People think there's no nature in cities, but they want to know more

We surveyed urban residents to study their knowledge, engagement, perception, and curiosity of bird biodiversity in four Peruvian cities in two different ecosystems (Lima and Huacho in the Pacific desert, and Nauta and Iquitos in the western Amazon). Surveys also included questions on social variables such as age, gender, education, outdoor activities, and years residing in the area. Bird point counts were carried out in the same sites where residents were surveyed. These sites were also visually assessed for greenspace cover and distance to nearest landmarks. Exploratory analyses (principal components analyses and factor analysis of mixed data) were carried out on environmental and survey data to summarize and select correlated variables for multivariate linear models. Amazonian city residents had higher “knowledge” and “engagement” scores than residents in the desert cities. Best performing models predicting “knowledge” scores suggested that urban residents learned about birds outside of the formal education sphere, although there were no strong common patterns among cities or in the full dataset. “Engagement” scores in the desert cities seemed to be linked to local and neighborhood greenspace and education, suggesting socioeconomic class plays a role. The overwhelming majority of all four cities’ respondents scored highly in “perception” and “curiosity” measures, implying that orientation toward nature is not lost in these four cities and that finding and promoting the human–nature connection in urban areas is a matter of asking the right questions and promoting existing nature practices and perspectives.


INTRODUCTION
Cities are full of people who can promote or subvert conservation and environmental priorities depending on their political and economic participation, attitudes, and connections with nature (Sanderson et al. 2018). When cities subvert conservation and environmental priorities, they encroach on arable land and natural habitat (e.g., Ewing 1997), support a human population with a higher demand for meat and other energy-intensive goods (e.g., de Haen et al. 2003), and negatively disrupt nitrogen cycling and microbial communities (e.g., Wang et al. 2017). By contrast, when they promote conservation and environmental priorities, cities have less energy use per capita than rural areas (Güneralp et al. 2017), house more people at a higher standard of living (Addanki and Venkataraman 2017), and present the opportunity to lay down a foundation for a smaller future impact on biodiversity (Sanderson et al. 2018). Thus, as the global population urbanizes, the human-nature connections of urban residents are a starting point toward understanding how humans lead the cities they live in on paths away from or toward sustainability.
Human connections with nature are influenced by the varied ways people relate to and value it (Chan et al. 2016). In cities, green spaces can provide direct benefits by promoting exercise (Wolch et al. 2014), providing new economic opportunities (Wolch et al. 2014), and serving as the setting for urban agriculture (Azunre et al. 2019). Cities also have trees mitigating the increased temperatures caused by urbanization (Derkzen et al. 2017, Lin andGou 2017), wetlands purifying water for use as tap water (de Groot et al. 2002), and biodiversity decreasing the chance of human infection by a generalist pathogen (Hassel et al. 2017). The aesthetic quality of nature has promoted scientific literacy through natural history activities (Wood et al. 2011), and both green and blue (water) spaces have facilitated community development through mental well-being and social cohesion (Wolch et al. 2014, Corburn 2017).
However, many cities are concrete canyons, not urban forests. In these gray spaces, scholars have theorized that humans lose their connection to nature, thus the desire to conserve it, through "the extinction of experience" (Pyle 1978, as cited in Pyle 2003. This process states that as the species richness to which humans are exposed decreases, the power of nature to incite humans into action is lost (Pyle 2003). Research into what causes extinction of experience has been conducted in the United States, Japan, Spain, Hong Kong, and the United Kingdom, among other countries (Botzat et al. 2016, Soga andGaston 2016). Most of this literature is restricted to experiences from the Global North, yet cities in the Global South are some of the fastest growing and are located in highly biodiverse areas (Pyle 2003). Whether the same patterns of loss of interest in nature by urban residents occur across the Global South has not been characterized in peer-reviewed literature but must be identified to provide pathways for peoples' participation in nature conservation and resource management (Castillo et al. 2006).
In this study, we asked: How do urban residents' knowledge, engagement, perception, and curiosity (KEPC) toward birds vary in four Peruvian cities and what factors may explain these variations? We chose birds as a study group because they are noticeable, hold aesthetic and cultural value, and can indicate environmental health (Ogden et al. 2014, Robinson 2019. Given the theory of extinction of experience, we hypothesized that bigger cities, particularly those in deserts, would have the lowest measures across all response variables. Within cities, we hypothesized that respondents who live in places with more greenspace and higher bird diversity would score higher on all response variables. . Maps constructed using R 4.0 in RStudio (R Core Team 2020, RStudio Team 2020) using the packages ggmap and ggsn (Kahle andWickham 2013, Santos Baquero 2019).

METHODS
To assess urban residents' KEPC and potential predictive factors, we collected environmental and bird diversity data at a site level, and socio-demographic data at an individual level.

City and site selection
We first chose the largest city (both in territory and population) in two Peruvian ecosystems (Lima in the Pacific Coastal Desert, and Iquitos in the Western Amazon), and then a smaller city about 3 hours away from each large city by highway (Huacho in the Pacific Coastal Desert, and Nauta in the Western Amazon; see Table 1 for general characteristics). This distance assured separation between cities, but proximity and inclusion in the same ecosystem. Difference in sizes gave another axis of comparison. Sites within cities were selected with the following criteria: > 200 m from any other site, < 500 m above sea level, easily accessible by car, and, together, spanning the expanse of the city. Number of sites per city differed on account of the difference in size of cities. The distance between sites assured separation, whereas the inclusion within the same elevational range (< 500 m asl) assured that the bird communities were influenced by the city landscape rather than by different elevations (see Fig. 1 for city maps).

Environmental and bird diversity data
Following Piland (2020), we classified each site visually using satellite images acquired from Google Earth to characterize both a local (radius = 100 m around site coordinates) and a neighborhood measurement (radius = 500 m) of greenspace and grayspace cover for each site (satellite images copyrighted to CNES/Airbus between 2016-2018, and accessed in 2019, Google Earth V.7.3.2.5776). Distance from nearest landmark (park, center of city, and water) edge to site coordinates was measured using the Google Earth distance tool on these same images. Satellite images processed were those available on Google Earth closest to bird point counts in 2017 and survey implementation in 2018. Bird point counts were conducted at every site coordinate in January-April 2017 as part of a related project (Grández et al., unpublished data). For this study, we used the following resulting https://www.ecologyandsociety.org/vol27/iss1/art25/ diversity metrics: total species richness (number of species seen at each site), average number of individuals (number of individual birds seen at each site averaged across visits), and Simpson evenness (a measure of species equitability based on number of individuals at site, where 1 means that species all occur at the same frequency). There were 97 sites assessed for environmental and bird diversity data.

Survey construction
The knowledge, perception, and attitude (KPA) protocol, first developed for public and community health social diagnostics (Gumucio et al. 2011), was adapted to assess urban residents' perception, knowledge, and attitudes toward bird diversity in their neighborhood (inspired by Celis-Diez et al. 2017). Four classes of response variables were chosen (KEPC) in order to capture more than just knowledge under the Western science paradigm, in an attempt to measure multiple ways of connecting with nature (Dickinson 2013 There were six surveyors for all four cities, with NCP  interviewing in all four cities, whereas the other five were split  with three in Iquitos and Nauta (XV, EV, and Claudia Pezo), and two in Lima and Huacho (MC and CP). Ten people were surveyed at every site with three exceptions: one in Lima where we surveyed nine respondents, another in Lima where we surveyed 11 respondents, and one in Nauta where we surveyed 11 respondents. Surveys were carried out between 9:00 AM and 6:00 PM every day of the week. Surveyors were trained in the study's aim and objectives, as well as how to ask questions. Respondents were informed of the study's objective and asked if they consented to taking part. Questions not about birds were asked first (Appendix 1: Questions 1-10), whereas questions on birds (Appendix 1: Questions 10-17) were asked in the same order and last. If there was a lack of clarity in the question, the question was explained using words not on the survey. In some cases, interviewers wrote down their interpretation of a respondent's gender rather than asking how they self-identify, and as such, all discussion of gender from here on out will be presumed to represent interpretation of gender by surveyor, rather than true gender.

Coding variables
As many questions in the survey were open-ended, responses required coding for analysis. Employment (Appendix 1: Question 4) was coded using the North American Industry Classification System (United States 2017) where the response was categorized into an "industry" variable. Information about the respondent's last residency change (Appendix 1: Question 6) was coded based on where their last residence was in relation to their current residence ("movement"). Activities outside (Appendix 1: Question 7) were classed into the following: "recreation" (main objective: to pass time, enjoy the outdoors, or socialize), "exercise" (main objective: to promote physical health), "extractive" (main objective: to consume), "work" (main objective: to produce), and "other" (main objective: outside the prior categories or where the respondent did not elaborate). In cases where people listed more than one activity, both were counted. In the analysis, dummy variables were created for each answer. Respondents were asked to specify topics of interest in regards to birds (Appendix 1: Question 17) and these were categorized broadly by whether the topic was centered on the bird or the human, and whether it related to that focus directly, indirectly, aesthetically, recreationally, or intrinsically. These categorizations are inspired by the human valuation of nature paradigm where nature's value to human society is relative to direct harm or benefit (Direct), indirect ecosystem services (Indirect), aesthetic/cultural meaning (Aesthetic), recreational opportunities (Recreational), or existence (Intrinsic; Primack 2014).

Data analysis
All analyses were conducted in R 4.0 and RStudio, using the tidyverse R packages for data wrangling and visualization (Wickham et al. 2019, R Core Team 2020, RStudio Team 2020). Descriptive statistics (minimum, maximum, median, mean, and standard deviation for numeric variables or frequency in percentage for factor variables) were calculated for the full dataset and for each individual city using R package summarytools (Comtois 2020). https://www.ecologyandsociety.org/vol27/iss1/art25/

Exploratory analyses
Principal component analysis (PCA) was used to explicitly summarize relationships between environmental and bird diversity variables in the full dataset and identify the dimension best characterizing the urban typology. The score on this dimension was used to describe each site (from here on, "greenspace score") in the models. Factor analysis of mixed data (FAMD) was used to summarize and visualize dimensions of attitudes on bird diversity by urban residents. Individuals were visualized on FAMD dimensions 1-4. All PCAs and FAMDs were run using the R package FactoMineR and visualized using factoextra (Lê 2008, Kassambara andMundt 2020).

Model construction
Multivariate linear regression models were conducted using each FAMD dimension 1-4 score as a function of the following variables: age, gender, education, work, outdoor activities, weekly hours outside, residency length, last move, observed species richness, average number of individual birds, greenspace score, and Simpson evenness. Site was not included as a random factor, because the site variables (observed species richness, average number of individual birds, greenspace score, and Simpson evenness) deconstruct the site and are hypothesized to be potential factors in determining respondents' KEPC. Outlier respondents, i.e., those with unique categorical score, were excluded from the models. Three sites from Iquitos were excluded because their Simpson evenness bird scores were significantly higher than the rest: 0.1-0.4, whereas the mean of the rest of the sites was 0.04 with a standard deviation of 0.01. This pattern seems to be driven in two sites by the lack of Brotogeris spp. flocks, resulting in a more even species abundance distribution, and in the third site a low species diversity (n = 3) also resulting in a more even distribution of individuals. Full dataset models were run with and without the City variable as an independent variable. Models were also run on subsets of the dataset by city. All models used a 10fold cross-validation as the resampling technique, and stepwise selection. Best performing models within each run were chosen based on the lowest value of root-mean-square-error (RMSE). These analyses were conducted using the caret and leaps R packages (Kuhn 2020, Lumley 2020.

Descriptive statistics
Full descriptive statistics are included in Appendix 2. There were 961 respondents in the four cities of Iquitos (n = 350), Nauta (n = 101), Huacho (n = 180), and Lima (n = 330 6.7% and 10.6%, respectively) or in administration and support (Lima, 6.6%). In Nauta, the workforce of respondents was further made up of educators (15.8%) and those who worked in agriculture, forestry, fishing, and hunting (14.9%). Lima's respondents work more frequently in market-based jobs (retail and administration and support) than those in the other three cities. Nauta's respondent workforce reflects the local economy, which revolves around agriculture, hunting, and fishing, and the fact that the city has schools from primary to post-secondary levels, with a higher proportion of educators and people who work directly with natural resources (Ministerio de Transportes y Comunicaciones 2008).
These differences between cities are also reflected in the nature of the activities done outside. Although in Huacho, Lima, and Iquitos respondents rarely worked outside (7.8%, 3.3%, 7.4% respectively), 14.9% of Nauta respondents reported working outside. Nauta also seems to be a more transient city with respondents' average residency length being 15.1 years, whereas in the other cities average residency length for respondents was between 18 and 20 years. The most common movement respondents in Nauta experienced was moving from rural to urban areas (35.6%), whereas in the other three cities it was within the same city (35% in Huacho, 37% in Lima, and 52.9% in Iquitos). Mobility within Nauta for respondents was relatively low (26.7% in city and 2% in the same neighborhood) and this may reflect the relatively recent growth given that it positions itself as an important port and connector to Iquitos from the rest of the country (Ministerio de Transportes y Comunicaciones 2008, Figallo and Vergara 2014; Table 1). Lima is the only study city whose respondents reported international moves (2.4%).
Respondents from the Amazonian cities, Iquitos and Nauta, gave more responses for the bird identification exercise, and on average identified more illustrations correctly than those of the coastal desert cities ( Fig. 2b-d). These cities also had more respondents with gardens and had a higher average number of reported species in their garden than those of the desert cities ( Fig. 3c-d). Both large cities, Lima and Iquitos, had a higher fraction of pet owners than their smaller counterparts (Fig. 3b). Despite these differences, over 90% of respondents had noticed birds in their neighborhoods ( Fig. 4b). Over half of respondents (64.1%, with individual cities ranging from 53.2% to 73.6%) believed the bird diversity in their neighborhoods to be low and, on average, estimated that about four different types of birds occurred in their neighborhood (species richness based on bird surveys varied from three to 35 in 87 sites, averaging at about 17 species/site; Fig. 4cd). Over 70% of total respondents in each of the four cities wanted to know more about birds (Fig. 5b). The most popular category of subject matter people wanted to know about birds was "birds intrinsic," a category consisting of information about the birds themselves, e.g., identification, habitats, behavior, migration, and evolution.

Exploratory analyses
The full environmental PCA to identify an appropriate dimension to characterize greenspace and grayspace was performed on 97 sites using the following log-transformed traits: percent land cover, at both local and neighborhood scales, of trees, buildings, grass, asphalt, and dirt; distance to nearest landmarks of park, water, and center of city; average number of individual birds;

Fig. 2.
Results for "knowledge" response variables: (a) Boxplots for factor analysis of mixed data (FAMD) Dimension 1 score by city; (b) Number of birds responded on bird identification survey by city; (c) Birds identified correctly when graded according to either an "accepted" common name or a scientific name; (d) Birds identified correctly when graded according to the most popular name given; and (e) Best performing models for each dataset (Huacho, Iquitos, Lima, Nauta, and the full combined dataset) with variable coefficients and quality measurements (root-mean-square-error [RMSE], R², and Bayesian information criterion [BIC]).
Simpson evenness; and species richness. Percent land cover at both scales of water and wetlands were excluded because they did not show normal distributions. Dimension 1 (PC1; "greenspace score") demonstrated an urban gradient, where higher numbers were associated with greenspace and trees and lower numbers with buildings and asphalt. Cities showed a progression along the urban gradient with sites in Huacho and Lima tending toward negative PC1 values and more urban characteristics (for example, higher % of building cover), Iquitos sites spread out on the middle of the axis, and Nauta tending toward positive PC1 values and more tree-heavy or grass-heavy sites. Observed species richness is most strongly associated with dimension 1 (less urban, more birds). Simpson evenness is also associated with dimension 1 (less urban, less evenness). The average number of individual birds at a site was positively associated with dimensions 1 and 2 (less urban, more birds). Interpretation of other dimensions and PCA loadings are included in Appendix 3.
The FAMD for all respondents resulted in a first dimension that summarized "knowledge" variables (explaining 23.98% of variance), a second dimension that summarized "curiosity" (12.82%), a third dimension that summarized "engagement" (11.31%), and a fourth dimension that summarized "perception" (10.24%; Table 2). Individual respondents' survey responses were mapped onto these four dimensions with 95% confidence ellipses to assess visually whether any of the independent variables (human or environmental) display clustering. The only variable that displayed strong clustering was city membership (see Figs. 2-5 for differences in score distribution by city). Categories describing a respondent's interest were excluded from this analysis because the large number of potential answers overwhelmed the construction of dimensions. Outlier surveys (surveys whose contribution to the construction of dimensions singularly defined it) were excluded and the FAMD was re-run. The final FAMD included 956 surveys. The best models for the response variable "knowledge" were quite different among cities (Fig. 2e). In Iquitos, a variety of work types most strongly affected the "knowledge" score (coefficients between 1 and 2), whereas working outside and a higher level of education were strong negative factors. Level of education in Lima also had a negative coefficient, although so small that it is likely not seriously affecting respondents' knowledge, whereas in Huacho and in Nauta, working in education was a strong indicator of knowledge about birds. In Huacho, working in agriculture had a negative influence on knowledge of birds. The strongest model for the knowledge dimension was the full dataset, including the city variable. In this model, residency in Nauta or in Iquitos alone was the best predictor of bird knowledge.
When assessing models with the response variable "curiosity" score, the strongest negative coefficients are found in work types that require time inside: e.g., in the Huacho model, working in manufacturing, or in the Iquitos model, working in administration and support, both with coefficients less than -1.0. The full dataset model is the strongest model if judged by BIC (BIC = -82.05) and no time spent outside has the strongest negative coefficient (-0.49). However, the distributions of scores and variables show that there are few strong differences in response variables across cities (Fig. 5a-c).
Factors predicting score of engagement, as measured by ownership of pets and gardens, and garden species richness, differed among cities (Fig. 3e). In Huacho and Lima, the most interpretable factor is the environmental dimension score: places with a higher "greenspace score", and thus more greenspace/trees, are more likely to have residents with pets and gardens. In Iquitos' best performing model, Simpson evenness had the largest coefficient. This coefficient was strongly negatively associated with the engagement score and may also be a stand-in for a greenspace score, where a higher Simpson evenness score suggests less species richness and a higher amount of grayspace. In Nauta, being a student made a respondent more likely to have a pet or a garden, although we do not know who in the family chose to have a pet or garden.
Finally, the best performing multivariate models for the fourth FAMD dimension score, "perception", at most explain 9.4% of the variance (see R² of the Lima model; Fig. 4e). The strongest coefficient was from Nauta where people working in the science and technology sector had a positive association (2.58). The strongest negative association was residency in Nauta in the full dataset. There were no common factors found between the five models.

DISCUSSION
"Extinction of experience" leads us to believe that residents of big cities, regardless of where they are, have fewer opportunities to interact with nature, and thus have lower scores across measures of knowledge, engagement, perception, and curiosity (KEPC) toward nature. However, we show that urban residents' KEPC may depend more on the socio-environmental context of the city they reside in than on the size of it. Survey respondents in the Pacific coastal desert have lower "knowledge" scores than those in the western Amazon, yet the best models in each city differed. Likewise, respondents in the desert have lower "engagement" scores than those in the Amazon, but the scores for respondents in the desert seem tied to availability of greenspace. "Perception" scores are low across all four cities, although the Amazonian cities have lower scores, and "curiosity" scores are high across all four cities. Together, these results suggest a potential urban humannature connection, grounded in the city in which people live.

Urban residents in the western Amazon "know" more about birds
Formal education has been thought to predict scientific and natural history literacy (Luck et al. 2009, Celis-Diez et al. 2017), yet urban residents from the Pacific coastal desert consistently scored lower along the "knowledge" axis and variables than urban residents from the western Amazon despite having more formal education (Fig. 2), suggesting that formal education, at least in the Pacific coastal desert, does not influence whether urban residents can identify local birds. The best model for the full dataset identified residence in Nauta or Iquitos as the strongest predictive factor for high "knowledge" score ( Fig. 2e). Residents from these cities also consistently scored higher in the bird identification exercise, despite having less formal education than those of the desert cities. This points to a stronger extracurricular environmental awareness. The regional economy for Nauta and Iquitos, as with the rest of the Amazon, revolves around its natural resources, with birds providing an attractant for tourism, for which Iquitos and Nauta are points of access (Figallo and Vergara 2014). By contrast, in the region of Lima, the cities Huacho and Lima are industrial and commercial centers (Lazarte 2015). Results from these cities show similar knowledge results to those found in Santiago, Chile, another capital city in a similar, albeit higher elevation, ecosystem (Celis-Diez et al. 2017, where most respondents identified two to three birds correctly out of 10).
Urban residents' ownership of gardens and pets is constrained by environment "Engagement" scores for the Pacific coastal desert cities suggest that socioeconomic factors are more important for determining these scores than in the Amazonian cities, where residents had more gardens with more garden species more consistently (Fig.  3c-d). The overall model identified a high greenspace score as a positive factor in determining engagement score, yet the cityspecific models differed in the role of the environment versus demographic factors (Fig. 3e). In Lima, the best model identified unemployment and moving from a rural area as negative factors -two factors that are indicative of low socioeconomic class in the capital. From a resources perspective, this makes sense: water access constrains the ability to have a garden, and water access (as well as more greenspace) is linked to higher socioeconomic levels (Ioris 2012, Bell 2015, Alata Ninapaytán et al. 2019. Water access for plants may not be as linked to higher socioeconomic levels in the Amazon where high levels of rain and the surrounding wetlands make water for gardens easily accessible. This is supported by the fact that non-environmental coefficients in the models for Iquitos and Nauta are low (Fig. 3e). Simpson evenness, however, had a very high coefficient in Iquitos and could function as a potential proxy for how urban a site is given that a higher Simpson evenness score suggests a lower species richness and a higher amount of grayspace, and thus less space or water for a garden or a pet.
However, these measures of quantifying "engagement" should be viewed with caution because the assumptions of what they measure come from another part of the world. What it means to have a garden or a pet is defined culturally and may vary from place to place. For example, some respondents said they did not have a pet, but then later mentioned feeding stray animals. Our questionnaire also used the word "jardín", connoting a flower garden, rather than "huerta", which suggests a vegetable or fruit garden. We quickly incorporated "huerta" into the conversation we had with respondents while implementing the survey; however, this may suggest that these percentages are underestimates of garden ownership in our study cities. Additionally, the concept of  suggesting that although water access and environment may play a part in these differences, so does culture and socio-geographic position. Culture and socio-geographic position also influence how we interpret our results. For example, having property in both rural and urban areas can be a mark of a higher socioeconomic class in the global North, requiring enough wealth to purchase and manage two properties (Mendham andCurtis 2010, Stiman 2020). This is not necessarily the case in the global South, where there are still prevalent concepts of community ownership and family connectivity along the rural-urban gradient, land ownership can be acquired through settlement rather than purchase (including in urban areas), and property taxes are not high or enforced (Browder and Godfrey 1997, Padoch et al. 2014, Diep et al. 2019. Therefore, ownership of gardens or pets and https://www.ecologyandsociety.org/vol27/iss1/art25/ whether it relates to engagement with nature depends on the city and where in that city you live.

Urban residents' perception of bird diversity in their city may also depend on ecosystem
Residents in all four cities perceived their neighborhood's bird diversity as low, yet the proportion of residents per city with this answer is higher in the Amazon than in the Pacific coastal desert (Fig. 4b-c), suggesting that how urban residents perceive their neighborhoods' bird diversity is a function of context. Cities are generally less biodiverse than their surroundings and cities' bird diversity measures are more similar to each other than to their surroundings (McKinley 2006, Aronson et al. 2014, Sol et al. 2020, but this does not mean that cities are completely devoid of bird diversity or that they are always less speciose than their surroundings (Piland 2020). In some cities, the input of energy and resources by humans supports more individual birds (Marzluff 2001), and this seems to be the case in Lima, where the average number of individual birds per site (41.07 ± 29.1) was higher than Huacho (35.72 ± 17.5) or Iquitos (34.5 ± 11.7). In Nauta, the average number of individual birds per site is 50.04 ± 23.8 and may still be relying on a mixture of urban and rural resources (a situation hypothesized with theoretical models; Anderies et al. 2007).
In the Amazon, cities' surroundings are some of the most bird speciose in the world; in comparison, the cities themselves are species-poor. In the Pacific coastal desert, the surroundings, although still speciose, harbor birds that live in patches or are harder to see. Thus, in comparison, the cities have a similar number of bird species and sometimes more (Piland 2020). In our study, we registered a range of three species in one site in Iquitos to 35 species in another site in Iquitos, and a mean species richness ranging from 11.5 species/site (Huacho) to 23.9 species/site (Nauta). When asked to quantify, residents on average said there were about four different species of birds regardless of how urban the site was or how many species were observed during point counts (Fig. 6). Yet, a higher proportion of people in Nauta and Iquitos categorized their neighborhood as having "low" bird diversity (Fig. 4c).

Urban residents across the board want to know more about birds
In all four cities, over 75% wanted to know more about the birds in their neighborhoods (Fig. 5b). From the group of residents that were interested in learning more about the birds in their neighborhoods, the majority of respondents were interested in information about birds themselves. The second most popular general category of response differed by region: those of the coastal desert were most interested in information about how birds relate to humans directly, e.g., how to keep them as pets or raise them, whether they transmit diseases, etc. (about 16% for both cities), whereas in the lowland rainforest, respondents were interested in information about what supports or interacts with the birds, in particular, their habitat, their ecological relationships, etc. (~17% in Iquitos and ~13% in Nauta). After these three categories, the most popular category for all four cities was information for information's sake. Many respondents said they would be interested in learning more simply because it is always good to know more. This suggests that although urban areas may diminish the opportunity to spend time in places considered to be "natural", it does not diminish "orientation", that is, the ability to perceive, be interested in, and be motivated to interact with natural elements (in these cities, with birds). This aspect has been shown to be more important than "opportunity" in other cities (such as in Brisbane, Australia; Lin et al. 2014), and has been theorized to be another casualty of urbanization as a driver of "extinction of experience" (Soga and Akasaka 2019).
Many studies, when they ask about "orientation," "knowledge," or "connection," focus on specific types of these that may not be appropriate for all populations (Dickinson 2013). Connectedness to and willingness to act in the interest of nature have been measured in different ways: the Connectedness to Nature Scale (e.g., Mayer andFrantz 2004, Olivos et al. 2011); visits, hiking, and backpacking in parks (e.g., Zaradic et al. 2009, Wright andMatthews 2015); and knowledge of Linnaean classifications of wildlife (e.g., Cotterill 1995, Celis-Diez et al. 2017). Yet, the Connectedness to Nature Scale has been shown to be biased and ineffective (Perrin andBenassi 2009, Zhang et al. 2014). Asking about visits to protected parks in the United States prioritizes the experiences of those that established the protected parks or those like them, often those who are in more racially or economically privileged classes (Byrne and Wolch 2009, Wolch et al. 2014, O'Brien 2015, Mock 2017. Knowledge of Linnean classifications assesses only one type of identification of wildlife that has a history of being inaccessible to those outside groups of power, and glorifies an individual whose original classifications included racist characterizations of humans (Boster et al. 1986, Spickard et al. 1995, Dickinson 2013. In this study, we show a multipronged approach to assessing human-nature connection, and recommend in further studies the use of methodologies where participants have more control over how connections are characterized, for example, photovoice methodologies (Wang and Burris, 1997).

The possibility of an urban human-nature connection
Respondents in all four of our study cities perceived and were interested in birds across neighborhoods, with different socioeconomic, cultural, and racial histories (Figs. 4 and 5). The fact that urban residents are most interested in information about the birds themselves rather than how to capitalize on them shows that there are other non-monetary benefits that nature may https://www.ecologyandsociety.org/vol27/iss1/art25/ provide for urban residents (for more information on potential benefits, see Masterson et al. 2019). Supporting these already existing orientations, while attempting to act more as a translator between different economies and politics rather than as a "change agent", can empower community participation in politics, selfdetermination, and, ultimately, environmental management and stewardship (Wali et al. 2017).
Additionally, our research runs counter to literature that prescribes a focus on children as a long-term investment to prevent the extinction of experience (Miller 2005), because children have the strongest possibilities in creating a connection to nature (Chawla 1999). During survey implementation, many survey respondents of all ages went on to engage in a conversation to suggest ideas on how to engage residents in noticing and learning about neighborhood birds. One respondent from Lima was very excited to think about ways to involve their child: they mentioned puppet shows about birds at the park, interactive activities at the park, and creating local field guides. Adult respondents in all four cities shared many thoughts and questions on birds that were hard to categorize within this context. Some of these questions included: "Why do we not have the lions and tigers that we see on TV? Why are the birds that we see here not on TV?"; "Do the birds sing because they're sad?". The focus on children is important, but children do not exist in a vacuum; they interact with adults who themselves can communicate their connection to nature (Schuttler et al. 2018). Harnessing the excitement of the parents and other adults is beneficial, not just in engaging in a larger conversation and answering questions together, but also for society. Limiting our scope to generational change artificially transfers responsibility to the next generation when we can start creating a more environmentally and socially just society now (Carruthers 2018).

CONCLUSION
The environmental movement's romanticization of wildness and pristine landscapes has painted nature as inaccessible for many urban residents and fomented the idea of extinction of experience in urban areas (Cronon 1995, Pyle 2003). Yet, these ideas of "wildness" and "pristine" are not well supported by experience or evidence. For example, most natural areas, such as the contiguous forests of the Amazon, are the settings in which humans have evolved and continue to evolve, and they have been formed by this relationship (Heckenberger et al. 2014). In The Nature of Cities, Marianne Krasny (Cornell University) writes of her surprise at the strength of the connection to nature she felt in the Lower East Side of Manhattan (Krasny 2015). Instead of feeling this connection from the sight of a pristine alpine landscape, or a wild oceanic wave, she feels it from meeting a family working a community garden. She continues to reflect that, in preventing "extinction of experience", it is necessary to define what the experience you're preserving is. We expected to see patterns of loss of KEPC and thus a growing indifference as cities and neighborhoods become more urban. However, although knowledge and engagement (garden and pet ownership) did generally decrease from smallest city to largest city, the complexity and diversity of the best performing models suggest interactions between cities' natural and social histories. Perception and curiosity measurements remained broadly constant across cities, underlining that the urban environment itself may not be a definitive factor in people's disconnection to nature. Respondents in these urban communities are noticing and are interested in nature, and could thus lead their communities toward sustainability. As conservation professionals, our role could lie in connecting residents to those who may support their initiatives financially, politically, or technologically, and, as residents, our role could lie in engaging with our own communities.

Consent Language
Please consider participating in our voluntary survey. The survey should only take 10 minutes to complete and the data will not be linked to your identity. As researchers, we are trying to understand how people in this city interact with nature and how well they know the birds in their surroundings. These data will be used for biologist Natalia Piland's doctoral thesis. If you would like to ask questions, please contact her at the email: npiland@uchicago.edu

Survey
Note for the surveyor: This survey is designed to understand some factors that may affect how a resident relates to nature in their neighborhood. These factors include age, education level, residency length, where they're from, how much time is spent on which activities outside. The human-nature relationship assessed in this survey is knowledge of birds and motivation in knowing more about birds. This will be correlated with information about each site, including birds observed and some environmental factors like percent of green areas (quantified through remote sensing).   PCA results Table A3.1. PCA loadings (rounded to 2 decimal places) for each environmental variable used. All shaded cells correspond to either the two highest or two lowest loadings for respective dimension. Those shaded in green indicate variables associated with "greenspace", in grey associated with "greyspace" or "concrete cover", and in beige associated with "dirt." In parentheses, the percent of variance explained by the respective dimension. Interpretation of PCA results: Dimension 2 (PC2) suggested a gradient of greenspace versus dirt open area. Dimensions 3 and 4 were less clearly interpreted, although they seem to be related to distance from the center of the city (PC3) and distance from the nearest park (PC4). The four cities displayed little differences along PC2, except for Huacho. The variables measured through bird surveys had less clear associations with the different dimensions. Observed species richness has relatively strong associations with dimensions 2, 3, and 4 (more dirt, less species; further away from center, less species; further away from nearest park, more species). Simpson evenness is associated with dimension 4 in the following manner: less urban, less evenness. Average number of individual birds is associated with dimension 2 such that less dirt, means more birds, and negatively associated with dimensions 3 and 4 (closer to center, more birds; in places closer to parks, more birds).