Wildfire Risk Modeling Exercise

Deadline to register is March 9  Register Now

The Wildfire Risk Modeling Exercise is designed as a no-fault forum for all organizations to apply their models in a common simulated environment that encompasses wildland, WUI, and urban environments. This format enables all participating organizations to learn how their models respond to different parameterized variables within natural and built environments and inform future model development.

1. All participating organizations are equally recognized for their scientific and technical merits, resulting in win-win outcomes.
2. At the start of exercise conduct in mid-March, all registered organizations receive the following:
3. Exercise Situational Manual outlining the rules of exercise play and scenarios
4. Exercise data to include real and synthetic data, as well as information about other supplemental data
5. Example notebooks
   1. Access to this virtual exercise platform for collaborating across organizations and submitting model outputs
      Participants run their models applying relevant information provided, synthetic data (also provided), and submit the following model outputs:
      1. Executable notebook (i.e. Jupyter Notebook)
      2. Executable service of the model output
      3. PDF (or other human readable format) report designed for the end user
      4. Model outputs are reviewed in alignment with the objectives, serving as a common foundation for model science innovation and advancement.
6. Findings from this exercise may be used to inform future investments in wildfire risk scientific discovery through future prize challenges and other opportunities.
7. Models are not evaluated or compared against a specific baseline model or other models. This is not a comparative modeling exercise.
8. Each model is reviewed for its unique strengths and sensitivity to parameterized variables in the exercise data sets provided. See below.
    

Eligibility

1. Any organization that has developed a wildfire hazard exposure, vulnerability, and/or risk model.
2. Eligible models must address one or more of the following types of fire: wildland, WUI, structure-to-structure or urban conflagration, and single structure.
3. Eligible organizations include, but are not limited to, academic/research institutions, private companies, non-profit organizations, and public sector entities that have existing or emerging wildfire risk models.

• How do I determine whether my organization or company is eligible to participate? For eligibility information, review the section above on eligibility criteria.
  


• When is exercise conduct? For dates on exercise conduct, review the section above on key dates.
  


• How can I ensure my organization's intellectual property is protected? Participating organizations are permitted to submit only model output information that does not contain information subject to non-disclosure agreements or intellectual property protections.
  


• How are participating organizations evaluated? All modeling outputs will be reviewed by a team of experts. Reviews will be conducted to understanding alignment with the exercise objectives outlined above. Participating organizations and their modeling outputs are not competitively evaluated or scored.
  


• What types of models does this exercise support? This exercise focuses on wildfire risk models. Models may provide deterministic, probability, likelihood, or exposure of any type of wildfire (i.e. wildland, WUI, urban conflagration). Models at any level of spatial extent and resolution can participate, which you identify as part of the registration. The synthetic data and scenes are developed to allow for flexibility in supporting a diversity of wildfire risk model types.
  


• How is Risk defined in the context of this exercise? Potential consequence which could occur to a system, society or a community, determined probabilistically as a function of hazard, exposure, vulnerability, severity of consequence, and capacity to withstand impact.
  


• Can organizations participate as teams with another organization(s)? Yes, organizations can participate as teams in the exercise. However, each organization needs to register separately and should indicate if they are working as a team with another organization.
  


• Are there any compensation or honorariums associated with participating in this exercise? There is no financial compensation or other incentives provided to any participating organization, company, or affiliated representatives.
  

About the Exercise Scenarios
Exercise scenarios will be provided for two (2) selected locations with different topography and vegetation, each encompassing an area of approx. 5km X 5km. Participants are highly encouraged to run and submit their models for both of the two selected locations.

1. Each location encompasses a small area of wildland, a transition area in the WUI with vegetation capable of carrying fire, and an urban environment comprised of multiple residential subdivisions and some commercial properties.
2. Exercise locations are not sufficiently large for some wildland fire risk models to run, such as those that are designed for large-scale fires in the natural landscape exclusively.
3. Exercise locations are designed for models to run in areas where fire in the natural landscape propagates into the WUI and threatens built infrastructure.
4. Each location will have real spatial coordinates and will use real road networks.
5. Building data is synthetic however, it's being designed to simulate real building stock in each location but with specific building characteristic attributes within the data set. Additional details on exercise data are provided below.
6. Organizations with a wide range of wildfire risk models can participate, this includes both static and dynamic models:
7. Static Models: Typically, statistical/empirical-based models that provide a snapshot of relative risk. May (or may not) have been developed using dynamic methods and simulation repetitions.
8. Dynamic Models: Often physics, reduced-physics, or semi-physics based and are time-resolved fire progressions that are run for scenario-specific outcomes.
9. Risk Indices: That generate a ranking or score of risk, or a value of risk expressed in monetary/economic terms.
10. For Dynamic Models: Synthetic fire perimeters for wildland fires will be included in the 5km X 5km landscape area, thus assuming ignition and active (slow moving) fire. Thus, the exercise scenarios focus heavily on quantifying the risk of fire impacts (and their severity), risk of vegetation carrying fire from the landscape into the WUI, and impacting the built environment.
11. Two (2) distinct weather condition sets will be provided (one for each location) that represent peak fire weather conditions for those locations.
12. Participants have flexibility in how they can or decide to run their model based on the variables and attribute-level data provided.
13. Keep in mind, one of the objectives of the exercise is to understand is which models are sensitive to which variables and at which levels.
14. No single model to incorporate all of the variables and attributes.

About the Exercise Data Resources
A package of exercise data is being produced and will be made available to all participants at the start of exercise conduct in mid-March. Exercise data is being designed to ensure maximum flexibility by a diversity of different models.

Participants are not required to use all exercise data attributes provided, rather you should only use the attributes relevant to your model. Participants are required to use the core data provided, and are requested not to replace building data with an alternative source, thereby upholding the integrity of the exercise.

Below is summary of the exercise data resources that will be provided.

1. Building Data:
   • Shape-based files and GeoTIFFs for the buildings.
   • Building data will include a full set of building-level attributes ranging from square footage, year built, roofing materials, window type, and other multitude of other attributes that may be relevant to some models.
2. Surface Data: Rasterized file for the surface.
3. Vegetation Data: Raster layer for vegetation with unique vegetation types identified that is capable of being voxelized. Exact resolution for veg is to be determined.
   • Will include a list of individual trees on the landscape with the following details: height to crown, diameter of base, species, etc. This will be more extensive for one of the two areas.
   • A shrub list will not be provided; however, some shrub information will come through the surface data layer.
   • Participants have the freedom to innovate and come-up with tree load, etc.
   • Maximum flexibility for different types of models with different capabilities with vegetation.
4. Supplemental Data: An inventory of potentially relevant publicly available data for other features, which provide participants with to consider what and how they may want to include other variables.
   • Examples include wind turbine infrastructure, US electrical transmission lines, water resources/reservoirs, railroad networks, etc.
   • Participants are not expected to use supplemental data but are welcome to innovate as long as the core exercise vegetation data and building data are used where applicable by a model.
5. Example Notebooks: Example Notebooks are being developed for this exercise as templates and tools for participants to use in interpreting the exercise data, preparing their data pipelines, etc.
   • The Example Notebooks will include a piece of code for reading the vegetation of the voxelized trees, as a tool to extract the data.

Organizations can participate as teams in the exercise. However, each organization needs to register separately and should indicate if they are working as a team with another organization.