaei, University of Nevada, Reno
In this study, a computationally efficient data assimilation method to assimilate observed fire perimeters in wildland fire simulation using WRF-Fire, a coupled wildland fire-atmosphere simulation platform, is introduced and validated to improve the fire simulations accuracy. Accurate wildfire simulation faces several challenges considering numerous sources of modelling uncertainties and errors. Data assimilation is an effective approach to reduce modelling uncertainties and errors by integrating the model with real-world observations such as fire perimeters. In this study, an operational data assimilation method is introduced in which the simulated fire perimeter is reignited from an observed fire perimeter with the goal of reducing the errors and uncertainties in WRF-Fire. This method is applied to the 2018 Camp Fire simulation using WRF-Fire. First, a baseline case based on the currently used operational model setup is utilized to simulate the Camp Fire. Results show non-negligible differences in terms of rate of spread and direction compared to the fire perimeters derived from NEXRAD radar observations. Next, the radar-driven fire perimeters are used to assimilate the fire perimeter at different times of the Camp Fire simulation, and the effects of assimilation in the forecasting results is investigated. The proposed data assimilation resulted in more accurate fire propagation compared to the baseline case while maintaining operational-suitable computational demand.

Predicting the Survival likelihood of Buildings in Wildfire Events

Hussam Mahmoud, Colorado State University
Most studies on reducing wildfire risk to communities focused on modeling wildfire behavior in the wildland to aid in developing fuel reduction and fire suppression strategies. However, minimizing losses in communities and managing risk requires a holistic approach to understanding wildfire behavior that fully integrates the wildland’s characteristics and the built environment’s features. In this presentation, we show how integrated concepts from graph theory can be used to establish a relative vulnerability metric capable of quantifying the survival likelihood of individual buildings within a wildfire-affected region. We test the framework by emulating the damage observed in the historic 2018 Camp Fire and the 2020 Glass Fire. We propose two formulations based on graph centralities to evaluate the vulnerability of buildings relative to each other. We then utilize the relative vulnerability values to determine the damage state of individual buildings. Based on a one-to-one comparison of the calculated and observed damages, the maximum predicted building survival accuracy for the two formulations ranged from 58 − 64% for the historical wildfires tested. From the results, we observe that the modified random walk formulation can better identify nodes that lie at the extremes on the vulnerability scale. In contrast, the modified degree formulation provides better predictions for nodes with mid-range vulnerability value

• PPT Slides

Structure-to-Structure Fire Spread: Accessory (shed) Structures

Faraz Hedayati, Insurance Institute for Business & Home Safety

It is possible for fire to spread from one building to another within a community causing widespread damage. Slope, building size, distance between buildings, wind direction and intensity influence the spread rate. There have been a few experimental studies regarding fire spread between buildings; most of these studies are based on radiation from flames jetting out through openings during an internal fire. In wildfires, on the other hand, the entire structure may be engulfed in flames, resulting in drastically different fire behavior. To bridge this gap, a multi-year project was initiated to study full scale wind-driven building to building fire spread.

In the first phase of the project, the fire spread from a burning shed (12 x 24 ft with a roof height of 10 feet) to a target building (30 x 40 ft with a mean roof height of 16 feet) was investigated. Fifteen 6-A wood cribs were placed in the sheds and placed in front of the target building's corner at different separation distances. Both buildings were exposed to 30 mph winds. Starting at a 40 ft separation between the shed and target, the distance was reduced by 10 ft increments, and different damage modes were explored on the target building. The results show that at 10 ft separation, the fire from the burning shed ignited the under-eave area within 10 minutes. On the other hand, at the distance of 40 feet, no damage was observed to the target building. Different damage modes on windows, siding and eaves were observed at 20 and 30 ft.

• PPT Slides

Heat Transfer and Fire Spread from Post-Flashover Compartment Fires

Joseph Willi, UL Fire Safety Research Institute
Heat transfer from structure fires to nearby “targets”, such as other structures or vehicles, has critical implications for establishing proper control measures and supporting code development in both urban and WUI environments. Three phases of experiments were conducted to investigate these potential mechanisms of fire spread.

Phase 1 experiments studied the behavior of different exterior building materials and windows during numerous exposures to a well-characterized fire source in the form of a heptane spray burner. The material samples were oriented with heat flux gauges at different offset distances under an oxygen consumption calorimeter.

Experiments in Phase 2 aimed to characterize heat flux exposures from post-flashover fires in a compartment attached to a false façade constructed with vinyl siding over insulation, T1-11, or concrete fiber board. Findings from Phase 1 helped determine the construction and heat flux gauge positions used in Phase 2.

Phase 3 consisted of six experiments that examined the potential for fire spread from a post-flashover compartment fire to a nearby structure. A “target” façade containing two double pane windows was exposed to the compartment/façade fire source characterized in Phase 2. The separation distance between the two structures was selected based on experimental results from Phases 1 and 2. Experiments were conducted both with and without intermediate fuels (shed, car, and deck) positioned between the two structures.

• PPT Slides

Session 6C: Wildfire Policy Decisions - Panel Discussion

Session Panelists: Bob Horton; Bob Roper; Alexander Maranghides; Dave Winnacker

Keynote #4: California Building Insurance - the state’s future standards and best practices for wildfire mitigation and resilience

Keynote Panelists: David Shew; Dave Winnacker; Roy Wright; Julia Juarez

• Note: No slide deck - panel discussion

Session 7A: Wildfire exposure, vulnerability factors, and defensible space development (Part 2)

Background and History of California's Defensible Space Regulations

Yana Valachovic, UC Cooperative Extension
California's fire protection strategies are nationally pioneering and continue to evolve in response to devasting wildfires and structure loss. As an example, in 1965, California established vegetation and fuel modification requirements on properties located in forest, brush, and grass-covered wildland areas. Subsequently, the state created fire hazard severity zones and identified where the state, local, and federal governments are responsible for fire protection and suppression. Recently Governor Newsom signed AB 3074 (2020). This legislation tasked the Board of Forestry and Fire Protection to develop enabling language to create a third defensible space zone, referred to as the “ember resistant zone”, also known as “zone zero”, adding to the existing two zones which extend to 30-feet from the home or building and 100-feet from the home or building, or to the property line. This zone is designed to give enhanced protection to the area immediately surrounding the first five feet of the structure and attached decks. This legislation gained support as a result of an increased understanding of the importance of wind-blown embers (firebrands) in home ignition and loss, based on published scientific reports and post-fire assessments after wildfires in California and other wildfire-prone areas. This presentation will share the background in the science and policy that drove changes in regulations and discuss where and when the “ember resistant zone” will take effect in the coming years.

• PPT Slides

Session 7B: NSF Funding Opportunities for Wildland Fire Science and Engineering

Session Speakers: Daan Liang, Kendra McLauchlan, Joy Pauschke, National Science Foundation

• Session Slide Deck

Session 8A: Insurance and Wildfire Mitigation

Wildfire Risk Reduction and Asset Protection

Valerie Brown, United Policy Holders

United Policyholders, working with wildfire experts, has compiled a list of proven, effective mitigation actions for your home, mirroring those provided by state agencies and IBHS.  With the California Department of Insurance moving forward with regulations that would require insurers to give discount to homeowners who take these specific risk mitigation actions, it's time for homeowners and communities to get involved in taking these actions to reduce their risk from wildfire and with the additional motivation of more affordability and availability of insurance coverage. 

• PPT Slides