A data complexity reduction analysis was performed to determine the associations between the wildlife visitation rates and urban factors. The analysis focused on the five common and rare species of mammals and birds identified as the standardized response variables. 11 geographical variables were considered as predictor variables. To simplify the data, Principal Component Analysis (PCA) was utilized, followed by direct gradient analysis, ordering first the urban factors and then the standardized wildlife visitation rate values (Figure 17).

No clear group differentiation is evident among all urban factors (Figure 17). However, the second component (Comp. 2) appears to be associated with the presence of humans and their transportation activities on trails (vectors Nearest Trail [NTr] and trails density [TD]), as well as the distance from highly urbanized areas like the city center (vector DCC). The first component (Comp. 1) distinguishes monitoring sites located in areas that have undergone significant transformation (Comp. 1 < 0.0) or areas where natural land covers and low transformation are prevalent (Comp. 1 > 0.0), as depicted on the map in Figure 17 (right). This is supported by the examination of vectors of forested areas (WOOD) and natural and water areas (WAT), which are diametrically opposed to transformed zones like urban parks (MOD) or completely urbanized areas (DEV). 

After analyzing the relationship between urban factors and standardized wildlife visitation rates for both common and rare species, it was discovered that most common bird and mammal species are strongly associated with component 1. This is demonstrated by the almost complete overlap of the green (common mammals) and blue (common birds) vectors with some urban factor vectors in the direction of this component. Larger mammals, such as coyotes and White-tailed deer, were more frequently found in natural areas, while smaller mammals, such as red squirrels, were more correlated with urban factors indicating transformation, such as trail density, making them more common in cities. The two Lepus species (White-tailed jackrabbit and snowshoe hare) were found to be correlated with distance to the nearest water body (NWB) and forested areas (WOOD). Only the Mallard, among the common bird species, tended to be spotted in natural areas dominated by forests, while others appeared to be more comfortable in urbanized areas. 

Rare species also showed a similar preference for component 1, particularly for birds. It is noteworthy that in this case, most of the rare mammal species, such as cougar, lynx, and American badger, were more related to component 2, but the size of the vectors indicated that the vectors in this component that represent human mobilization are those that these animals would try to avoid. This was confirmed by analyzing the original data frame (Table 1), where these animals were detected in sites such as 89 and 37, which are located on the periphery, far from downtown in naturally wooded areas (Figure 17 - right). Lastly, it is important to note that sites located in the negative values of both components (Comp 1<0.0 and Comp 2<0.0), which are opposite to DCC and NTr vectors, and where no common and rare species tend to be spotted, indicate that the closer the monitoring site is to downtown, the less likely it is to detect common and rare species. 

Table 2 shows that only the analysis of common species has statistical significance (p-value < 0.05). For rare species, the lack of records in the monitoring data affects the significance of the conclusions. As previously mentioned, most species are correlated with component 1, with 60% of species specifically showing this correlation. Almost all the mammal models were significant, with only the red squirrel showing no significance. However, the red squirrel has a strong negative correlation with component 2, particularly with trail density (TD), where they are typically spotted. Among birds, only the crow and magpies were significant common species with a strong negative correlation with component 1. The magpie is typically found in urbanized areas, while the crows are found near water bodies. Although the rare species analysis was not significant, Table 2 values confirm that rare species, especially birds, are more associated with component 1. 

In order to determine thresholds in urban areas that could be useful in making decisions about future monitoring or urban planning strategies, it was used multivariate regression tree analysis (MRT) to identify groups of common mammal species (Figure 18-a) and bird species (Figure 18-b) that were detected in different monitoring sites, while imposing the predictor variable (urban feature) as a constraint to create clusters. The main goal of this analysis was to select the predictor factor and threshold that explained the variance in the wildlife visitation rate of the common species. By doing so, it became possible to identify values in urban factors that could help identify more or fewer individuals of different species. This information can be used to support new monitoring sites, such as increasing camera trapping in areas further from trails so that other species can be identified, or to develop new site-specific strategies, such as ecological restoration to increase visitation.

This analysis focused solely on the common species that exhibited statistical significance (Table 2). In terms of mammals, the presence of forested areas (WOOD) and proximity to trails (NTr) were found to be the most significant urban factors influencing the frequency of species detections. Most common species were detected when the forest surrounding the monitoring sites exceeded 5%, except for the red squirrel, which displayed more activity in less forested areas (< 5%). Larger mammals such as coyotes and white-tailed deer were detected more frequently when the distance to trails was greater than 40 m, whereas smaller mammals were detected in sites closer to trails (< 40 m). However, the snowshoe hare appeared to be more adaptable, as it was detected in sites further from trails too.

As for birds, three factors seemed to influence the frequency of detection for the most commonly found species. Although the direct gradient analysis indicated some adaptation of birds to urbanized areas, the MRT analysis showed that these species were only detected in 9 sites where the surrounding area had over 60% transformed land cover (MOD), and detection increased if these sites were located more than 11 km from the city centre. Therefore, there may be other factors affecting bird detection rates, such as the percentage of water bodies surrounding the sites, where mallards were detected if the percentage of water bodies was nearly 50%, as observed in two sites (n = 2).

The aforementioned MRT analysis should be approached with caution, as the error exceeded 60%. Nevertheless, this is the first insight that the WildEdmonton monitoring program has obtained regarding which urban factors can influence visitation rates. These results can serve as a baseline for decision-making in municipal planning.

• For research projects that involve multiple predictor and response variables, multivariate statistical analyses can be an effective approach. In this particular study, data complexity reduction tools such as PCA and direct gradient were used to establish links between 11 predictor variables (urban factors) and 20 response variables (10 common and 10 rare species of mammals and birds). Additionally, multivariate regression methods (MRT) were employed to predict thresholds in urban features that could explain the variation in wildlife visitation rates for common species. 

• This project included a geographical analysis of normalized biodiversity visitation data of mammals and birds detected in the City of Edmonton through the WildEdmonton program from 2018-2021. The data was classified into three groups based on wildlife visits/day. The North Saskatchewan River Valley had the highest number of individuals detected per day for both mammals and birds. Mammals' visitation rates remained highly concentrated around the North Saskatchewan River Valley, whereas bird detection per monitoring day had low spatial variability, with almost all monitoring sites recording the same value per day.

• The direct gradient analysis with PCA illustrating the relationship between urban factors and wildlife visitation rates shows that the presence of humans and their transportation activities on trails, as well as the distance from highly urbanized areas, are associated with the second component. Meanwhile, the first component distinguishes monitoring sites located in areas that have undergone significant transformation or areas where natural land covers and low transformation are prevalent. Common bird and mammal species are strongly associated with component 1, while larger mammals tend to be found in natural areas, and smaller mammals are more correlated with urban factors indicating transformation. Rare species also showed a preference for component 1, particularly for birds. It is noteworthy that rare mammal species were more related to component 2, but vectors in this component representing human mobilization are those that these animals would try to avoid. Sites located in negative values for both components indicate that the closer the monitoring site is to downtown, the less likely it is to detect common and rare species. Overall, these findings can help inform urban planning and conservation efforts aimed at preserving wildlife in urban areas. 

• The multivariate regression analysis illustrated that common mammal species are more easily detected at sites surrounding at least 5% forest and located more than 40 m away from trails. Common bird species are more easily detected at sites surrounding more than 60% transformed land covers (MOD) and located at least 11 km away from the city center. Sites surrounded by almost 50% water bodies could potentially detect more birds. 

• The research findings should be interpreted with caution due to some limitations in the data used. One limitation is that not all monitoring sites were active for the same period. Additionally, spatial variability may have influenced the results since most of the sites are located along the North Saskatchewan River Valley. Climate seasonality was not considered in the analysis, which is a factor that can affect wildlife visitation and alter the species detected during different seasons. However, to address some of these limitations, the data was normalized by the number of days each site was active, and data transformations (inverse and square root) were applied to ensure that all response variables had the same weight in the multivariate analysis. It is important to consider these limitations when interpreting the results and to use them to complement more detailed studies. 

• Overall, the multivariate statistical analyses conducted in this study have provided important insights into the complex relationships between urban factors and wildlife visitation rates of common and rare species. The findings of this study can be used as scientific support for decision-making in the formulation of future monitoring or urban planning strategies, such as increasing camera trapping in some areas further from trails to identify other species or implementing site-specific strategies like ecological restoration to increase visitation.