Implications of Volunteer Bias in Ebird Data. (Under the Direction of Dr

ABSTRACT

PERKINS, DEJA JNAI. Blind Spots in Citizen Science Data: Implications of Volunteer Bias in eBird Data. (Under the direction of Dr. Madhusudan Katti).

Cities are complex socioecological systems that require data collection methods that are equal in their ability to capture ecological patterns and social inequity. Social inequity in cities is apparent through lasting environmental legacies, such as in tree cover, which reflects historical patterns of inequity resulting from systematic forces (segregation and racism) with lasting effects on how cities function economically and ecologically. These legacies influence the distribution of bird diversity and abundance of native species within cities. For proper management and distribution of resources for nature within cities, we need sampling methods that are able to detect the human drivers of bird distribution within cities.

Citizen Science is an increasingly popular approach for environmental data collection, especially in urban contexts, with many benefits for both researchers and participants. eBird and systematic point counts are two bird census methods commonly used to inform research, management, and policy decisions. Crowd based citizen science projects like eBird and systematic counts have been used for over 154 tangible science actions including research and monitoring, IUCN Red Listing, and informing policy. With citizen science based data informing policy and management decisions, it is important to ensure that data collection methods sample equally with respect to social inequity. This study explores the spatial distribution of two citizen science methods, eBird checklists and systematic point counts, to assess volunteer site selection biases, and if the two methods differ in their ability to detect patterns of social inequity. The study compares the two methods across four cities in two regions: Raleigh NC, Durham NC,

Tucson AZ and Fresno CA. Stationary eBird checklists submitted between April and May 2015 were compared to systematic point count surveys in Tucson and Fresno between April 15th – May 15th 2015. Stationary eBird checklists from April – May 2019 were compared to systematic point count surveys in Raleigh and Durham from April 15th – May 31st 2019.

Using ArcMap 10.6, map visualizations display the spatial distribution of unique survey and checklist locations across median household income block groups in each of the study areas.

Contingency analysis carried out in JMP Version 14.2 SAS Institute Inc. reveal that eBird checklists have a consistent pattern of spatial bias in the distribution of sampling locations, with overrepresentation of some (typically middle) income groups, and underrepresentation of others, particularly the lowest income groups. The study shows that systematic point counts are a better method for urban data collection due to a more intentionally, even distribution of sampling locations across the urban landscape. If properly sampled, systematic surveys can reflect the true pattern of income distribution across the urban landscape in respect to area. Bivariate analysis show that if properly sampled, systematic surveys can detect a relationship between income and bird species richness whereas stationary eBird checklists typically do not have the sample size to detect such relationships. Cities are complex socioecological systems with different underlying sociocultural variables that shape natural communities depending on the regional climatic and cultural history of the area. This study reveals that cities require data collection methods with proper sampling size and spatial distribution to detect underlying variables that influence bird species richness in urban landscapes.

by Deja Jnai Perkins

A thesis submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Master of Science

Fisheries and Wildlife Conservation Biology

Raleigh, North Carolina 2020

APPROVED BY:

______Dr. Madhusudan Katti Dr. Stacy Nelson Committee Chair

______Dr. Louie Rivers

DEDICATION

I dedicate this Thesis to myself, for persevering through it all to reach the finish line.

BIOGRAPHY

Deja Perkins is a graduate student pursing a master’s degree in Fish, Wildlife, and

Conservation Biology with a heavy focus on urban ecology. Growing up between worlds in the

South Side and the Suburbs of Chicago, inspired her to study urban ecosystems and the influences that shape cities and the organisms that call it home. As a young child, she sought out every opportunity available to learn about the environment and the species that live within it. She pursued her undergraduate education at Tuskegee University where she became passionate about environmental justice and extension. Currently she is the volunteer coordinator and research technician for the Triangle Bird Count, a citizen science project based out of the Lab of

Reconciliation Ecology at NC State. This project aims to monitor the diversity of urban birds in the Triangle area of North Carolina. She aims to study how socioecological variables influence birds in the urban landscape, and explore how citizen science data collection protocols may be influenced by the pre-existing and historical structures of the landscape. She hopes to continue exploring the complexities of the urban environment, to help create ecologically and socially equitable, sustainable urban centers; as well as continue finding creative ways to inspire communities to observe nature in their neighborhoods.

iii

ACKNOWLEDGMENTS

I would like to thank all the volunteers who collected data and participated in the

Triangle Bird Count, as well as Jennie McFarland, Christopher Hensley, Pedro Garcia, and

Stephanie Slonka for the 2015 Fresno and Tucson Bird Count Datasets the managers of the

Tucson Bird Count for the 2015 survey year. I would also like to thank my undergraduate lab technicians, Kolby Altabet, Evelyn Rowan, Hadley Buckner and Shannon Dolan for assisting in data collection and processing. I would like to thank the Southeast CASC for providing funding for the first year of my program, and mentoring me in project management.

I would like to thank my lab mates, friends, family, committee and my mentors for helping me through this process. Thank you to my committee members, Dr. Madhusudan Katti,

Dr. Stacy Nelson, and Dr. Louie Rivers, for being a continuous stream of support, a safe space to talk about injustice, and encouraging me to be my best, full authentic self throughout this process. Thank you to Jason Ward for introducing me to the world of birding and being a source of support. Lastly, I would like to thank my parents for continuously encouraging me to follow my dreams.

TABLE OF CONTENTS

LIST OF TABLES………………………………………………………………………………..vi LIST OF FIGURES ...... vii Introduction ...... 1 Urban Influences ...... 1 Data Collection for Urban Studies ...... 3 eBird ...... 6 Systematic Counts ...... 7 Current Study ...... 8 Methods ...... 9 Study Area...... 9 Surveys ...... 14 Volunteer Recruitment ...... 16 Data Analysis ...... 17 Results ...... 18 Discussion...... 24 Conclusion ...... 27 References ...... 30 Appendixes...... 35 Appendix A ...... 36 Appendix B ...... 37 Appendix C ...... 38 Appendix D ...... 39 Appendix E ...... 40 Appendix F...... 41 Appendix G ...... 42 Appendix H ...... 43 Appendix I ...... 45 Appendix J ...... 51

LIST OF TABLES

Table 2.1. Social attribute comparison between the four study areas…………………………9

Table 4.1. Results of the contingency analysis between the 4 study areas……………………21

LIST OF FIGURES

Figure 2.1 Map of Raleigh (eBird)……………………………………………………………..10

Figure 2.2 Map of Raleigh (Systematic counts) ...... 10

Figure 2.3 Map of Durham (eBird) ...... 11

Figure 2.4 Map of Durham (Systematic counts) ...... 11

Figure 2.5 Map of Tucson (eBird) ...... 12

Figure 2.6 Map of Tucson (Systematic counts) ...... 12

Figure 2.7 Map of Fresno (eBird) ...... 13

Figure 2.8 Map of Fresno (Systematic counts) ...... 13

Figure 4.1 Raleigh Contingency Analysis Graph ...... 19

Figure 4.2 Durham Contingency Analysis Graph...... 19

Figure 4.3 Fresno Contingency Analysis Graph ...... 20

Figure 4.4 Tucson Contingency Analysis Graph ...... 20

Figure 4.5 Tucson eBird Regression Test ...... 22

Figure 4.6 Tucson Systematic Regression test ...... 22

Figure 4.7 Fresno eBird Regression test ...... 23

Figure 4.8 Fresno Systematic Regression test ...... 23

Figure 4.9 Raleigh eBird Regression test ...... 23

Figure 4.10 Raleigh Systematic Regression test ...... 23

Figure 4.11 Durham eBird Regression test ...... 24

Figure 4.12 Durham Systematic Regression test ...... 24

vii

1. Literature Review

Urban Influences

Urban landscapes are complex systems, and they are important for conservation. Pickett et al. (2011) stated that we should think of studying the “ecology of cities” instead of the

“ecology in cities” because urban ecology is a coupled socioecological system, and it has many influences that shape both the built and natural environment. Urban ecosystems are comprised of both people and nature, and therefore must also include human social and economic manifestations, in addition to introduced [non-native organisms] plants and microbes (Pickett et al. 2011). Ecologically, the regional climatic and biogeographical factors, phylogeny and evolutionary history, species interactions, individual species life history, and ecological functionality, all play a role in determining community assembly in cities (Aronson et al. 2016;

La Sorte et al. 2018; Hensley et al. 2019; Morelli et al. 2020). However, ecological factors alone cannot explain all the patterns seen in urban environments. Human factors such as resident socioeconomic status, culture, neighborhood structural features, governance mechanisms, and landscape planning, have all been shown to have an important influence on urban habitat (Loss and Brawn 2009; Aronson et al. 2016; Katti et al. 2017).

Humans are actors in the environment, manipulating everything in cities. They create heterogeneous habitats through highly fragmented landscapes with patches of greenspaces in the form of parks, residential yards and undeveloped land; as well as in the vegetation, building and paving materials, and the entire morphology of the urban landscape (Pickett et al. 2011).

Humans not only manipulate the environment through what and where they build, but also through lasting legacy effects. Legacy effects are the impacts that previous events, processes and phenomena have on current properties or processes. Legacies can range from natural to

human driven. In the bioregional context, native biomes, climate, topography, initial vegetation and pre- urbanization land use, all represent the natural conditions in which a city was established and grew, which in turn influences how legacies unfold (Roman et al. 2018). Human drivers of legacy effects reflect specific historical periods, colonial history, laws and regulations, urban parks, and civic beautification movements. Other human drivers include neighborhood urban form (shifting residential patterns and lot design) and socioeconomic change (past residents and social groups) (Roman et al. 2018), and cultural inertia (Katti et al. 2017).

Minorities are disproportionately burdened with environmental dis-amenities and experience fewer amenities than the privileged majority (Mochai and Saha 2007). The “space to plant” is a legacy of the urban built environment that stems from a history of deliberate and systematic racial discrimination in housing and urban development (Roman et al. 2018). Trees are an ecosystem service, especially in the urban environment, providing many health and environmental benefits to humans, and habitat for wildlife. A recent meta-analysis shows that lower income areas with more racial minorities have less tree canopy cover, which can make health problems worse for already disadvantaged groups (Roman et al. 2018). These urban inequities are the result of historical and systematic forces (segregation and racism), which have shown to have lasting effects on how cities function economically and ecologically (Locke

2020). Redlining is a racially discriminatory US housing policy established in the 1930s by the

Home Owners Loan Corporation to limit access to homeownership and wealth creation from racial and ethnic minorities (Locke 2020). This caused a host of adverse effects ranging from high unemployment rates and poverty, to residential vacancies. The impacts of redlining legacies still exist today, but its impacts on urban environments and ecosystems are still unclear. Studies from Baltimore, MD, show how historical redlining policies influence the current location of

parks and trees. Areas classified as “D” (the lowest rating which discouraged banks from lending and investing in these areas) in the past, have on average 23% tree cover today, in comparison to neighborhoods classified as “A” (the highest rating, preferred for bank lending), which show almost twice that amount (43%) (Locke 2020).

Disadvantaged groups enjoy fewer amenities and are disproportionately burdened with polluted and hazardous environments close to where they live, in comparison to the privileged majority (Pickett et al. 2011). Wealthier neighborhoods typically display a “biological luxury effect”, where more native species are found in higher income areas, also in higher abundance

(Melles 2004). Studies in Europe, Australia, and America have shown that bird diversity also reflects land use and socioeconomic patterns (Strohbach et al. 2009). Given that birds rely on urban vegetation structure, human socioeconomic status and urban form legacies should also have an impact on bird community assembly. Due to this relationship, we should be able to use urban bird diversity patterns to detect patterns of ecological inequality in cities.

Data Collection for Urban Studies

Conducting comprehensive studies in urban areas is challenging because cities are large and contain varying environments. To be useful, biological surveys need to be efficient, cover a variety of land uses, and encompass a range of spatial scales (Turner 2003). Traditional field methods can be difficult to implement in the urban landscape due to the complexities of navigating road structures and access permissions for public and private property. The use of citizen science has become very popular in recent years as a way to crowd-source data collection and increase scientist engagement with the general public (Dickerson et al. 2010). Collecting representative data from large diverse areas requires a considerable amount of effort and citizen science can help achieve this need (McCaffrey 2005). The citizen science model utilizes a

dispersed network of volunteers to assist in professional research developed by or in collaboration with researchers (Cooper et al. 2007). Most citizen science projects involve surveillance monitoring over large geographical regions, which is then used at a later date by researchers, allowing for spatial and temporal comparisons of the study organism or phenomenon (Dickerson et al. 2010).

Citizen science is popular within the field of urban ecology because it allows the pairing of ecological data with human cultural attributes such as income, education, and water use

(Disckerson et al. 2010). The spatial biases in data collection, causing increased observations in more densely populated areas, makes it a valuable tool for research in urban areas in comparison to rural environments due to the higher volume of participants (Johnston et al. 2020). Citizen science also allows researchers to engage with non-scientists at larger scales than what is achievable through professional science alone (Johnson 2014). Urban ecology projects make participation easier for potential volunteers due to the accessibility of habitats; familiar

“backyard” species, lacking the need for specialized knowledge; and providing convenient means for introducing scientific activities for individuals who lack experience in the scientific process (Pandya 2012). Aside from researcher benefits, citizen science has numerous benefits for participants as well. It provides opportunity for the community to improve their scientific literacy and sense of place by inviting them to be involved in ecological research in their own backyards or neighborhoods (Evans 2005). It creates an awareness of the local environment that can turn into local advocacy, and most importantly, it allows people to participate in science rather than become a recipient of outreach efforts (Pandya 2012).

However, participation in citizen science does not reflect the demographics of the US.

The communities that can benefit most from citizen science projects are not the majority of

participants of citizen science projects. Historically underrepresented minority groups such as

African Americans, Latinos, and Native Americans participate less in citizen science than majority groups, and affluent participants outnumber less affluent participants (Pandya 2012).

Barriers to citizen science can include lack of access to natural settings; lack of transportation; lack of familiarity with science and scientific processes; as well as conflict between participation and other responsibilities (Pandya 2012).

About 66.1 million Americans (31%) over the age of 16 participate in bird watching

(Eubanks et al. 2004). Due to its popularity as a hobby, bird-monitoring projects have been the most successful at integrating citizen science (McCaffrey 2005). The public has been contributing to the monitoring and understanding of bird distribution and abundance through citizen science for many years. The oldest and largest wildlife survey in the world, the Christmas

Bird Count (CBC) which began 1990 (Saver and Druege 1990), as well as other projects such as the North American Breeding Bird Survey (NABBS), Celebrate Urban Birds, and Project

Budburst, all have long standing datasets showcasing the effectiveness of citizen science projects for long term data collection and monitoring (Pandya 2012). Currently there are two methods that are widely used in urban bird surveys: eBird and systematic point counts. Both methods rely on citizen science, or public participation for data collection. Participation in bird watching however, like citizen science, tends to lean towards the affluent and older White demographic.

On average, birders are 88-97% White, 51-60 years old, and have a median household income of

$55-$96k (Eubanks, Stoll and Ditton 2004). The time and money alone are additional barriers to inclusive participation in recreational bird watching.

eBird Platform

The eBird platform builds upon the tradition of the public contributing to bird monitoring data. It uses a variety of information technologies to engage a global network of bird watchers to report observations to a centralized database (Wood et al. 2011), housed at the Cornell Lab of

Ornithology at Cornell University (Kolstoe and Cameron 2017). Although it has other applications, eBird is primarily targeted towards bird watchers, or “birders”, as a user group. The platform was created to use data collection protocols that match the way people go birding

(Wood et al. 2011). There are 3 different types of checklists that users can report; traveling, stationary, and incidental. Traveling checklists occur when people observe birds along a route or any area that requires a distance component. Stationary checklists are submitted when people observe birds from a single location (without any distance), while incidental checklists occur when people see a bird they want to report while they were not actively birding. Many birders are classified as “Listers” who aim to maximize the cumulative number of species they are able to see per trip (Kolstoe and Cameron 2017). Listers make this method effective because they spend a lot of effort (duration, number of checklists, miles and money) to observe birds/collect data. To ensure the accuracy and quality of the observations, a regional staff person reviews submitted checklists to ensure observations are likely given the geographic region and time of year (Wood et al. 2011).

One of the fundamental challenges in the face of rapid global change is ensuring that conservation decisions are informed by the best available science (Sullivan and Phillips et al.

2017). The eBird platform aids researchers in that challenge. Currently there are more than

500,000 eBird users worldwide, making the platform unmatched as a potential source of year round bird data (eBird.org). The open access dataset has also made it easier for anyone to answer

questions about an individual species or community of birds across a variety of geographical landscapes and temporal spans. People have used eBird data in many different ways from informing conservation planning to action (Sullivan et al. 2017). The eBird interface is popular for projects requiring bird-based data because it is a cheap, effective method of assisting with biodiversity management in urban greenspaces (Callaghan et al. 2017). It has been used for over

154 tangible science actions including: research and monitoring; developing the IUCN Red List; conservation planning, site/habitat management, and to create laws and policy (Sullivan et al.2017).

Many researchers have previously questioned the reliability of eBird data because it is a citizen science based platform (Kamp 2016; Weirsma, 2010). However, studies have shown that eBird data is comparable to standardized shorebird surveys by trained observers (Callaghan and

Gawlik 2015), and that eBird surveys have higher total species richness and Shannon's diversity indexes than standardized surveys, when studies are completed on the park scale (Callaghan and

Gawlik 2015). Studies in Australian greenspaces comparing data from structured bird surveys and eBird checklists found that there is no difference in species richness or diversity index between the two methods (Callaghan and Martin 2018). The increased effort of eBird surveys through an increased number of observers, time spent surveying and spatial coverage, also results in higher detection when using this method (Callaghan and Martin 2018).

Systematic Bird Count Surveys

Systematic surveys have been used to collect bird data since the North American

Breeding Bird Survey (BBS) in 1996. The BBS was established in 1965 by the USFWS to monitor birds that regularly breed, winter, summer, or migrate through the US (Butcher 1990).

The BBS uses volunteers to conduct 3-minute roadside surveys along specific predetermined

routes. It is a highly standardized method where a few locations are surveyed over a short period of time. Breeding Bird Survey datasets are commonly used to inform management plans

(Tulloch et al. 2013). Similarly to other citizen science bird monitoring programs (Christmas

Bird Count), the BBS avoids data collection in inner cities. With close to 80% of the US being classified as “urban,” and the increasing growth of populations within cities, as well as the increase in urban sprawl, it is increasingly important to have a replicable method to survey the urban matrix (Roman et al. 2018). To account for the need of a reliable urban survey method,

Will Turner created a methodology focusing explicitly on metropolitan areas (Turner 2003).

Turner established the Tucson Bird Count (TBC) in 2002 as a citizen science program now run by the Tucson Audubon Society in partnership with the University of Arizona. Similarly to BBS, the TBC relies on a volunteer base of skilled observers who can ID the 25 most common species of the area by sight and sound. Its survey design is based on the BBS, using a 1km x 1km grid over the city of Tucson AZ. The survey points are distributed randomly across the grid, focusing on one point per cell. TBC surveys increase the number of survey locations and reduce the space between locations. Since it is an urban count, travel time and costs are relatively small, therefore additional effort in the form of increased sites are used instead of repeated visits

(Turner 2003).

Current Study

To make sure we have cities that are ecologically, socially and economically sustainable places to live, we need data that is unbiased in terms of social equity. The current study focuses on a multi-city comparison of the two methodologies currently used for urban bird data collection: systematic counts and eBird checklists. The aim is to 1) explore if the two citizen science methods differ in spatial coverage with respect to underlying social inequities; and 2)

determine if species richness is greater in higher income areas. We do this by 1) comparing if the methods are equal in the distribution of survey locations across income groups; and 2) exploring if there was a relationship between income and bird species richness.Insert your main body text here. Organization and format depends on your style guide. The “Body 1” style for this template indents the first line of each paragraph, but you can adjust this in the “Paragraph” settings, if your style guide requires different spacing.

2. Methods Study Area

The study reflects data collected in Raleigh and Durham, North Carolina, as well as two western cities including Tucson, Arizona and Fresno California. The multi-city comparison allows for a better understanding of what socioeconomic factors may be influencing urban bird diversity and how this relationship may differ across cities. Comparing the methods in two different regions allows for a better understanding of regional differences, and how the biome might play a role in this socio-ecological relationship.

Table 2.1 A comparison of the demographic attributes of the four study areas. The GINI coefficient is a measure of inequality that displays the distribution of wealth in each city.

Demographics Raleigh Durham Tucson Fresno Population(2018) 469,298 316,739 545,975 530,093 Median Household Income $63,890 $58,190 $41,625 $47,189 Maximum Household Income $191,944 $169,583 $134,444 $204,688 % of Persons in Poverty 14% 14% 23% 27% % of Population White 53% 43% 44% 27% % of Population Black 29% 37% 5% 8% % of Population 11% 14% 43% 49% Hispanic/Latino % of Population Asian 4% 6% 3% 14% % of Population over 65 10% 13% 13% 11% GINI COEFFICIENT 0.47 0.47 0.46 0.499 CITY AREA (Acres) 92,660 69,310 145,100 73,410 SAMPLE SIZE eBird 209 148 84 24 SAMPLE SIZE Systematic 356 277 696 260

Raleigh

Raleigh is located in the Piedmont region of North Carolina and primarily consists of oak forests, coniferous woodlands, floodplain forests and riverine aquatic communities. It is a part of a subtropical climate with moderate springs and fall, but hot summers with high intensity storms

(Inkilainen 2013). This highly forested urban landscape is also one of the fastest growing urban regions in the US with a current population of 469,298 (US.Census.gov). Like much of the southern US, Raleigh was a historically segregated town. Remnants of segregation are reflected throughout the city in parks and other areas for recreation. Although it has a lower percentage of individuals in poverty (14%), unequal patterns of income distribution are visible on a map.

Raleigh is a poster child for urban renewal with many of its historically Black neighborhoods undergoing a mass wave of gentrification.

Figures 2.1. and 2.2: Maps of unique checklist (eBird) and survey (Systematic count) locations across median household income blocks in Raleigh, NC. Income is displayed in increments of 30 thousand dollars, with lighter shades of purple displaying lower income values. Richness is displayed in Figure 2.1 according to size and color. The darker shade of blue, and the larger the circle the higher the richness. Richness was not displayed in Figure 2.2 due to the lack of sufficient survey coverage.

Durham

Durham is a historically Black city located 40 minutes from Raleigh, NC that has undergone a mass wave of gentrification. Historically it was a city booming from the tobacco industry with its own Black Wallstreet, but “urban renewal” projects typically decimated Black neighborhoods while building up downtown Durham. As of 2019, Durham has approximately

321,488 residents, 43% White and 37% Black (census.gov). The city has a history of redlining that result in legacy effects that influence both housing and natural amenities. Durham is a part of the Piedmont region of North Carolina with habitats ranging from dry coniferous woodlands, oak forests, mesic forests as well as floodplain forests and riverine aquatic communities. The city is a highly forested urban landscape, with about 40% tree cover (2015 Durhamnc.gov).

Figures 2.3 and 2.4: Maps of unique checklist (eBird) and survey (Systematic Count) locations across median household income blocks in Durham, NC. Income is displayed in increments of 30 thousand dollars, with lighter shades of purple displaying lower income values. Richness is displayed in Figure 2.1 according to size and color. The darker shade of blue, and the larger the circle the higher the richness. Richness was not displayed in Figure 2.2 due to the lack of sufficient survey coverage.

Tucson

Tucson is located in the western half of the United States in a more arid region. Over

800,000 people live in Tucson’s metropolitan area which is roughly 1300km^2. Original habitats surrounding Tucson are upland Sonoran desert scrub with various trees, shrubs, and cacti (Turner

2003). Desert vegetation is scarcer in the suburban and urban areas of Tucson. Instead many developed areas host non-native ornamental plantings and shade trees making the area resemble a savannah habitat than the more natural desert scrub or mesquite woodland habitat (Turner

2003). There are only a few areas in Tucson that contain water year round. Tucson is primarily

White (44%) and Latino (43%), with a high population of retirees (13% over 65). Tucson is a popular bird watching destination, due to its geographic location. Its measure of inequality (46%) is the lowest out of all four study areas, however it still has high levels of poverty compared to the other study areas (23%).

Figures 2.5 and 2.6: Maps of unique checklist (eBird) and survey (Systematic count) locations across median household income blocks in Tucson, AZ. Income is displayed in increments of 30 thousand dollars, with lighter shades of purple displaying lower income values. Richness is displayed in Figure 2.1 according to size and color. The darker shade of blue, and the larger the circle the higher the richness.

Fresno

Fresno is located in California’s Central Valley. The area was historically grassland and riparian woodland but much of the land to the south, east, and west of the city were converted to agriculture, and rangeland to the north and northwest (Hensley 2018; Katti et al. 2017).

Agricultural lands to the north and east were converted into suburban neighborhoods and shopping centers (Katti et al 2017). Fresno typically experiences an annual precipitation of 12.8 inches but has recently been undergoing a drought since 2012 (Hensley 2018). In 2013, the city switched to a water metering system which identified that 70% of water use was being used for irrigation (Katti et al. 2017). Due to the drought, in 2015 municipalities were instructed to reducewater use by 30%, which in turn affects the landscape (Katti et al. 2017). Income influences the amount of water that can be delegated to landscape use, revealing an environmental justice issue in relation to water use and landscaping. Fresno has a higher Asian immigrant population than the other study areas, as well as the highest measure of inequality

(Gini coefficient 49%) out of all four study areas.

Figures 2.7 and 2.8: Maps of unique checklist (eBird) and survey (Systematic count) locations across median household income blocks in Fresno, CA. Income is displayed in increments of 30 thousand dollars, with lighter shades of purple displaying lower income values. Richness is displayed in Figure 2.1 according to size and color. The darker shade of blue, and the larger the circle the higher the richness.

Surveys

Systematic Count

The systematic count surveys in this study are modeled after the methodology used in the

Tucson Bird Count and follow a similar protocol as established by Turner (2003). A 1km by 1km grid was placed over the city limits of the study area and observation points were randomly generated, ensuring only 1 point per cell. The systematic count represents a method that is consistent with effort that is spatially randomized in respect to income. Counts were conducted within a window of 30 minutes before sunrise until 4 hours after sunrise. Unlike the Tucson Bird

Count which has a 5 minute survey period and unlimited observation radius, the Fresno and

Triangle Bird Counts require a 10 minute observation period and want to record all birds seen and heard within a 40 meter radius. The Tucson and Fresno Counts run from April 15 - May

15th, while the Triangle Bird Count utilizes a survey period of April 15 - May 31st. Although the

Tucson and Fresno Bird Counts have continuously recorded data from 2001 and 2008

respectively, only data recorded within the spring season of 2015 were used to ensure complete coverage for a cross-city comparison. Data from Raleigh and Durham only use observations from the 2019 (the first survey year for the Triangle Bird Count) season since the 2020 survey season was cancelled due to COVID 19.

All of the systematic counts rely on experienced volunteers who can identify the 25 most common species of the area. Each volunteer was provided with a map of the survey location,

GPS coordinates and standardized data forms at the training workshops. All data sheets and protocols were also made available online for individuals who were not able to attend workshops in person. In addition to bird observations, observers were asked to collect information on site noise level and site location, as well as any supplemental observations made outside of the survey window. Counts were not conducted in inclement weather (fog, prolonged rain) or if wind exceeded 12mph (continuous movement of tree branches, or loose paper and dust are raised). If sites were inaccessible or located on private property, sites were moved to the nearest public access point and the GPS coordinates and site descriptors were reported to update the database.

After completion, data forms were submitted by mail and entered manually into an online spreadsheet.

eBird Surveys

The second core dataset was obtained from the eBird online platform. eBird data is collected by citizens and hobbyist bird watchers worldwide. Volunteers submit a digital checklist of bird species seen and heard at either a nearby hotspot, personal location or location from a map, all based on geographic coordinates. Date, number of observers, number of individuals per species, trip duration, trip type, distance, and geographic coordinates are also recorded for each submitted checklist. This study only utilized checklists that were classified as stationary protocol

type; where volunteers counted birds within a 30m diameter and within a known period of time, without any distance components. eBird also classifies checklists into traveling (most common checklist type) and incidental, however stationary protocols are most similar to the Systematic

Bird Count protocols across all 4 study areas. For comparison to Fresno and Tucson Counts, this study selected eBird checklists that were submitted April through May 2015 to allow for more complete coverage spatially. Since the TriBC is a newer project, Raleigh and Durham checklists were restrained to April and May of 2019.

Volunteer Recruitment

Systematic Count Recruitment

Recruitment was targeted at birders with a relatively high skill level in each of the respective cities. For the Triangle Bird Count (Raleigh and Durham points), volunteers were recruited through digital and printed flyers, regional birding email listservs, NPR news segment, and personal communication. In an effort to be more inclusive, the Triangle Bird Count (TriBC) allowed volunteers to participate regardless of their level of bird watching experience, but prioritize volunteers who were classified as intermediate or experienced birders. Intermediate and experienced bird watchers are able to identify common backyard species and many migratory species by sight (intermediate) and sound (experienced). Two workshops were held for the TriBC during the spring 2019 season to familiarize volunteers with the protocol and review common species that might be seen during the survey period. Training materials were also provided online for volunteer review. The Tucson Bird Count (TBC) recruited volunteers primarily through its local Audubon chapter due to the program partnership (Turner 2003).

Similarly to the TBC, the Fresno Bird Count (FBC) recruited through printed and digital flyers,

Fresno Audubon Society, regional birding listservs, classes at Fresno State, and personal communication, and provided training workshops for volunteers (Hensley 2018).

eBird Recruitment

Birders using the eBird platform register when they make an online account. Users are allowed to make observations whenever and wherever they choose. Targeted eBird recruitment was not used for this study. It relied on existing registered users of the eBird platform who made observations during April and May 2015 (Fresno and Tucson) and 2019 (Raleigh and Durham).

The amount of eBird checklists varied by study city.

3. Data Analysis

The following section reports the analysis completed for each of the research objectives.

All statistical tests were carried out in JMP®, Version 14.2. SAS Institute Inc.

3.1 Objective 1: Comparison of method spatial distribution of points across income groups

To test the hypothesis that survey point distributions are not equal across survey methods

I ran a Contingency Analysis. To complete the Contingency Analysis and Chi Square Test, median household income was binned by increments of 30 thousand with a minimum value of $0 and a maximum value of $210k. In an effort to reduce possible inaccuracies in the Chi Square

Test due to low values, such as 0 and 1, within the original bins of the higher income groups, higher values were combined into a single group ranging from $150k - $210k. The Contingency

Analysis allowed me to compare the proportion of observation points (unique locations) in each income group and across method type.

To test if the methods differ in spatial distribution of points across income groups, I ran a

Chi Square Test of Independence using “Income” and “Method” as variables to test if the proportion of unique locations in each income group differed based on the method used for data

collection. All the survey locations in the systematic count method were included in the Chi

Square test, regardless of if they were surveyed in the given year or not. This allows us to account for a more accurate representation of both Raleigh and Durham since these surveys had low volunteer participation in its inaugural year of the survey. By using all the survey points in this test, we can compare what Raleigh and Durham’s systematic count method would look like with sufficient volunteer participation to ensure complete spatial coverage of all survey points.

3.1 Objective 2: Determining the relationship between species richness and income

To test the hypothesis that species richness increases as income increases, I ran a

Bivariate Analysis for each method and city. Each survey site (systematic count) and checklist location (eBird) was summarized to obtain the total number of species or species richness at each unique location regardless of date or time surveyed. Observation locations were categorized by method, listed as either “Systematic Count” or “eBird” in each city, as well as by which income block group it was located within. Income groups were derived originally from the Social

Explorer dataset from the 2016 US Census using the “Median Household Income” of each census block group. The median household income layer was imported into ArcMap 10.6 and intersected with the shapefile containing bird observation data. Using Arc Map 10.6, the bird survey data was visualized across the four cities for each method.

4. Results

4.1 Objective 1: Comparison of methods for spatial distribution of points across income groups

The results of the analysis show that eBird and systematic counts are not sampling across the income groups in an equally proportionally representative manner. Comparing the two methods across four cities, I found a significant difference between methods in the proportion of

unique locations distributed among the different income groups (Chi sq = 194.9, P = 0.001).

When looking at each individual city, I found a significant difference in the proportion of unique locations distributed among the different income groups in Raleigh (Chi sq = 12.3, p = .03) and

Tucson (Chi sq = 41.5, p = .0001). In Durham (Chi sq = 5.77, p = .329) and Fresno (Chi sq =

6.07, p = .299) the differences were insignificant.

Raleigh Durham

Figures 4.1 and 4.2: Results of the contingency analysis displayed as a stacked bar graph. The bar on the far right displays the proportion of points in each method without income. The proportion of points is displayed along the y- axis and median household income grouping is displayed across the x-axis. Income is grouped in increments of 30 thousand from 0 to 210 thousand, with the exception of group 6 which represents an increment of 60 thousand due to low sample size across the four cities. The numbers along the x-axis represent the following values: 1 = 0-30k; 2= 30-60k; 3 = 60-90k; 4 = 90-120k; 5 = 120-150k; 6 = 150-210k.

Fresno Tucson

Figures 4.3 and 4.4: Results of the contingency analysis displayed as a stacked bar graph. The bar on the far right displays the proportion of points in each method without income. The proportion of points is displayed along the y- axis and median household income grouping is displayed across the x-axis. Income is grouped in increments of 30 thousand from 0 to 210 thousand, with the exception of group 6 which represents an increment of 60 thousand due to low sample size across the four cities. The numbers along the x-axis represent the following values: 1 = 0-30k; 2= 30-60k; 3 = 60-90k; 4 = 90-120k; 5 = 120-150k; 6 = 150-210k.

Table. 4.1: Results of the contingency analysis displaying the proportion of points located in each income group across both methods in all four study areas.

4.1 Objective 2: The relationship between species richness and income

The relationship between income and species richness was significant in Tucson (P

=.0001) using the systematic count method. The relationship between income and species richness was insignificant in all the other cities regardless of the method used. In Tucson, Fresno, and Raleigh, the relationship between income and species is positive although weakly correlated.

Tucson eBird Tucson Systematic Count

Figures 4.5 And 4.6: Results of the regression test displaying the relationship between income and species richness. Significant relationships were only found within the results of the test for the Tucson Systematic count.

Fresno eBird Fresno Systematic Count

Figures 4.7 and 4.8: Results of the regression test displaying the relationship between income and species richness. The relationships were not found to be significant for the eBird Checklists (small sample size) or the Systematic Counts (underlying variables).

Raleigh eBird Raleigh Systematic Count

Figures 4.9 and 4.10: Results of the regression test displaying the relationship between income and species richness. The relationships were not found to be significant for the eBird Checklists (underlying variables) or the Systematic Counts (insignificant spatial coverage).

Durham eBird Durham Systematic Count

Figures 4.11 and 4.12: Results of the regression test displaying the relationship between income and species richness. The relationships were not found to be significant for the eBird Checklists (underlying variables) or the Systematic Counts (insignificant spatial coverage).

5.Discussion

5.1 Objective 1: Comparison of methods for spatial distribution of points across income groups

In the systematic count method, the grid-based spatial sampling is designed to ensure that effort is equally distributed across the city and therefore proportionally across income groups. In

Raleigh and Durham, the proportion of locations in each method is more evenly distributed than in Tucson and Fresno where the systematic count had a much higher proportion of survey locations than the eBird Method. In each city, the systematic count method had more unique locations than the eBird method. eBird users typically visit the same locations or “hotspots” multiple times resulting in the sampling of fewer unique locations. This sampling scheme is

similar to other sampling methods used outside the urban environment. The systematic method focuses on sampling more locations over repeated samples since travel time in the urban environment is not intensive (Turner 2003). This might explain why the western cities had a larger difference in sample size between the two methods. There could also be a difference in the

“culture” of recreational bird watching in the western cities compared to the eastern cities. Birder demographics are generally White, middle class, and middle aged, whereas the demographics of

Fresno CA do not match that pattern (higher % Asian, high poverty, (lower?) % of people over

65 years old). Tucson is a more popular birding and retirement destination; however, since the

Tucson Bird Count has a long-term partnership with the local Audubon chapter, more birders may be participating in the TBC than eBird during the time frame of this study.

The results of the Contingency Analysis and Chi Square test show that there is a significant difference between the two methods in their spatial distribution of points, suggesting that the systematic point count method has a more even, or better, coverage than what is collected through eBird. In most cities, more unique locations were found to be within the $30 -

$60k income group, regardless of method, except for Durham where the highest proportion of points was found in $60 - $90k. When looking at the pattern of income, both methods can identify the dominant income group of the landscape. However when looking at if each method samples equally across all income groups we see the distribution of points across income groups is different depending on the method used. For example in Tucson, both methods show that the top two income groups are 1) $30 - $60k and 2) $60 - $90k. However, the contingency analysis reveals that when using the eBird method, 50% of the sampled locations are in the top income group ($30 - $60k), while only 24% of the locations are within the second group ($60 - $90k).

The systematic count method has locations more evenly distributed across the top two income groups with 35% ($30 - $60k) and 34% ($60 - $90k) respectively. In the other cities we see a similar pattern of unequal distribution between income groups depending on the method used. We see the eBird method consistently having a higher proportion of effort (sampling locations) in different income groups than the systematic count method. For example in Raleigh, the eBird method had more effort in the $60 – $90k income bin, but if the effort was more evenly distributed like in the systematic count, there should be more effort in the $30 - $60k income areas. Overall, I found a consistent pattern of spatial bias in the distribution of eBird sampling locations, with overrepresentation of some (typically middle) income groups, and underrepresentation of others, particularly the lowest income groups. This likely results from unconscious or implicit bias on the part of eBird volunteers, reflecting their own demographic characteristics and spatial distribution, when they freely choose where to conduct their bird observations. In contrast, the systematic counts are designed to ensure uniform spatial sampling across the entire city, and volunteers are directed to specific survey locations to conduct point counts, thereby minimizing observer biases in spatial sampling. This difference in bias between the methods as implications for our understanding of the social-ecological drivers of bird distribution in cities, of environmental and social legacy effects, and also for policy and management decisions relying on these kinds of citizen science datasets.

5.3 Objective 2: Determining the relationship between species richness and income

Although many studies suggest that there is a relationship between income and species richness, our regression test suggests that there are other underlying factors that may these relationships at our study sites. The underlying cause of the bias influencing this pattern might differ in each city because each city has a different culture, history, and spatial pattern of wealth.

For example in Raleigh, there is a very high amount of tree cover within the urban landscape that could possibly explain why we see high species richness across all income groups. Durham may have less tree cover than Raleigh, but it has also been undergoing urban renewal, which could be altering the vegetation structure within the urban matrix. The spatial pattern of wealth is also different in these two cities. In Raleigh, higher income areas are found in West and North

Raleigh while lower income areas are found in Southeast Raleigh. In Durham the wealth is concentrated near Duke University and downtown Durham, with pockets of lower income.

In the western cities of Tucson and Fresno, water usage and landscaping could be underlying factors of the bias causing the relationship between income and richness, where higher income allows for more water use/irrigation (Katti et al. 2017). The spatial distribution of wealth in Tucson results in the higher income areas being concentrated to the perimeter of

Tucson, reflecting a preference for proximity to the desert among wealthier people (often retired or winter visitors from colder northern climes) settling in Tucson. In Fresno, the majority of the city is low income and there are only a few pockets of wealth, especially along the northern edges of the city, closer to the San Joaquin river and the foothills of the Sierra Nevada, which are also sources of access to greater natural habitat and bird diversity.

6. Conclusion

I have discovered that there is an issue of volunteer bias in eBird data, and while the data is widely available for research use, my results urge caution in application of results from eBird datasets alone. If using eBird datasets in an urban study, there will be bias in the results based on the income (and likely racial, ethnic, and other social variables) distribution of the study area.

The difference in sampling effort displayed across income groups between the two methods is due to participant bias. When allowed a choice in where to sample, volunteers do not sample in a

way that accurately reflects the income distribution of the city. Due to the nature of the hobby, eBird users sample fewer locations with increased visits based on site potential for high bird diversity. The systematic count method is designed to be evenly distributed across all income groups in respect to area, and is able to capture a more accurate picture of the urban landscape.

The findings of this study are important to keep in mind for anyone who may be using eBird data within the urban landscape. The urban matrix is dynamic, not only through its landscape in the diversity of habitat types and land uses, but also because it is influenced by coupled socio-ecological systems. When using crowd-sourced datasets for urban planning or ecological research, we must keep in mind participant biases in our methodology and that it may influence how we see the landscape. It is important to use methods that are able to detect the underlying social drivers to create an accurate picture of the urban landscape.

While my study was able to identify that a systematic methods is a better option spatially, additional underlying variables are influencing bird richness in the study area. It is important to note that the variables influencing bird diversity in Western cities are different from the variables that influence cities in the East. The nature of those variables will vary in each city depending on the spatial pattern of wealth, culture, and history of the city. I would like to further explore natural variables such as tree cover, as well as other cultural and economic variables like patterns of urban renewal, to determine what their influence is, if any, on bird richness in the urban landscape. I would like to quantify urban habitats by zone usage (commercial, residential, industrial, park etc) and primary habitat type to better understand what variables may be influencing bird richness in cities.

Not only should we look into what current variables are shaping the bird community, but also what historical structures have helped to shape the communities we see today. An additional

use of this data would be to overlay it with historical redlining maps to see if the current pattern of eBird point distribution displays a similar relationship between income and bird locations or if the areas where people have chosen for bird watching has changed. It would be interesting to know if human movement within the city (where we build and invest resources) influences bird communities. It would be beneficial to explore if eBird locations are mostly parks or residential areas to better understand how volunteers are affecting bird data and what these biases might mean for utilizing eBird data at finer citywide scales.

All these variables are important avenues to explore because of the push to make cities more sustainable and resilient to climate change, not only for people but also for wildlife that inhabit these spaces. These considerations are also important from an environmental justice perspective, for a broader policy goal of ensuring people have equitable access to nature

(including birds) regardless of their socioeconomic or racial/ethnic background, or their spatial location within the urban landscape matrix. This research will have value to urban planners, ecologists and natural resource managers, currently and in the future, as we consider the best way to conserve habitat, and develop spaces within urban areas to make cities more ecologically, socially, and economically sustainable.

REFERENCES

Aronson, M.F.J., Nilon, C.H., Lepczyk, C.A., Parker, T.S., Warren, P.S., Cilliers, S.S. et al.

(2016). Hierarchical filters determine community assembly of urban species pools.

Ecology, 97, 2952-2963.

Butcher, G.S., Fuller, M.R., McAllister, L.S. & Geissler, P.H. (1990). An Evaluation of the

Christmas Bird Count for Monitoring Population Trends of Selected Species. Wildlife

Society Bulletin (1973-2006), 18, 129-134.

Bonter, D.N. & Cooper, C.B. (2012). Data validation in citizen science: a case study from

Project FeederWatch. Frontiers in Ecology and the Environment, 10, 305-309.

Callaghan, C.T., Bino, G., Major, R.E., Martin, J.M., Lyons, M.B. & Kingsford, R.T. (2019).

Heterogeneous urban green areas are bird diversity hotspots: insights using continental-

scale citizen science data. Landscape Ecology, 34, 1231-1246.

Callaghan, C.T. & Gawlik, D.E. (2015). Efficacy of eBird data as an aid in conservation

planning and monitoring. Journal of Field Ornithology, 86, 298-304.

Callaghan, C.T., Major, R.E., Lyons, M.B., Martin, J.M., Wilshire, J.H., Kingsford, R.T. et al.

(2019). Using citizen science data to define and track restoration targets in urban areas.

Journal of Applied Ecology, 56, 1998-2006.

Cooper, C.B., Dickinson, J., Phillips, T. & Bonney, R. (2007). Citizen science as a tool for

conservation in residential ecosystems. Ecology and Society, 12, 11.

Dickinson, J.L., Zuckerberg, B. & Bonter, D.N. (2010). Citizen Science as an Ecological

Research Tool: Challenges and Benefits. Annual Review of Ecology, Evolution, and

Systematics, Vol 41, 41, 149-172.

Eubanks, T.L., Stoll, J.R. & Ditton, R.B. (2004). Understanding the Diversity of Eight Birder

Sub-populations: Socio-demographic Characteristics, Motivations, Expenditures and Net

Benefits. Journal of Ecotourism, 3, 151-172.

Evans, C., Abrams, E., Reitsma, R., Roux, K., Salmonsen, L. & Marra, P.P. (2005). The

Neighborhood Nestwatch Program: Participant Outcomes of a Citizen-Science Ecological

Research Project. Conservation Biology, 19, 589-594.

Gerrish E, Watkins SL (2018) The relationship between urban forests and income: A 486 meta-

analysis. Landsc Urban Plan 170(April 2017):293–308.

Hensley, C.B. (2018). Effects of urbanization on bird assemblages in three southwestern U.S

cities. Fresno State Digital Repository URI: http://hdl.handle.net/10211.3/204556

Hensley, C.,, Trisos, C., Warren, P.S., MacFarland, J., Blumenshine, S., and Katti, M. 2019.

Effects of Urbanization on Native Bird Species in Three Southwestern US Cities.

Frontiers in Ecology and Evolution. https://doi.org/10.3389/fevo.2019.00071

Inkiläinen, E.N.M., McHale, M.R., Blank, G.B., James, A.L. & Nikinmaa, E. (2013). The role of

the residential urban forest in regulating throughfall: A case study in Raleigh, North

Carolina, USA.

Johnson, M.F., Hannah, C., Acton, L., Popovici, R., Karanth, K.K. & Weinthal, E. (2014).

Network environmentalism: Citizen scientists as agents for environmental advocacy.

Global Environmental Change-Human and Policy Dimensions, 29, 235-245.

Johnston, A., Moran, N., Musgrove, A., Fink, D. & Baillie, S.R. (2020). Estimating species

distributions from spatially biased citizen science data.

Kamp, J., Oppel, S., Heldbjerg, H., Nyegaard, T. & Donald, P.F. (2016). Unstructured citizen

science data fail to detect long-term population declines of common birds in Denmark.

Diversity and Distributions, 22, 1024-1035.

Katti, M., Jones, A.R., ÖZGÖÇ ÇAĞLAR, D., Delcore, H. D., Kar Gupta, K. 2017. The

Influence of Structural Conditions and Cultural Inertia on Water Usage and Landscape

Decision-Making in a California Metropolitan Area. Sustainability, 2017, 9, 1746.

http://dx.doi.org/10.3390/su9101746.

Kinzig, A.P., Warren, P., Martin, C., Hope, D. & Katti, M. (2005). The Effects of Human

Socioeconomic Status and Cultural Characteristics on Urban Patterns of Biodiversity.

La Sorte, F.A., Lepczyk, C.A., Aronson, M.F.J., Goddard, M.A., Katti, M. MacGregor-Fors, I.,

Mörtberg, U., Nilon, C., Warren, P.S., Williams, N.S.G., Yang, J. 2018. The phylogenetic

and functional diversity of regional breeding bird assemblages is reduced and constricted

through urbanization. Diversity and Distributions, 2018; 00:1–11.

https://doi.org/10.1111/ddi.12738. Full text: http://rdcu.be/I7Pu/.

Leong, M., Dunn, R.R. & Trautwein, M.D. (2018). Biodiversity and socioeconomics in the city:

a review of the luxury effect. Biology Letters, 14, 20180082.

Locke, D.H., Hall, B., Grove, M.J, Pickett, S. T.A., Ogden, L., Aoki, C., O’Neil-Dunne, J. P.M.

(2020). Residential housing segregation and urban tree canopy in 37 U.S cities. Available

from:

https://www.researchgate.net/publication/338410701_Residential_housing_segregat

ion_and_urban_tree_canopy_in_37_US_Cities

Loss, S.R., Ruiz, M.O. & Brawn, J.D. (2009). Relationships between avian diversity,

neighborhood age, income, and environmental characteristics of an urban landscape.

Martin, C.A., Warren, P.S. & Kinzig, A.P. (2004). Neighborhood socioeconomic status is a

useful predictor of perennial landscape vegetation in residential neighborhoods and

embedded small parks of Phoenix, AZ.

Martin, V.Y. (2017). Citizen Science as a Means for Increasing Public Engagement in Science:

Presumption or Possibility? Science Communication, 39, 142-168.

Melles, S.J., 2005. Urban bird diversity as an indicator of human social diversity and economic

inequality in Vancouver, British Columbia. Urb. Hab. 3, 25–48.

Miller, J.R. & Hobbs, R.J. (2002a). Conservation Where People Live and Work. Conservation

Biology, 16, 330-337.

Morelli, F., Tryjanowski, P., Møller, A. P., Katti, M., and Reif, J. 2020. Partitioning the Effects

of Urbanization on Biodiversity: Beyond Wildlife Behavioural Responses to a Multilevel

Assessment of Community Changes in Taxonomic, Functional and Phylogenetic

Diversity. Frontiers in Ecology and Evolution. February 2020, Vol 8, Article 23.

https://doi.org/10.3389/fevo.2020.00023

Pandya, R.E. (2012). A framework for engaging diverse communities in citizen science in the

US. Frontiers in Ecology and the Environment, 10, 314-317.

Pickett, S.T.A., Cadenasso, M.L., Grove, J.M., Boone, C.G., Groffman, P.M., Irwin, E., Kaushal,

S.S., Marshall, V., McGrath, B.P., and Nilon, C.H. (2011). Urban ecological systems:

Scientific foundations and a decade of progress.

Roman, L. A., Pearsall, H. E, Theodore S., Conway, T. M., Fahey, R. T., Landry, S., Vogt, J.,

van Doorn, N.S., Grove, J. M., Locke, D.H., Bardekjian, A. C., Battles, J. J.,

Cadenasso,M. L., van den Bosch, C. C., Konijnendijk; Avolio, M., Berland, A., Jenerette,

G. D., Mincey, S. K., Pataki, D.E., Staudhammer, C. (2018). Human and biophysical

legacies shape contemporary urban forests: A literature synthesis. Urban Forestry &

Urban Greening. 31: 157-168. https://doi.org/10.1016/j.ufug.2018.03.004.

McCaffery, R.E (2005). Using citizen science in urban bird studies. Urban Habitats 3, 70-86

Sandström, U.G., Angelstam, P. & Mikusiński, G. (2006). Ecological diversity of birds in

relation to the structure of urban green space.

Sauer, J. R., and S. Droege, editors. 1990. Survey designs and statistical methods for the

estimation of avian population trends. US. Fish Wildl. Serv., Biol Rep. 90(1). 166 pp

Tulloch, A.I.T., Possingham, H.P., Joseph, L.N., Szabo, J. & Martin, T.G. (2013). Realising the

full potential of citizen science monitoring programs. Biological Conservation, 165, 128-

138.

Turner, W.R. (2003). Citywide biological monitoring as a tool for ecology and conservation in

urban landscapes: the case of the Tucson Bird Count.

Watkins, S.L. & Gerrish, E. (2018). The relationship between urban forests and race: A meta-

analysis.

Wehn, U. & Almomani, A. (2019). Incentives and barriers for participation in community-based

environmental monitoring and information systems: A critical analysis and integration of

the literature.

Wiersma, Y.F. (2010). Birding 2.0: Citizen Science and Effective Monitoring in the Web 2.0

World. Avian Conservation and Ecology, 5, 13.

Wood, C., Sullivan, B., Iliff, M., Fink, D. & Kelling, S. (2011). eBird: Engaging Birders in

Science and Conservation.

APPENDICES

Appendix A

Table 2.1:A comparison of the demographic attributes of the four study areas. The GINI coeffiecient is a measure of inequality that displays the distribution of wealth in each city.

Demographics Raleigh Durham Tucson Fresno

Population(2018) 469,298 316,739 545,975 530,093 Median Household Income $63,890 $58,190 $41,625 $47,189 Maximum Household Income $191,944 $169,583 $134,444 $204,688 % of Persons in Poverty 14% 14% 23% 27% % of Population White 53% 43% 44% 27%

% of Population Black 29% 37% 5% 8 % % of Population 11% 14% 43% 49% Hispanic/Latino % of Population Asian 4 6% 3% 14% % % of Population over 65 10% 13% 13% 11% GINI COEFFICIENT 0.47 0.47 0.46 0.499 CITY AREA (Acres) 92,660 69,310 145,100 73,410 SAMPLE SIZE eBird 209 148 8 24 4 SAMPLE SIZE Systematic 356 277 696 260

Appendix B

Appendix C

Appendix D

Appendix E

Appendix F

Raleigh Durham

Figures 4.1 and 4.2: Results of the contingency analysis displayed as a stacked bar graph. The bar on the far right displays the proportion of points in each method without income. The proportion of points is displayed along the y-axis and median household income grouping is displayed across the x-axis. Income is grouped in increments of 30 thousand from 0 to 210 thousand, with the exception of group 6 which represents an increment of 60 thousand due to low sample size across the four cities. The numbers along the x-axis represent the following values: 1 = 0-30k; 2 = 30-60k; 3 = 60-90k; 4 = 90-120k; 5 = 120- 150k; 6 = 150-210k.

Fresno Tucson

Figures 4.3 and 4.4: Results of the contingency analysis displayed as a stacked bar graph. The bar on the far right displays the proportion of points in each method without income. The proportion of points is displayed along the y-axis and median household income grouping is displayed across the x-axis. Income is grouped in increments of 30 thousand from 0 to 210 thousand, with the exception of group 6 which represents an increment of 60 thousand due to low sample size across the four cities. The numbers along the x-axis represent the following values: 1 = 0-30k; 2 = 30-60k; 3 = 60-90k; 4 = 90-120k; 5 = 120- 150k; 6 = 150-210k.

Appendix G

Table. 4.1: Results of the contingency analysis displaying the proportion of points located in each income group across both methods in all four study areas.

Appendix H

Tucson eBird Tucson Systematic Count

Fresno eBird Fresno Systematic Count

Raleigh eBird Raleigh Systematic Count

Durham eBird Durham Systematic Count

Appendix I

Triangle Bird Count Protocols

Please read all instructions prior to surveying your route(s). Your safety is our number 1 priority. Adherence to these instructions will lead to more accurate and useful results, as well as a fun and safe experience. We’d like to give credit to the Tucson Bird Count and Fresno Bird Count for developing the following protocols.

Scouting Being that this is the first survey for the TriBC, it would be beneficial to visit your site before the actual survey. We have tried to screen census sites so that they are in a publicly accessible area.

In the event that a site on your route is not accessible (in a yard, in a restricted area, etc.), choose an alternate, publicly accessible site (sidewalk, street, park, etc.) as close as possible to the original site on your route map.

A few routes have remote sites, plan accordingly and for safety reasons, do not do remote routes alone. The same may apply to some urban routes in certain neighborhoods.

Do not attempt to conduct counts on any private property without consent of the owner, and only count on private property if no publicly accessible location is nearby. If the original site is in a yard or alley, move it to the nearest publicly accessible spot (street, park, etc). If you move a site, mark the exact location of the new site on your route map with a red ‘X’ and the Site ID number, and submit details on the new site location, providing GPS coordinates if possible. Your markings will be used to update sites in computerized TriBC maps for data analysis and for locating the same site in future years.

Equipment Checklist:

□ Transportation (car or bike for areas with roads, boots for more ‘remote’ areas) □ GPS unit (recommended, but you can also use your phone GPS) □ Binoculars □ Pencil or pen with dark ink □ Watch with second hand timer or phone timer □ Field guide

□ Clipboard □ TriBC instructions, route map, and data sheets □ Water, sunscreen, hat, etc. □ Data Sheets: To download a copy of the TriBC data sheets go to: trianglebirds.org

When to Run Routes Survey your route on a morning of your choosing during the survey period, April 15 - May 15. Many routes have one or more sites near roads that are noisy with weekday traffic; weekend mornings generally provide better counting conditions.

Acceptable Weather Occasional light drizzle or a brief shower may be acceptable, but fog, steady drizzle, or prolonged rain should be avoided. Counts should not be done if the wind exceeds 12 mph (loose paper and dust are raised, small branches on trees move).

Starting and Traversing Sites in a Route ● Choose a path connecting all the sites along your route in a way that will minimize (roughly) your total travel time (Site ID numbers are for identification only; you don’t need to survey them in numerical order). ● Begin counting at your first site as close to sunrise as possible (as early as 30 minutes before sunrise). ● No counting should take place more than 4 hours after sunrise. If all sites along a route are not surveyed within this time, survey unfinished sites another morning. Report that this was done in the “Notes” section of your data sheet, with a brief explanation (e.g., “sites x, y, and z involved hiking”). Counting Birds

These are 10-minute, point counts of all species seen or heard within a 40 m radius (130 ft) ● Only one observer should conduct the 10-minute point count at each site. ● Count from a stationary point outside of a car. ● Count every bird seen or heard by the primary observer during the 10-minute period. ● On the point count data sheet check the S box for seen, the H box for heard, or both. ● Only count birds within an estimated radius of 40 meters. ● Using arabic numbers (1, 2, 3 etc), mark how many of each species of birds are seen or heard within the appropriate distance band. 0-5 meters, 5-10 meters, 10-20 meters, or 20-40 meters. ● It is important to count each bird only once, even if it leaves the site.

● For large groups of birds, estimate the number. Be conservative in your counts if you are uncertain how many individuals there are. If you think you hear 3 American Crows, but can only be sure there are 2, then write down “2.” ● Do not exceed 10 minutes because you are sure a certain "good bird" is there and not calling — valid negative data are as important as positive in this survey. ● If you observe, but do not identify, a bird during a point count, write down descriptive features in the “Notes” section and spend time after the 10 minute period working on the ID. Record such birds as being in the point count. ● Don't use any method of coaxing birds ("spishing", tape playbacks). It's important that all point counts be done consistently to produce reliable results. ● Turn and face a different direction about every 1 minute, or you will miss birds behind you. Even if you count mostly by ear, this will help you pick up quieter birds.

Supplemental Observations: Birds detected by other observers in the group, detected outside the 10-minute window, or detected during your transit between sites, should be recorded in the Supplemental Observations Data Sheet. It is only necessary to record on this sheet those species that were not observed already during the Point Count for that site.

Rare or Unusual Birds: Any reports of rare species in the Triangle region or unusual in the area being surveyed should be supported by including some details of the observation in the "Notes" section. Include all features you used to determine ID.

Temporary Noise, Interruptions If a temporary noise (e.g., passing car) or interruption (e.g., inquisitive resident) interferes with your ability to count birds at a site, pause the count (and the clock) for the duration of the interruption, and resume counting when the interruption has passed. Total time counting birds (that is, not including interruptions) for the point count should be 10 minutes. Birds observed during an interruption, but not otherwise during the 10-minute count, should be reported in the Supplemental Observations column.

Constant Excessive Noise If constant noise interferes with your observations at a site, try to return to survey the site on a morning when it's quieter. The goal is to accurately survey birds at all sites — documenting the birds in a parking lot is as useful as recording the birds in

a wash. Thus, surveying the original site at another time is preferable to moving the site. If returning is impractical, move up to 50 walking paces (but no more) to a spot where the noise is reduced. Report noisy conditions (see Recording Data, below).

Submitting Data Please submit your data within 1 week of your count date. Please mail: (1) your Route Map, (2) updated site locations, and (3) all completed Route Data Forms to the address below. You may also wish to keep a copy for your records.

Mail completed forms to: Triangle Bird Count c/o Madhusudan Katti Dept of Forestry & Environmental Resources Campus Box 8008 NC State University Raleigh, NC 27695

Contact Information If you have any questions, don't hesitate to contact the TriBC Project Coordinator, Deja Perkins [email protected]

TriBC Principal Investigator: Dr. Madhusudan Katti, [email protected]

Triangle Bird Count SciStarter page: https://scistarter.com/project/18979-Triangle-Bird- Count

Recording Data

This section details the fields on the TriBC Route Data Form sheets. Record all data on these sheets, only 1 site per sheet. Use as many sheets as are required for your route, but do not include sites from more than one route on the same sheet.

ROUTE section: Record the Route Number and Primary Observer name on each sheet, and names of Others Present on the route's first sheet.

SITE section: Site ID #. Each site gets a column. Write the full Site ID number here.

Date & Site Start Time: For each site, record the date, and the time at which the point count was started.

Site Status: Check one box. Check “new or moved” if the site is in a physically different location from previous years.

Site Location: Site at which point count is conducted. This description must accurately describe the count location up to 30 feet. This information will be used to ensure that future TriBC's count birds at the same locations.

Sites on or near streets. Report the adjacent street address (e.g., “in front of 4549 Maple St”) or nearest street intersection. Report side of street or corner of intersection (e.g., “20 ft SE of the SE corner of Bullard and Blackstone”). If no marked address is near, record distance and direction along the street from the nearest street. Include any nearby landmarks, especially those likely to appear on maps (e.g., “Woodward Park, 100 ft NE of parking lot”). Including GPS coordinates (in addition to the above) if possible, though not essential if detailed street locations are provided.

Noise level: Moderate = your ability to detect birds by sound is hindered somewhat; Extreme = hindered almost completely.

NOTES section: Use this section for details of site relocations, sightings of unusual species, recording accessibility issues (e.g., "in the middle of golf course, need to receive permission from clubhouse or need to call X number in advance"), or any other important information about a site. Begin any notes with the relevant Site ID number.

BIRDS section: Write the Site ID # at the top of each column. In the species column on the left, use the Species List to write in the common name, or 4-letter AOU code (if you know it), of the species encountered. If writing the common name, please include the whole name with no abbreviations.

In the Point Count column, record the number of individuals of each species seen/heard during the 10-minute point count. If you use tick marks to tally birds as

you observe them, be sure to also write a distinct Arabic number (1, 2, 3, etc) of individuals for each species observed, and circle it to avoid confusion.

In the Distance Bands columns, record the number of individuals of each species seen in the appropriate box. Please use arabic numbers if possible.

Appendix J Triangle Bird Count Route Data Form Only record 1 site per page. Use additional pages as necessary. See Protocols for details of fields. Primary Observer (conducts point counts) Others present

ROUTE Complete all fields for all sites Site ID # Date Site start time Site location Describe location and access if site is new or moved, or if existing description can be improved. SITE Otherwise write “same”. Temperature (°F) Wind level none light moderate

Noise dB (if applicable) Noise level (circle one) none moderate extreme

detail observations of unusual birds (include all distinguishing characters used to make ID)

NOTES

*Only record one site on this page* Please check the S column for seen and the H column for heard or both. Also check the box for the distance band where the bird was seen or heard. Site ID # Distance Bands

Species Name # Birds S H 0-5m 5-10m 10-20m 20- 40m

BIRDS

Supplemental Observations Data Sheet Record any species that you may happen to see that were not included in your point counts. This could include any species at the site (but not during count) or en route to count. Site ID: Site ID: