The Threat of the Death Cap in Southwest British Columbia
Results and Maps
The distribution of the Amanita phalloides sites in Vancouver can be seen on the interactive map below, with a total of 66 sites. By clicking on each of the red symbols a pop-up with relevant information appears; such information includes tree IDs, flourishing number, tree species, diameter, size class, date planted, height range and 3 class type. The size class can be useful in determining the relative age of trees, which can be a key factor in predicting and analyzing the dispersal of the mushroom, classes range from small, to medium, large, very large, and huge (Starbucks). Most of the trees being from the Carpinus betulus species, and two cases of Quercus rubra and Fagus sylvaticus.
​
By clicking on the top left arrow in the left top corner of the map, it is possible to toggle between the different available layers. One of the layers represents the 31 schools which are within 500m of each respective site. It is possible to find their names, and could potentially be useful in deciding where to conduct an information session. Through the map, it is also possible to search for specific addresses, and possible to change the base map options.
Amanita phalloides Sites in Vancouver, BC, Canada
Note: if the interactive map does not appear, access the map by clicking on the link beside the title above. To access the map through the link you must have a free ESRi account.
​
In addition to the interactive map, a static map representing the sites of the death cap locations was further constructed in order to support policy making and future decision making in regards to risk reduction and raise awareness. The construction of this map serves as a support for future overview and tracking of the death cap in Vancouver. It provides a basis for mycology club volunteers to observe, record, map and keep track of sittings around the city. The map can also serve to share public information about the threat and presence of the mushroom
Multi Criteria Evaluations (MCEs)
The above maps compare the risk associated with all known host trees. The left map shows the risk associated with all known locations (other than in parks) of the following tree species: English Oak, Sweet Chestnut, Beech, Filbert, Linden, Garry Oak and Northern Red Oak. The right map shows the risk associated with all known locations. The same Analytical Hierarchy Process (AHP) was used for both MCE outputs to ensure consistency in the weighting to show the true difference in the risk associated with all known host tree species locations (other than Hornbeam) compared to risk associated with all Hornbeam tree locations. As the maps show, there is greater risk area for the Hornbeam locations. The red zones in both maps highlight potential priority sampling locations for the fruition season.
The above map series attempts to show the change in risk of Death Cap presence under conditions of no lawn watering compared to current lawn watering levels. Both sensitivity analysis and realistic weighting (20% for demographic factors and 80% for proximity factors) outputs are included for comparison. These map outputs suggest that the total risk area decreases when there is greater lawn watering. However, these results are misleading. In retrospect, the weighting regime for the demographic factors for the excluding lawn watering maps should have been changed to not only discount the lawn watering factor but also change the other factors weightings to account for the loss of the lawn watering weight. If this had been done, the excluding lawn watering maps would surely have different outputs. Thus, further analysis is required for a more conclusive understanding of the effect of lawn watering, but the steps taken to produce the above maps can be used as road map to achieve this future understanding.
The above maps compare the sensitivity analysis, which was constructed with equal weights at every MCE level, to the output produced by the AHP weighting regime. This shows the change in risk levels based on the weighting regimes. As the left map shows, when all weights are equal the total area of high or highest risk is very large. In comparison, the right map, suggests that the total area of high or highest risk is quite small and restricted to more finite points as (opposed to large areas) when the weighting of the factors is not equal and more realistic.
​
This map series shows the maps that contribute to achieving our Final Output. As the map sequence shows, the weighting regime is based on the Analytical Hierarchy Process. The Final Output serves as our best guess for the areas in Vancouver with the greatest risk. As the map shows, a large portion of the city is at least at moderate risk of Death Cap presence. The red and orange highest risk and high-risk zones show the areas where future research should be focused.
Maxent Analysis
Click on figure to enlarge​
Figure 1. Martclip is the temperature in March and contributed the most. Vegcover is vegetation cover and contributed the second highest.
​
A jackknife test to show individual variable contributions to training gain if they were used in isolation was also conducted. Below are the results:
Click on figure to enlarge​
This model above shows the potential distribution of the Death Cap mushroom based on its sighting in Mission, B.C. The model varies quite drastically compared to the other model. This is due to the very small sample size of one used to train this model. Given more information about the extent of the mushroom in Mission could yield better results. The environmental layers utilized are precipitation, temperature, soil water stress, evapotranspiration, surface material, and vegetation cover. The table below shows how each variable contributed to the model.
Figure 2. Evapotranspiration was the only variable with a meaningful contribution to the model output. This is likely due to the specific locality of the sample used and the resolution of the environmental layers which was 1000 meters.
​
A jackknife test to show individual variable contributions to training gain if they were used in isolation was also conducted. Below are the results. As is clear through the jackknife test this model is heavily biased and would need more samples to perform well.
Click on figure to enlarge
This model shows the potential distribution of the Death Cap mushroom based on sightings and ingestion cases in Victoria, B.C. The model was trained using 6 sample sites around the Victoria area including Oakbay, Government House, Foul Bay road, Richmond Avenue, Crystal Gardens, and Eastdowne road. The environmental layers utilized are precipitation, temperature, soil water stress, evapotranspiration, surface material, and vegetation cover. The table below shows how each variable contributed to the model.
Click on figure to enlarge​
Figure 3. February temperature and precipitation in July are the most important variable in this model. Variations between models exist due to the different environmental factors present at the sample species localities.
​
A jackknife test to show individual variable contributions to training gain if they were used in isolation was also conducted.
Below are the results:
Click on figure to enlarge​
This map shows trees species that could possibly host the Death Cap mushroom. The top 6 were chosen based on model performance results with trees that had the highest probability of being present in a habitat being chosen. The environmental layers utilized are precipitation, temperature, soil water stress, evapotranspiration, surface material, and vegetation cover. The table below shows how each variable contributed to the model. Overall, these models performed well and drew predictions from several of the environmental layers supplied to Maxent.
The results of variable contributions and jackknife tests is shown in this document:
Go to Dicussion
The first Maxent model shows the predicted habitats of the Death Cap mushroom based on the verified locations of the mushroom in the document supplied to us from the CDC via Vivian Mao. The environmental layers utilized are precipitation, temperature, soil water stress, evapotranspiration, surface material, and vegetation cover. The table below shows how each variable contributed to the model.
