Standardized Criteria and Constraints
The MCE model used to predict future spread of Hemlock Woolly Adelgid incorporated spread through road vectors using the distance from roads (where low distances were better for spread) and road size to indicate traffic (where major roads were better for spread). In addition, proximity to current HWA and temperature under two emissions scenarios were incorporated in the MCE model, with lower distances from current HWA distributions and higher temperatures being more conducive to spread. The spread of the invasive species into Canada is limited by mean winter temperatures below -10 degrees Celsius. This was factored into the MCE by reclassifying the data for both scenarios as a constraint. All other criteria were standardized on a scale of 0-100. Figure 4 and 5 represent the final standardized product of the criteria and constraints before their input into the MCE.
Figure 4. Standardized criteria and constraints required for the MCE model for climate scenario RCP 4.5.
Figure 5. Standardized criteria and constraints required for the MCE model for climate scenario RCP 8.5.
Vulnerability of Eastern Hemlock
The multi-criteria evaluation model yielded a layer showing susceptibility to spreading for both climate scenarios. These layers were then multiplied with the Eastern Hemlock density. The resulting layer represents areas of Eastern Hemlock vulnerability to the spread of Hemlock Woolly Adelgid for both climate scenarios. Figure 6 and 8 show the vulnerability for the RCP 4.5 and the RCP 8.5 scenarios, respectively. In addition, after confirming the presence of hot spots using Moran's I (where p-value = 0.000391), a hotspot analysis was conducted on the areas of vulnerability for both scenarios, and the results are shown in Figure 7 and 9.
The final output represents the most vulnerable Eastern Hemlocks as bright red and the least vulnerable as dark blue. The yellow areas of Hemlocks represent the threshold for concern for management and monitoring purposes. Several trends were observed for the vulnerability of Eastern Hemlock under the lower emissions climate scenario RCP 4.5. First, areas of highest vulnerability occurred where the Hemlock density was the greatest, especially along the Eastern shore of Georgian Bay, as seen by the clusters of red Hemlocks in Figure 6. In addition, Figure 6 shows that clusters of high vulnerability represented by orange follow along roads, indicating that vulnerability is higher along road vectors, due to human movement facilitating their spread. Finally, the vulnerability is generally higher in the south and decreases towards the north, with the highest possible spread occurring in Quebec at latitude 60o41.88 S. Vulnerability is lowest in the northern portion of the study area because mean winter temperatures are lower in the North affecting HWA survival. Distance from the current distribution of HWA also increases with higher latitudes, meaning that spread is less likely further North. However, there are a few outliers which do not follow these trends, such as the few areas of high vulnerability among low vulnerability in the Northwestern extent of Hemlock distribution.
The hot spot analysis for the vulnerability of Eastern Hemlock under the RCP 4.5 scenario identified coldspots as blue and hotspots as red, with yellow values indicating there were no clusters. The results from this analysis, shown in Figure 7, indicate that the most hot spots occur East of Georgian Bay. Cold spots occur in New Brunswick and Nova Scotia, as well as just West of Lake Ontario. There is also an isolated area of hot spots in the Northwestern extent of Hemlock distribution.
Figure 6. Vulnerability of Eastern Hemlock to the spread of Hemlock Woolly Adelgid under climate scenario RCP 4.5.
Figure 7. Hotspots of Eastern Hemlock vulnerability under climate scenario RCP 4.5.
The vulnerability of Eastern Hemlock under the climate scenario RCP 8.5 yielded very similar results to the RCP 4.5 scenario, as shown in Figure 8. The highest vulnerability occurs in the same areas and follows sections of high Hemlock density. Additionally, high vulnerability occurs along road vectors and vulnerability decreases further North in the RCP 8.5 scenario as well. This climate scenario also showed the same outliers of high vulnerability in the Northwestern extent.
Figure 9 shows vulnerability hotspots for the RCP 8.5 scenario, and again these results closely follow the ones from RCP 4.5. Land East of Georgian Bay has a large number of hot spots, whereas cold spots occur mostly in New Brunswick and Nova Scotia, as well as just West of Lake Ontario. Again, there is an isolated section of hot spots in the Northwestern extent.
Figure 8. Vulnerability of Eastern Hemlock to the spread of Hemlock Woolly Adelgid under climate scenario RCP 8.5.
Figure 9. Hotspots of Eastern Hemlock vulnerability under climate scenario RCP 8.5.
The two resulting layers from the MCE model showing spread under the lower emissions RCP 4.5 and the higher emissions RCP 8.5 scenarios were also compared to identify differences in spread and the impact that climate change might have. It is expected that the higher emissions scenario (RCP 8.5) would allow for greater and further spread North than the lower RCP 4.5 scenario since higher emissions result is higher temperatures. They were compared using subtraction with the raster calculator, and the difference in spread is shown below in Figure 10. The map represents areas with no difference in the spread as green, areas with the least difference in yellow and the most difference in red. As seen in Figure 10, there was no difference between the spread in the Southern extent of the study area, and only a slight difference in the north, where spread under RCP 8.5 extends slightly farther along the northern edges. This increase in spread to the north under the higher emissions scenario is a result of slightly higher mean winter temperatures causing the constraint of -10 degrees Celsius to move farther north. The average increase of spread to the north was found to be 12km.
Figure 10. Comparison between spread under RCP4.5 and RCP8.5 scenarios.
Strengths and Weaknesses of Model
A few limitations to this model include the inability to account for spread through animal and wind vectors. Birds are an important factor in the spread of HWA as they can transport them long distances in a short time. However, it was difficult to include bird migration data in the MCE. Literature also shows that HWA travels northward at a rate of 12.5km/yr, however, this time element was difficult to incorporate into the MCE model (Evans and Gregoire, 2006). Additionally, it is challenging to quantify the contribution of wind to spread and therefore this factor was too complicated to include in the MCE as well. Furthermore, a main limitation of the model was the lack of a high-quality Eastern Hemlock inventory for Canada. This limitation has been identified and currently, two bodies of research are independently compiling this inventory so that stands at most risk can be monitored. Further benefits to this inventory include identification of disjunct populations of hemlock, where hemlock is less likely to be infected due to the isolation from other hemlock populations (Hart, 2008). However, despite the lack of higher quality Eastern Hemlock inventory, the Hemlock distribution and density layer used for the model was still able to provide results which showed vulnerable areas and therefore can still be of use for management and monitoring of HWA.
The model used for the analysis also had several strengths; it efficiently combined risk for spread along roads based on distance and traffic with the risk of spread under different temperatures and different distances from the current distribution. This allowed for the clear identification of individual areas of high and low vulnerability to spread, as well as hot spots of vulnerability requiring more immediate attention. In addition, although only slight differences were found between the best case climate scenario, RCP 4.5, and the worst case scenario, RCP 8.5, the model provided a clear method of comparing the two.