Data
Based on my literature review: I use the following 3 criteria to evaluate bird collision risk at UBC:
surface area of building glazing, proximity to and density of trees, proximity to and density of soft landscape features, and building shadows.
I then explored and compiled data from online sources.
1. Building Glazing​
At the present time, there is no data for the amount or percentage of glass in buildings. Instead, I used the building construction years table to studied the architecture style patterns and then visited certain buildings to collect approximate measurements of the area glazed on building facades. From here, I used the building footprints and building height data to calculate the surface area of building facades and the percentage of glazing. My field data collection process is described in detail here.
Building Construction Years Table
Source: UBC Campus and Community Planning Building Data 2018 (not public)
Acquired through a contact in the department.
Paths and Roads Shapefile
Building heights
Sources: City of Vancouver Open Data Catalogue - Building Footprints 2009
Google Earth, UBC Campus and Community Planning Project Floor Plans (for buildings built in 2018)
UBC Administrative Boundary Shapefile
​
This is data is used to delineate the area that is within UBC's administrative jurisdiction and provide context to the final map.
​
Source: UBC GeoData - context
Source: UBC GeoData - Routes 2016
2. Tree Density
Tree Points Shapefile
Source: UBC GeoData - Landscapes 2008
3. Landscape Density
Landscape Shapefile
Source: UBC GeoData - Landscapes 2015
4. Building Shadows
From their study of building bird collision risk, Kahle, Flannery, and Dumbacher (2016) found that access to sunlight might affect glare and reflection at key times of the day, thus increasing collision rates. To model this effect, I used building footprint and building height data (shown above) to analyze the shadows buildings casted on each other and identify facades that would have more sunlight access.
Field Data Collection - Building Glazing
The amount of building surface area glazed is critical bird collision risk factor at UBC (Cavers et al., 2015; Huang & Porter, 2015). As such, in order to gather data on amount of glazing on buildings, I visited buildings in person to record the surface area of windows and glass. Field data collection can be extremely time consuming and it is near impossible to measure glazing for every building on campus, as such, I originally planned to do a stratified random sample of buildings by architectural styles.
​
To understand my building population, I first did a building study of architectural styles using data on building construction years.
1. Building Study
Using Microsoft Excel pivot tables and charts, I analyzed the patterns of construction year and architecture style of buildings at UBC. For the purpose of this study, I only examine core institutional buildings and exclude student residences (including fraternity and sorority centres), recreation and athletic buildings, hospital facilities, and residential neighbourhood developments. Aside from general architectural style patterns, the purpose and programming of a building significantly affect the building form and design. As such, the glazing surface area of the excluded building types are likely to be more varied and heterogeneous from core institutional buildings.
​
​
Architectural Styles of Buildings at UBC
-
Fairview Campus FRVW
-
Collegiate Gothic CG
-
Wood frame and Stucco Construction WF
-
Modernist MOD
-
Brutalist BRUT
-
Post-Modernist PMOD
-
1990s High Modernism/High Glazing GLAZ
-
2000s Post-Modernist PPMOD
-
2010s Present Style PRES
​
​
The architectural styles were first informed by a literature review of Kalman's (1994) comprehensive book A History of Canadian Architecture which detailed the years that the styles were popularized and each style's primary design characteristics. I combined these style categories with the UBC Library Archive's Chronology of Buildings (University of British Columbia Archives, n.d) which provides some historical context to early construction (ex. the original Fairview Campus buildings built in 1911-1918). As architectural styles often spanned at least a decade or longer, I paid special attention to how I assigned architectural styles to buildings built at the end or beginning of a decade.
2. Sample Methodology and Selection
Originally, I had planned to choose a stratified random sample using the 9 architecture styles (strata) for a sample of 35 buildings. However, this was extremely time consuming and after a day of data collection, I decided to narrow the sample size down to 9 randomly selected buildings.
3. Digitization
.jpg)
1 meter measuring tape for scale
Mathematics Building
Measuring window surface area in J Image
1. Go to north facade of building and place 1 meter measuring tape against the surface where windows are located. Take a picture. Repeat for each facade clockwise.
2. Measure glazing surface area using J Image, a Java-based open source image processing program. I chose to use J Image based on Gentile and Cheung's (2017) methodology.
3. Record glazed surface area. Total glazed surface area/building facade surface area = Percent of building that is glazed.
4. Calculate the glazing percentage for all 9 buildings and look for patterns based on year and architectural style.
Of the 9 buildings sampled, modernist style buildings (West Mall Annex) have the lowest percentage of surface area glazed of 1.9% and the 90s high glazing buildings have the highest percentage of surface area glazed of 45%
4. Classification and Generalization
low glazing facade
medium glazing facade
high glazing facade
It is important to note that glazing percentage varies between buildings of the same architectural style, and as I was only able to sample 9 buildings, I decided to classify glazing into 3 classes in 15% intervals rather than a larger more specific number of classes.
The rest of the un-sampled buildings under each architectural style were then generalized to the one of the three classes.
To verify if buildings of the architectural style roughly matched the 3 glazing classes, I randomly selected buildings from each architectural style eyeballed the glazing surface area from Google searched images.
Data Preparation
>Data Cleaning
Buildings (UBC shapefile)
​
The following building types were deleted and excluded from analysis.
​
Building category =
Residential
Recreational
Hospital
​
>Data Transformation
Building footprint simplification
​
Building sides were simplified to a tolerance of 3 meters so that small corners or edges were generalized to once facade.
​
Building Polygon > Polyline
​
Building polygons were converted to polylines and split at vertices to create the facade shapefile.
​
Buildings (City of Vancouver 2009 shapefile)
​
Overlay with updated UBC footprints and delete CoV footprints of demolished buildings.
​
City of Vancouver building footprint Dissolve
​
Dissolve CoV building features to one base footprint using the maximum height.
Tree Points > Polygon
​
Using 5 meter buffers, tree points were converted the polygons to replicate canopy surface area.
A 5 meter diameter for tree canopies was chosen based on approximate measurements from Google Earth.
​
>Spatial Join
UBC Building Footprints + City of Vancouver 2009 footprints
​
Perform a spatial join with City of Vancouver 2009 footprints to get building heights.
Buffer Analysis
Buffer Analysis
Studies and bird friendly design guidelines generally identify that trees and soft landscape features within 2 to 20 meters are of concern to bird building collision risk (Gentile & Cheung, 2017; Klem et al., 2009). However, trees that are touching or within 2 meters from glazed facades can create a protective effect as birds cannot build up enough momentum to fatally crash into glass (Gentile & Cheung, 2017).
A buffer analysis was conducted for the 2 to 20 meter radius of buildings for trees and 20 meter radius of buildings for landscape features.
Trees Buffer Analysis
​​
-
2m buffer to building footprint
-
Merge 2m buffer and building footprint
-
20m buffer to building footprint
-
Erase 2m merged buffer to 20m buffer
-
Intersect tree polygons to erased buffer
This creates a shapefile for trees that are within 2-20m from buildings.
​
​
-
20m buffer to building footprint
-
Intersect tree polygons to 20m buffer
-
Summarize: sum of tree surface area within 20m buffer.
-
Join: to analyze tree surface area per building
Landscape Buffer Analysis
​​
​Landscape features like lawn, shrubs, and plant bed vegetation is less likely to be high enough to touch or be adjacent to buildings. As such, I looked at landscape features within a 20m buffer.
​
-
20m buffer to building footprint
-
Intersect landscape polygons to 20m buffer
-
Summarize: sum of landscape surface area within 20m buffer.
-
Join: to analyze landscape surface area per building
High Glazing Facade Buffer Analysis
​
Based on observation, I found that the percentage of facade glazing was higher on the front or entrance of a building which is connected to or facing pedestrian paths. To identify the front/high glazing facades of a building, I used the select by location function to find facades within the following proximity of a path:
-
3m from local paths
-
7m from sidewalks
-
15m from primary paths (Main Mall)
Shadows Analysis
To locate building facades that would possibly have less reflection, I performed sun shadow volume analysis which visualizes how buildings cast shadows onto each other. I hypothesize that as glazed facades have less access to sunlight, reflections are likely to be dampened.
At the moment, research on the affect of sunlight on glazed facades to bird collisions has not been done and due to the uncertainty of this criteria, it is excluded from my final multi-criteria evaluation.
To use the sun shadow volume tool in ArcMap, features must be input as multipatch features.
​
Building footprints were first converted to TIN features with z values = maximum height of buildings, then converted to multipatch features.
​
The shadows are based the sun position on September 1st, 2018 at 12:00PM. This is a peak time for bird collisions as migration happens throughout fall and there is the most sunlight at solar noon.
Multi-Criteria Evaluation (MCE)
I performed a MCE using the following criteria:
-
surface area of building glazing
-
density of trees
-
density of soft landscape features (wild, plant bed, lawn)
​
MCE is a method which combines individual criteria to display the most "suitable" locations where criteria values are concentrated. For this project, I am locating buildings that are the most suitable for retrofits or have the highest risk for bird collision.
​
Criteria Normalization
So far, the criteria data has values of all different ranges. To make sure I'm comparing 'apples to apples', I first normalize the criteria values to a scale of 0 to 1. 0 being the least suitable and 1 being the most suitable.
I performed a fuzzy linear transformation of the building with min values = 0 (assuming that there are buildings with 0 glazing and area of vegetation around them and max values = the highest value in the data set.
Percent of Building Glazing
1.9% to 45%+
Tree Density
Low (<200 m²) to High (>3500m²)
Soft Landscape Density
Low (<500 m²) to High (>14000m²)
Most suitable for bird friendly retrofits = highest potential bird collision risk
Weighting
Once criteria have been normalized to a common scale, I determined the weight, meaning how much each factor matters in relation to the others, using the http://www.123ahp.com site. This site calculates the weights of factors through an Analytic Hierarchy Process (AHP) wherein the criteria are ranked on a numerical scale out of 1.00.
All weighted
The criteria importance or weights for each criteria above is informed through my literature review and observations.
Soft landscape is categorized into 3 types: wild, plant bed (shrubs and gardens), lawns.
As lawns are least likely to attract birds for foraging or habitats, it is weighed least important.
Sensitivity Analysis
How would the selection of priority buildings for retrofits change if a different set of weights were used? In order to explore this possibility, I also created an equal soft landscape weighted MCE and all equal criteria weighted MCE.
Equal weight Landscape
All Criteria Equal Weight
All criteria weighed 0.200 (20%)
This is an unlikely scenario as many studies have cited that glazing percentage is a high determinant of bird collisions and cannot be compared to the effects of lawn landscape coverage.
All weighted MCE and equal landscape weight MCE maps were then combined into a final map where buildings were high priority for both weighing schemes. Click here to see the results.