Chapter 23 Hazard Model 2.0

23.1 What is the Hazard Model 2.0?

The Hazard Model 2.0 incorporates baseline spatial proximity risk plus many possible in-situ risk factors for CWD introduction and exposure of wild cervids to draw a landscape map of hazards. The hazard landscape is used to proportionately distribute sampling quotas across states, provinces, or indigenous nations.

23.2 What Questions Does the Model Answer?

Question 1. What is the baseline spatial risk due to proximity to the closest known infection? The geographical location of each sub-administrative unit may matter when it comes to CWD introduction. The Hazard Model 2.0 allows you to see the risk that each area will turn positive next year based simply on the geographical distance to the nearest known positive detection, wherever that detection may be.

Question 2. How can I identify areas with the highest CWD risk? The Hazard Model 2.0 incorporates all pertinent risks factors for each sub-administrative area, allowing the user to pinpoint areas of particularly high risk. The cumulative summaries are telling; higher the number, the higher the risk.

The following risk factors may be used to calculate total risk:

  • Natural cervid movement

    • Dispersing free-ranging cervids

  • Natural congregation

    • Mineral licks
    • Water bodies
    • Seasonal migration

  • Artificial congregation

    • Feedgrounds
    • Guzzlers or water tanks
    • Baiting stations
    • Agricultural practices

  • Business activities

    • Captive cervid facilities (e.g., cervid farms and hunting enclosures)
    • Taxidermists
    • Meat processors
    • Wild cervid rehabilitation centers
    • Carcass rendering facilities
    • Incineration facilities
    • Venison food banks

  • Individual human actions

    • Landfill disposal of imported carcasses
    • Carcasses dumped into the environment
    • Shed antler collection
    • Roadkill collection
    • Local practices

Question 3. How can I find the areas with the highest hazard? Prions must exist before specific risk factors matter. For example, if there are no prions in the system, then it doesn’t matter that cervids are congregating in a location, even though congregation is listed as a risk factor. Rather the risk becomes a hazard when prions enter the equation, perhaps though spatial spread. Thus, the hazard is the total risk factors in an area, actuated by the magnitude of the baseline spatial risk.

Question 4. How much does each risk factor influence the total hazard? Some risk factors drive the overall hazard. The sensitivity analysis on the visualizations page allows you to see which risk factors or weights might have disproportionately higher influence over the total hazard. The risk factor with the largest sensivitity constitues the most influential parameter in the model.

Question 5. How do I distribute sampling effort across my jurisdiction to target areas where CWD is most likely to appear or increase in prevalence? The administrative area-wide sampling point quota, as provided by the user, is distributed proportionately per sub-administrative area based on the hazard scores. You assign the appropriate surveillance category to each sub-administrative area according to CWD status and other risk factors. You choose the level of sampling (i.e., number of points) and the model proportionately distributes samples.

Question 6. How do surveillance points convert to the number of cervids my team needs to sample? Surveillance points per sub-administrative area can be used by agencies to determine their surveillance plans. Points, rather than counts, may be used to focus sampling efforts on sex and age classes in which CWD is most likely to be detected. You can select your own risk weights in the Annual Surveillance Collection to guide your surveillance efforts.

Note: The Hazard Model 2.0 does not produce surveillance targets based on statistical confidence. If you wish to determine sample sizes to reach statistical assurance that CWD is absent in an area, use the Sample Size Quotas Using Clustering Model or the Efficient Sample Size Calculator.

Note: The Hazard Model 2.0 does not consider cost. If you wish to determine sample sizes to minimize cost, use the Sample Allocation Model.

23.3 Output Details

Comparative maps, data charts, and tables of the following outputs by sub-administrative area may be viewed on the Visualization page. In addition, the model execution page allows you to download your tabular outputs as .csv or .json files.

  • Baseline Spatial Risk: Risk arising from spatial proximity to the nearest known CWD infection
  • Total Risk Score: Risk arising from tallied landscape or herd features
  • Total Hazard Score: Hazard that considers both the baseline spatial risk and the total risk score
  • Sensitivity Analysis by Sub-Admin Unit: The influence that each risk factor has on the overall hazard
  • Sensitivity Analysis all Sub-Admin Units: A comparison of influential risk factors across sub-administrative areas.

23.4 Abbreviated Tutorial

For best results, enter all possible data into the Warehouse:

  1. Enter as much agency data as possible into the sample, landscape survey, hazard risk weights, disposal, cervid facility, processor, demography, seasonal migration, and congregation data collections.
  2. Select Hazard Model 2.0 and open a new model run.
  3. Enter your desired parameters, as detailed below.
  4. If desired, create Visualization associated with the model run.
  5. Explore the model logs, input file, and output files used in the run.
  6. If the model did not run, check the model logs to understand required data that was missing.
  7. If desired, use the model results in an Annual Surveillance Collection entry in combination with a chosen weighting scheme.

23.5 Parameters Needed to Execute the Model

  • Reference name: Enter a name to help you identify your model run for future reference. Note that this field does not affect the model run.

  • Applicable season year (Optional): Select the season year (for your reference) from the drop-down list. Note that this field does not affect the model run.

  • Notes (Optional): Enter any notes about the model run details. Note that this field does not affect the model run.

  • Risk factors: Select the set of risk factors relevant in your area of interest. Many of these risk factors depend on data existing in the corresponding data collections. If you do not have data for a risk factor, then the Warehouse will not let you select the risk factor at all. If data does exist for a risk factor, then you can select the exact dataset to use for that risk factor. Other risk factors arise from continental geospatial data that already exists in the Warehouse. For these risk factors, you simply need to click the check box to tell the model to consider this risk factor or not. Note: You must choose at least one risk factor to run the model.

  • Risk weights: Select the set of risk weights that are relevant in the area of interest. You may either use a pre-set template (Pacific Northwest or Arid Southwest) or you may create your own set of risk weights using the Hazard Risk Weights Data Collection.

  • Sub-administrative areas: Enter the set of sub-administrative areas that you wish to consider in the model. Note: You must choose at least one area to run the model.

  • Seasonal migration: If data are available, select the seasonal migration paths to consider in the model.

Note: To enable the Hazard Model 2.0 to give you the most useful results, it is important to enter as much information as you can in all the data collections.

23.6 Output Details

To access the output, go to the “Previously executed models” page from the Warehouse navigation sidebar and identify the successful model execution whose output you’d like to save. Click the icon of three dots in the Attachments column of the corresponding row, which will show a drop down list of files generated by the model. File content will vary depending on the model specifications.

23.6.1 Results

File: OutputHazards.csv

Output description: A data table containing all the sub-admin units used in the model with the following columns:

  • SubAdminID - Unique Warehouse identifier for sub-administrative area
  • FullName - Name of sub-administrative area, if available, e.g. Tompkins County or DMU 4
  • BaselineRisk - Risk associated simply with the spatial proximity to the nearest known infection, wherever that may be
  • TotalRisk - Risk associated with landscape, herd, and habitat features
  • TotalHazard - Total hazard created by integrating baselines risk with total risk
  • e_Baseline - The elasticity (influence) of the baseline risk on the total hazard
  • e_FreeRangingCervidDispersal - The elasticity (influence) of cervid dispersal on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_MineralLicks - The elasticity (influence) of mineral licks on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_WaterBodies - The elasticity (influence) of water bodies on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_SeasonalMigration - The elasticity (influence) of seasonal migration on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Feedgrounds - The elasticity (influence) of feed grounds on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Guzzlers - The elasticity (influence) of guzzlers on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_AgriculturalPracites - The elasticity (influence) of agricultural practices on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_CaptiveCervidFacilities - The elasticity (influence) of captive cervid facilties on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_RehabilitationFacilities - The elasticity (influence) of rehabilitation facilities on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Taxidermists - The elasticity (influence) of taxidermists on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Processors - The elasticity (influence) of processors on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_RenderingFacilities - The elasticity (influence) of rendering facilities on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Incinerators - The elasticity (influence) of incinerators on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Foodbanks - The elasticity (influence) of foodbanks on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Landfill - The elasticity (influence) of landfills on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Dumping - The elasticity (influence) of dumping areas on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Sheds - The elasticity (influence) of shed collection areas on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • e_Roadkill - The elasticity (influence) of roadkill on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_FreeRangingCervidDispersal - The elasticity (influence) of the expert weight given to cervid dispersal on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_MineralLicks - The elasticity (influence) of the expert weight given to mineral licks on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_WaterBodies - The elasticity (influence) of the expert weight given to water bodies on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_SeasonalMigration - The elasticity (influence) of the expert weight given to seasonal migration on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Feedgrounds - The elasticity (influence) of the expert weight given to feed grounds on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Guzzlers - The elasticity (influence) of the expert weight given to guzzlers on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_AgriculturalPracites - The elasticity (influence) of the expert weight given to agricultural practices on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_CaptiveCervidFacilities - The elasticity (influence) of the expert weight given to captive cervid facilties on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_RehabilitationFacilities - The elasticity (influence) of the expert weight given to rehabilitation facilities on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Taxidermists - The elasticity (influence) of the expert weight given to taxidermists on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Processors - The elasticity (influence) of the expert weight given to processors on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_RenderingFacilities - The elasticity (influence) of the expert weight given to rendering facilities on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Incinerators - The elasticity (influence) of the expert weight given to incinerators on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Foodbanks - The elasticity (influence) of the expert weight given to foodbanks on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Landfill - The elasticity (influence) of the expert weight given to landfills on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Dumping - The elasticity (influence) of the expert weight given to dumping areas on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Sheds - The elasticity (influence) of the expert weight given to shed collection areas on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.
  • ew_Roadkill - The elasticity (influence) of the expert weight given to roadkill on the total hazard. Note: This elasticity will appear as zero if the risk factor is not selected by the user.

23.6.2 Visualizations

For more details about how to create a visualization refer to the Hazard Model 2.0 Visualization page. Each visualization page is split in half where the top half has a data visualization and the bottom half has a corresponding data table. The model property defines the values or results shown.

23.7 Details on the Theory

Thompson N, Sernaker S, Hanley B, Cook J, Hollingshead N, Hubbs A, Reed H, LaHue N, Lieske C, Gillin C, Reeder A, Munk B, Wood L, Justice-Allen A, Duvuvuei O, Crockett E, Wycoff S, DeVivo M, Westacott H, Cook W, Heffelfinger L, Wild M, Epps C, Walsh D, Nelson C, Thacker C, Beckmen K, Schuler K. The Hazard Model 2.0: Extension of a risk-based chronic wasting disease surveillance model to the western United States and Canada. In preparation

23.8 Code

The code is publicly available at https://github.com/Cornell-Wildlife-Health-Lab/Hazard-Model-2.0.