Beating Around the Bush: Uncovering the Hidden Link Between Urban Trees and Wildlife Activity

Introduction

Urban trees play an important role in shaping city environments. Beyond improving air quality and adding color to city streets, street trees can also provide shelter, food sources, and movement pathways for wildlife living in urban areas. At the same time, cities are fragmented environments where wildlife often comes into contact with human infrastructure, increasing the risk of injury or the need for human intervention.

The diversity of New York City provides an insightful setting for analyzing these dynamics, as each borough differs in tree coverage, wildlife presence, and urban structure. Some boroughs have more street trees and larger parks, while others are more densely developed. This variation allows for comparisons between tree distribution and reported wildlife incidents across the city.

Using NYC Open Data from the 2015 NYC Street Tree Census and the Urban Park Ranger Animal Condition Response dataset, this project explores the relationship between street trees and wildlife incidents across New York City boroughs. Specifically, this analysis asks:

1. how street trees are distributed across boroughs
2. how wildlife incidents vary by borough
3. whether borough-level tree characteristics are associated with wildlife incident patterns

This project uses descriptive and exploratory analyses to examine how urban trees and wildlife activity coexist within a large metropolitan city such as New York City.

Required Packages

For this project, I used the following R packages to clean, analyze, and visualize the data.

library(tidyverse)
library(janitor)
library(skimr)
library(ggplot2)
library(knitr)
library(supernova)
library(rcompanion)

Data and Methods

Data Sources

This analysis uses two publicly available datasets from NYC Open Data. The first dataset is the 2015 NYC Street Tree Census, which documents the location and characteristics of street trees across New York City. For this study, the Street Tree Census was used to calculate the total number of street trees in each borough, serving as a borough-level measure of street tree abundance.

The second dataset is the Urban Park Ranger Animal Condition Response dataset, which records requests for wildlife assistance, relocation, and rescue responded to by NYC Park Rangers. This dataset was used to calculate the total number of wildlife-related incidents in each borough, serving as a proxy measure of wildlife activity.

Although the two datasets capture different aspects of the urban environment—street trees located primarily along streets and wildlife incidents occurring primarily within park properties—they were analyzed together at the borough level to explore broader spatial patterns in urban ecology across New York City. All analyses were conducted using aggregated borough-level summaries.

I began by downloading publicly available datasets from NYC Open Data and importing them into R as CSV files for data cleaning, analysis, and visualization.

# Import NYC Open Data CSV files
tree_data <- read_csv("2015_Street_Tree_Census.csv")

## Rows: 683788 Columns: 45
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (26): created_at, curb_loc, status, health, spc_latin, spc_common, stewa...
## dbl (15): tree_id, block_id, tree_dbh, stump_diam, postcode, community board...
## num  (4): boro_ct, x_sp, y_sp, census tract
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

animal_data <- read_csv("Urban_Park_Ranger_Animal_Condition_Response.csv")

## Rows: 7613 Columns: 22
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (15): Date and Time of initial call, Date and time of Ranger response, B...
## dbl  (3): Duration of Response, # of Animals, Hours spent monitoring
## lgl  (4): PEP Response, Animal Monitored, Police Response, ESU Response
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.

Data Cleaning and Preparation

Before analysis, I cleaned and prepared the data by selecting relevant variables, addressing missing values, and ensuring consistent formatting across datasets.

#Standardize column names
tree_data <- tree_data %>% clean_names()
animal_data <- animal_data %>% clean_names()

# Select variables needed for tree-level analysis
tree_clean <- tree_data %>%
  select(borough, health, tree_dbh, latitude, longitude)

# Select variables needed for wildlife incident analysis
animal_clean <- animal_data %>%
  select(borough, species_description, animal_class, animal_condition, property)

# Check for missing borough values
sum(is.na(tree_data$borough))

## [1] 0

sum(is.na(animal_data$borough))

## [1] 0

# Convert key categorical variables to factors and clean missing values
tree_clean <- tree_clean %>%
  mutate(
    health = factor(health, levels = c("Poor", "Fair", "Good"))
  )

animal_clean <- animal_clean %>%
  mutate(
    animal_condition = na_if(animal_condition, "N/A"),
    animal_condition = factor(animal_condition),
    animal_class = factor(animal_class)
  )

Descriptive Analysis (Plots)

To describe how street trees and wildlife incidents vary across New York City, I summarized the total number of urban trees and reported wildlife incidents in each borough and visualized these counts using a bar chart.

Street Tree Distribution Across Boroughs (Bar chart)

tree_summary <- tree_clean %>%
  group_by(borough) %>%
  summarise(
    total_trees = n(),
    pct_good_health = mean(health == "Good", na.rm = TRUE),
    avg_dbh = mean(tree_dbh, na.rm = TRUE),
    .groups = "drop"
  )

ggplot(tree_summary,
       aes(x = reorder(borough, -total_trees),
           y = total_trees)) +
  geom_col(fill = "chartreuse4") +
  labs(
    title = "Total Street Trees by Borough",
    x = "Borough",
    y = "Total Street Trees"
  ) +
  theme_minimal(base_size = 12) +
    theme(plot.title = element_text(size = 17, family = "serif", face = "bold"),
        axis.title.x = element_text(size = 12, family = "serif"),
        axis.title.y = element_text(size = 12, family = "serif"))

Total number of street trees recorded in the 2015 NYC Street Tree Census, summarized by borough.

Interpretation

The bar chart shows the total number of street trees in each New York City borough. Queens has the largest number of street trees, followed by Brooklyn. Manhattan has the fewest street trees among the five boroughs. Overall, the distribution highlights substantial differences in street tree availability across boroughs, reflecting variation in land area, residential density, and urban structure.

Wildlife Incidents Across Boroughs (Bar chart)

animal_summary <- animal_clean %>%
  group_by(borough) %>%
  summarise(
    total_incidents = n(),
    pct_injured = mean(animal_condition == "Injured", na.rm = TRUE),
    .groups = "drop"
  )

ggplot(animal_summary,
       aes(x = reorder(borough, -total_incidents),
           y = total_incidents)) +
  geom_col(fill = "salmon4") +
  labs(
    title = "Wildlife Incidents by Borough",
    x = "Borough",
    y = "Total Incidents"
  ) +
  theme_minimal(base_size = 12) +
  theme(plot.title = element_text(size = 17, family = "serif", face = "bold"),
        axis.title.x = element_text(size = 12, family = "serif"),
        axis.title.y = element_text(size = 12, family = "serif"))

Total number of reported wildlife incidents across NYC boroughs.

Interpretation

The bar chart displays the total number of reported wildlife incidents across New York City boroughs. Manhattan reports the highest number of incidents, followed by Queens and Brooklyn. The Bronx and Staten Island report fewer incidents overall. These differences highlight variation in reported wildlife activity across boroughs, which may reflect differences in park usage, population density, and reporting patterns.

Combining Tree and Wildlife Data at the Borough Level (Table)

After examining street trees and wildlife incidents separately, I combined the two borough-level summaries to explore how tree abundance and wildlife incidents relate to one another. By merging the tree and wildlife datasets by borough, this analysis allows for direct comparison between the number of street trees and the frequency of reported wildlife incidents across New York City.

borough_summary <- left_join(tree_summary, animal_summary, by = "borough")

summary(borough_summary)

##    borough           total_trees     pct_good_health     avg_dbh      
##  Length:5           Min.   : 65423   Min.   :0.7586   Min.   : 8.474  
##  Class :character   1st Qu.: 85203   1st Qu.:0.8142   1st Qu.: 9.694  
##  Mode  :character   Median :105318   Median :0.8149   Median :10.493  
##                     Mean   :136758   Mean   :0.8059   Mean   :10.591  
##                     3rd Qu.:177293   3rd Qu.:0.8152   3rd Qu.:11.739  
##                     Max.   :250551   Max.   :0.8265   Max.   :12.558  
##  total_incidents  pct_injured    
##  Min.   :1151    Min.   :0.2168  
##  1st Qu.:1221    1st Qu.:0.2443  
##  Median :1614    Median :0.2834  
##  Mean   :1523    Mean   :0.2738  
##  3rd Qu.:1684    3rd Qu.:0.2921  
##  Max.   :1943    Max.   :0.3327

kable(borough_summary, caption = "Borough-level summary of street tree characteristics and wildlife incident counts across New York City.")

Borough-level summary of street tree characteristics and wildlife incident counts across New York City.

borough	total_trees	pct_good_health	avg_dbh	total_incidents	pct_injured
Bronx	85203	0.8264938	9.693649	1221	0.2167577
Brooklyn	177293	0.8142379	11.738884	1614	0.2920723
Manhattan	65423	0.7586141	8.473641	1943	0.2442936
Queens	250551	0.8152487	12.557870	1684	0.2833856
Staten Island	105318	0.8149386	10.492746	1151	0.3326829

Wildlife Incidents Relative to Street Tree Availability (Standardized bar chart / rate per 10,000 trees)

To account for differences in street tree availability across boroughs, wildlife incidents were standardized by tree count and expressed as incidents per 10,000 street trees.

borough_summary <- borough_summary %>%
  mutate(incidents_per_10k_trees = total_incidents / total_trees * 10000)

ggplot(borough_summary,
       aes(x = borough, y = incidents_per_10k_trees)) +
  geom_col(fill = "dodgerblue4") +
  labs(
    title = "Wildlife Incidents per 10,000 Street Trees by Borough",
    x = "Borough",
    y = "Incidents per 10,000 Street Trees"
  ) +
  theme_minimal(base_size = 12) +
  theme(plot.title = element_text(size = 17, family = "serif", face = "bold"),
        axis.title.x = element_text(size = 12, family = "serif"),
        axis.title.y = element_text(size = 12, family = "serif"))

Number of wildlife incidents per 10,000 street trees in each New York City borough.

Interpretation

The bar chart shows the number of wildlife incidents per 10,000 street trees in each borough. Manhattan has the highest rate of incidents per tree, even though it has fewer street trees overall. In contrast, Queens has the lowest rate of incidents per tree despite having the most street trees. This suggests that wildlife incidents are more concentrated in boroughs with fewer street trees and more spread out in boroughs with higher street-tree density.

Spatial Distribution of Street Trees (Binned spatial density plot / heatmap)

In addition to borough-level summaries, I examined the spatial distribution of street trees across New York City to better understand how trees are physically distributed throughout the city. Mapping tree density provides context for borough-level patterns and highlights areas where street trees are more concentrated or sparse.

ggplot(tree_clean, aes(x = longitude, y = latitude)) +
  geom_bin2d(bins = 50) +
  scale_fill_viridis_c(
    option = "plasma",
    name = "Street Tree Density"
  ) +
  coord_fixed() +
  labs(
    title = "Street Tree Density Across New York City",
    subtitle = "Higher values indicate greater concentration of street trees",
    x = "Longitude",
    y = "Latitude"
  ) +
    theme_minimal(base_size = 12) +
  theme(plot.title = element_text(size = 17, family = "serif", face = "bold"),
        axis.title.x = element_text(size = 12, family = "serif"),
        axis.title.y = element_text(size = 12, family = "serif"))

Spatial density of street trees across New York City. Each grid cell represents the concentration of street trees based on latitude and longitude, with yellow indicating higher tree density

Interpretation

In this spatial density binned map areas shown in yellow represent higher concentrations of street trees, while areas shown in purple indicate lower tree density. Higher-density regions are visible across much of Queens and Brooklyn, whereas Manhattan and Staten Island display more fragmented patterns with fewer high-density clusters. Overall, this visualization highlights substantial spatial variation in street tree density across the city, reflecting differences in urban form, land use, and available green space.

Park-Level Patterns in Wildlife Incidents (Faceted horizontal bar chart)

To further explore where wildlife incidents occur within each borough, I examined park-level patterns by identifying the parks with the highest number of reported wildlife incidents. This allows for a more localized view of wildlife activity within urban green spaces.

animal_clean %>%
  count(borough, property) %>%
  slice_max(n, n = 3, by = borough) %>%
  ggplot(aes(x = reorder(property, n), y = n)) +
  geom_col(fill = "springgreen4") +
  coord_flip() +
  facet_wrap(~ borough, scales = "free_y") +
    theme_minimal(base_size = 10) +
  theme(plot.title = element_text(size = 17, family = "serif", face = "bold"),
        axis.title.x = element_text(size = 12, family = "serif"),
        axis.title.y = element_text(size = 12, family = "serif")) +
  labs(
    title = "Most Frequent Parks for Wildlife Incidents by Borough",
    x = "Park",
    y = "Incident Count"
  )

Top three parks with the highest wildlife incident counts in each borough.

Interpretation

Across boroughs, wildlife incidents are concentrated in a small number of major parks. Van Cortlandt Park stands out in the Bronx with the highest number of reported wildlife incidents, while Central Park shows the highest counts in Manhattan. Prospect Park in Brooklyn also exhibits elevated incident levels relative to other parks in the borough. In contrast, parks in Queens and Staten Island show lower incident counts overall. These patterns suggest that larger parks with extensive natural areas may experience more wildlife activity and a greater likelihood of reported human–wildlife interactions.

Species Involved in Wildlife Incidents (Faceted horizontal bar chart)

To better understand wildlife incident patterns, I examined which animal species are most frequently involved in reported incidents within each borough.

top_species <- animal_clean %>%
  count(borough, species_description) %>%
  group_by(borough) %>%
  slice_max(n, n = 5) %>%
  ungroup()

ggplot(top_species,
       aes(x = reorder(species_description, n), y = n)) +
  geom_col(fill = "mistyrose4") +
  coord_flip() +
  facet_wrap(~ borough, scales = "free_y") +
  labs(
    title = "Most Common Wildlife Species Involved in Incidents by Borough",
    x = "Species",
    y = "Number of Incidents"
  ) +
    theme_minimal(base_size = 10) +
  theme(plot.title = element_text(size = 17, family = "serif", face = "bold"),
        axis.title.x = element_text(size = 12, family = "serif"),
        axis.title.y = element_text(size = 12, family = "serif"))

Most common wildlife species involved in reported incidents across NYC boroughs. Bars represent the number of incidents involving each species, with panels shown separately for each borough.

Two raccoons resting in a tree in Central Park shortly after dusk, illustrating the presence of adaptable urban wildlife in New York City parks. Photo credit: Chris St Lawrence

Interpretation

Across all boroughs, raccoons are the most frequently reported species involved in wildlife incidents, suggesting they are a dominant presence in urban park environments throughout New York City. Domestic animals such as dogs and cats also appear consistently across boroughs, reflecting frequent interactions between pets, wildlife, and human-managed spaces. Some borough-specific patterns emerge, including higher involvement of red-tailed hawks in Manhattan and white-tailed deer and snapping turtles in Staten Island, which likely reflects differences in habitat type, park size, and proximity to less-developed natural areas. Overall, these patterns illustrate how urban wildlife incidents vary not only in frequency but also in species composition across boroughs.

Inferential and Exploratory Analyses

Building on the descriptive patterns observed above, I conducted a series of exploratory statistical analyses to examine whether borough-level differences in urban trees and wildlife incidents were associated with measurable differences in tree characteristics or wildlife outcomes. These analyses are intended to complement the descriptive results and explore potential relationships, rather than to establish causal effects.

Differences in Average Street Tree Size Across Boroughs (One-way ANOVA)

To examine whether street tree characteristics vary across New York City, I tested whether average tree diameter at breast height (DBH) differed by borough. A one-way ANOVA was used to compare mean tree size across the five boroughs, allowing for an assessment of whether tree size distributions are similar or distinct across urban contexts.

anova_tree <- aov(tree_dbh ~ borough, data = tree_clean)
summary(aov(tree_dbh ~ borough, data = tree_clean))

##                 Df   Sum Sq Mean Sq F value Pr(>F)    
## borough          4  1241407  310352    4178 <2e-16 ***
## Residuals   683783 50788948      74                   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

supernova(anova_tree)

##  Analysis of Variance Table (Type III SS)
##  Model: tree_dbh ~ borough
## 
##                                    SS     df         MS        F   PRE     p
##  ----- --------------- | ------------ ------ ---------- -------- ----- -----
##  Model (error reduced) |  1241407.100      4 310351.775 4178.336 .0239 .0000
##  Error (from model)    | 50788948.448 683783     74.276                     
##  ----- --------------- | ------------ ------ ---------- -------- ----- -----
##  Total (empty model)   | 52030355.548 683787     76.091

Interpretation

A one-way ANOVA revealed a statistically significant difference in average street tree diameter (DBH) across New York City boroughs, F(4, 683,783) = 4178.34, p < .001. This result indicates that mean tree size varies by borough. However, the associated effect size was small (η² = 0.024), suggesting that borough-level differences account for only a small proportion of the overall variation in tree DBH. While boroughs differ in average tree size, most variability in tree diameter is likely driven by other factors such as species composition, tree age, and local environmental conditions.

Association Between Borough and Wildlife Condition (Chi-square test of independence)

In addition to overall incident counts, I examined whether the condition of wildlife involved in reported incidents varied by borough. A chi-square test of independence was used to assess whether the distribution of animal conditions (e.g., healthy, injured, deceased) differed across boroughs, which helps identify whether certain boroughs experience different wildlife outcomes.

table_boro_ac <- table(animal_clean$borough, animal_clean$animal_condition)
chisq.test(table_boro_ac)

## 
##  Pearson's Chi-squared test
## 
## data:  table_boro_ac
## X-squared = 113.65, df = 12, p-value < 2.2e-16

cramerV(table_boro_ac)

## Cramer V 
##  0.07477

Interpretation

A chi-square test of independence indicated a statistically significant association between borough and wildlife condition, χ²(12) = 113.65, p < .001. This suggests that the distribution of wildlife outcomes (e.g., healthy, injured, unhealthy, DOA) varies across New York City boroughs. However, the strength of this association was small, as indicated by Cramér’s V = 0.075. This implies that while borough-level differences in wildlife condition exist, borough alone explains only a small portion of the variation in wildlife outcomes, and other factors likely play a larger role in shaping wildlife condition across the city.

Exploratory Relationship Between Street Tree Abundance and Wildlife Incidents (Simple linear regression)

Because this analysis is exploratory and based on a small number of borough-level observations, I first examined the relationship visually using a scatter plot. This allows for an intuitive assessment of whether total street tree abundance appears related to wildlife incident counts.

## 
## Call:
## lm(formula = total_incidents ~ total_trees, data = borough_summary)
## 
## Residuals:
##       1       2       3       4       5 
## -258.91   57.84  479.46   67.18 -345.57 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)  
## (Intercept) 1.409e+03  3.764e+02   3.745   0.0332 *
## total_trees 8.280e-04  2.462e-03   0.336   0.7588  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 376 on 3 degrees of freedom
## Multiple R-squared:  0.03633,    Adjusted R-squared:  -0.2849 
## F-statistic: 0.1131 on 1 and 3 DF,  p-value: 0.7588

ggplot(borough_summary,
       aes(x = total_trees, y = total_incidents, label = borough)) +
  geom_point(size = 3) +
  geom_smooth(method = "lm", se = FALSE) +
  geom_text(vjust = -0.8) +
  annotate(
  "text",
  x = 225000,
  y = 1250,
  label = paste0("R² = ", round(r2, 2)),
  size = 3.5,
  family = "serif") +
  labs(
    title = "Street Tree Abundance and Wildlife Incidents by Borough",
    x = "Total Street Trees",
    y = "Total Wildlife Incidents"
  ) +
    theme_minimal(base_size = 9) +
  theme(plot.title = element_text(size = 17, family = "serif", face = "bold"),
        axis.title.x = element_text(size = 12, family = "serif"),
        axis.title.y = element_text(size = 12, family = "serif"))

Scatter plot showing the relationship between total street trees and total reported wildlife incidents across New York City boroughs. Each point represents a borough. No clear linear trend is observed, consistent with the results of the exploratory linear regression.

Finally, a simple linear regression was used to examine the relationship between total street trees and total wildlife incidents across boroughs, treating this analysis as exploratory given the small number of observational units.

simple_regression <- lm(total_incidents ~ total_trees, data = borough_summary)
summary(simple_regression)

## 
## Call:
## lm(formula = total_incidents ~ total_trees, data = borough_summary)
## 
## Residuals:
##       1       2       3       4       5 
## -258.91   57.84  479.46   67.18 -345.57 
## 
## Coefficients:
##              Estimate Std. Error t value Pr(>|t|)  
## (Intercept) 1.409e+03  3.764e+02   3.745   0.0332 *
## total_trees 8.280e-04  2.462e-03   0.336   0.7588  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 376 on 3 degrees of freedom
## Multiple R-squared:  0.03633,    Adjusted R-squared:  -0.2849 
## F-statistic: 0.1131 on 1 and 3 DF,  p-value: 0.7588

r2 <- summary(simple_regression)$r.squared

Interpretation

As shown in the scatter plot, there is no clear linear relationship between street tree abundance and wildlife incidents across boroughs. This visual pattern is supported by the regression analysis, which found no significant association between total street trees and total wildlife incidents (β = 0.00083, p = .76, R² = .04). Boroughs with high tree counts, such as Queens, did not consistently report higher numbers of incidents, while Manhattan exhibited a high incident count despite having fewer street trees. Together, these findings suggest that street tree abundance alone does not explain borough-level variation in wildlife incidents.

Discussion and Implications

Conclusion

Overall, this analysis helps show how street trees and wildlife incidents are distributed across New York City and how these patterns differ by borough. The descriptive results highlighted clear differences in tree counts, wildlife incident totals, and the species and park locations most often involved. When wildlife incidents were adjusted based on tree availability, differences across boroughs became even clearer, suggesting that the total number of street trees alone does not explain where incidents happen.

The exploratory statistical analyses supported this pattern. Although average street tree size differed across boroughs, there was no strong or statistically significant relationship between the total number of street trees and wildlife incidents. Wildlife condition also varied by borough, but the small effect size suggests these differences should be interpreted carefully. Taken together, these findings suggest that wildlife activity in urban environments is influenced by many factors beyond street tree abundance, such as park characteristics, land use, and human activity.

Audience & Relevance

These findings are relevant to New Yorkers, urban planners, policymakers, community organizations, and wildlife management agencies interested in how green infrastructure and wildlife interact in dense urban environments. Knowing where wildlife incidents occur most often, and how they relate to tree distribution and park locations, can help support decisions around urban greening, park management, and public education.

For residents, these findings provide insight into how wildlife interactions vary across neighborhoods. For policymakers and planners, the results underscore the importance of considering ecological context and spatial patterns, rather than relying solely on broad measures such as total tree counts, when designing interventions aimed at improving both human and wildlife well-being in cities.

Connection to Open Data

This project was completed using publicly available datasets from NYC Open Data, including the 2015 NYC Street Tree Census and the Urban Park Ranger Animal Condition Response dataset. Access to these open datasets allows students, researchers, and community members to independently explore real-world urban and environmental questions.

By bringing together multiple open datasets, this analysis shows how open data can be used to examine the relationship between urban trees and wildlife activity. This project shows how publicly accessible data can be used in a transparent and reproducible way to explore environmental questions and connect them to urban planning and policy discussions.