Modeling aesthetic ecosystem services in megacity streetscapes

Bridging perception and landscape structure: mapping urban streetscape aesthetics as nature’s contributions to people | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Bridging perception and landscape structure: mapping urban streetscape aesthetics as nature’s contributions to people Douglas W. Cirino, Nicolas Mouquet, Jean Paul Metzger This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5109482/v3 This work is licensed under a CC BY 4.0 License Status: Under Review Version 3 posted 12 You are reading this latest preprint version Show more versions Abstract Urban streetscapes serve as essential public domains shaping human perception, sense of community identity and well-being. An integral aspect of this perception lies in the appreciation of the "environmental scenic beauty", reflecting individuals' personal comfort and sense of connection to their surroundings. Our study investigates the nuanced factors influencing the provision of urban aesthetic ecosystem services — a key non-material benefit with implications for mental well-being and urban quality of life here framed as a Nature’s Contribution to People (NCP). Employing online surveys, deep learning analyses, and spatial modeling, we bridge ground-level perception with landscape-level features, exploring the intricate interplay between green and built areas in shaping aesthetic preferences in São Paulo’s streets — the largest megacity in the Southern Hemisphere and a highly diverse urban environment. We found that the perceived beauty of streets is positively affected by the heterogeneous arrangement of vegetation and built-up areas and by the three-dimensionality of trees — and not solely by the quantity of greenery. Surprisingly, socioeconomic profiles of respondents exhibit no discernible impact on aesthetic evaluations, suggesting consensus across people with diverse social characteristics. Using convolutional neural networks trained on our survey, we predicted aesthetic scores for over 350,000 street images, yielding for the first time a map of the scenic beauty ecosystem service of an entire megacity. This aesthetic map uncovers significant mismatches between supply and demand for aesthetic services, exposing urban inequalities. By revealing these drivers and spatial patterns, our framework provides actionable insights for policymakers — linking perception and landscape-level planning — and offers a pathway to cultivate more socially equitable and aesthetically meaningful urban environments. Earth and environmental sciences/Environmental social sciences Social science/Environmental studies Scientific community and society/Geography Social science/Geography Urban Sustainability Relational Values Cultural Ecosystem Services Landscape Heterogeneity Environmental Justice Urban Ecology Megacities Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 INTRODUCTION The environment in which people reside and their perception of it are intimately intertwined with human well-being and health. This connection gives rise to a spectrum of effects, ranging from beneficial to detrimental (Beil & Hanes, 2013 ; Cirino et al., 2022 ; Dobbs et al., 2011 ; Teixeira et al., 2019 ). An integral aspect of this perception lies in the appreciation of the "environmental scenic beauty", reflecting individuals' personal comfort and sense of connection to their surroundings (Knez et al., 2018 ). Often regarded as a cultural ecosystem service (CES), this aesthetical non-material perception, linked to a sense of place, social cohesion, and spirituality (Cabana et al., 2020 ), is related to relational values (Chan et al., 2016 ) and ultimately is a nature contribution to people (NCP) ((Díaz et al., 2018 ). This value is frequently sought in natural settings or scenic locations, such as in natural parks, wilderness areas, and other remote spots offering beautiful views (Sandifer et al., 2015 ), but regardless of the contribution of those once in a time landscape experiences , this aesthetic service plays a vital role in our daily lives, particularly within urban landscapes (Lenda et al., 2023 ) where most individuals spend most of their time. In urban environments, people are less prone to having daily contact with natural and semi-natural environments, leading to an extinction of the human-nature experience (Soga & Gaston, 2016 ). While greenery plays a central role in shaping the aesthetic appeal of urban streets, the spatial configuration of vegetation, built infrastructure, and their interplay are critical but underexplored dimensions in urban ecosystem service provision (Chen et al., 2016; Weber et al., 2008 ). This study builds on landscape ecology principles to analyze how heterogeneity and spatial arrangement contribute to the aesthetic cultural ecosystem service, expanding beyond the traditional focus on green cover alone(Chen et al., 2016a ; Weber et al., 2008 ). The streetscape aesthetics are part of the artificial, natural and semi-natural aspects of the urban ecosystem, and are referred here as a Cultural Ecosystem Service. Particularly, the interspersing of green and built-up areas, the heterogeneity of the landscape or the diversity of green or blue areas should affect the perception of urban aesthetics (Daniel et al., 2012 ; Sabbion, 2018 ). The beauty of public spaces, such as streets, should play an important role in people's well-being (Bratman et al., 2019 ). Streets, being primarily for public use, serve as open pathways for all citizens and visitors and are the main commuting location for urban populations. The appearance of streets influences more than just their visual appeal—it also affects community pride, feelings of safety, ease of walking, and the expression of culture (Henderson et al., 2016 ). Therefore, streets provide a shared, public aesthetic and cultural experience. Despite the considerable impact of urban environments, especially streets, on human quality of life (Hartig & Kahn, 2016 ), there has been a scarcity of studies examining how people perceive the aesthetics of city streets. Most of those studies are small-scale investigations that offer valuable insights into citizen preferences concerning urban environments, but with limited ability to extrapolate or make assumptions on larger scales (Chen et al., 2016; Weber et al., 2008 ). One notable initiative that attempted to overcome this limitation is the Place Pulse Project (Dubey et al., 2016 ), which applied deep learning to predict perceptions of safety, liveliness, boredom, wealth, depression, and beauty across 56 cities using Google Street View imagery. Place Pulse was a groundbreaking effort that has since been adapted for studies assessing urban perception in various cities (Rossetti et al., 2019 ; Zhang et al., 2018 ). However, while these models successfully predicted how people perceive certain attributes of streets, they did not explicitly analyze the underlying landscape features that drive these perceptions. In other words, they could classify "beautiful" streets but could not determine which urban and ecological characteristics contribute to this beauty, they used extensive survey and cutting-edge technology, but did not connect the human perception to real landscape and ecological attributes. Here, we deepen this approach by integrating landscape ecology principles into urban perception studies, analyzing landscape features related to human perception. Our study bridges the perspectives of urban dwellers (seeing from the front) and landscape ecologists and urban planners (seeing from above) by analyzing both image-level features (composition, color, complexity) and landscape-level metrics (patch configuration, spatial heterogeneity, three dimensionality and urban form). Here, we examine how people perceive the beauty of streetscapes — the 'daily landscapes' — using São Paulo, the largest megacity in the Southern Hemisphere, as our model system. Beyond that, the understanding of how streets are essentially part of the urban ecosystem is limited, and it is urgent to treat those human-made environments under an ecological lens. Our objective is to overcome this narrow and dichotomous view – between the natural and the build environment – to a more holistic and relational approach between the city’s semi-natural daily landscapes and their dwellers. By integrating the perception of the landscape to the interplay between build and biological elements of this ecosystem we can overcome the limitations on understanding the provision of CES in everyday life. Although highly anthropogenic, urban streetscapes sustain ecological processes making them legitimate units for analyzing nature’s aesthetic contributions to people. To operationalize this perspective, we developed an integrated approach combining perceptual data, landscape metrics, and deep learning to connect what people see at street level with what ecologists and planners measure from above (Fig. 1 ). We conducted an extensive online survey with over 3,000 participants encompassing diverse social and demographic profiles. Each participant evaluated pairs of Google Street View images (SVIs) depicting São Paulo’s streetscapes, selecting which street they perceived as more beautiful. These pairwise comparisons were then translated into continuous aesthetic values through Elo score computation (Elo, 2008 ). The resulting scores provided a robust, perception-based measure of urban beauty across 420 sampled images, representing the socio-morphological diversity of São Paulo—the largest megacity in the Southern Hemisphere. Aiming to understand the landscape factors shaping these perceptions, we analyzed each georeferenced image within its surrounding urban fabric using high-resolution satellite data (Fig. 1 ). Landscape metrics were extracted to quantify vegetation composition, spatial heterogeneity, patch configuration, and three-dimensional urban structure, encompassing both green and built components. This multiscale integration allowed us to test whether beauty is primarily driven by vegetation amount or by the interplay between green and gray elements—how buildings, trees, and open spaces coexist within the urban mosaic. Finally, using all the responses of the survey, we trained a deep learning model to generalize these relationships, predicting aesthetic scores for over 350,000 street-level images across the entire city. The resulting map, at 50 m resolution, provides the first comprehensive visualization of São Paulo’s aesthetic ecosystem service distribution. Together, these analyses reveal how urban form, ecological structure, and spatial configuration jointly influence people’s aesthetic experiences, offering a framework to link perception, landscape ecology, and urban sustainability in an empirically grounded way. RESULTS We tested the effect of personal characteristics of the respondents as potential effects on the selection of the images in the survey, using generalized linear mixed models to compare different socioeconomic aspects between them. We did not find any significant differences between the individual characteristics (such as age, gender, culture and social level) and their aesthetic preference (examples provided in Fig. S1 – S11), we thus computed the Elo scores by pooling across the 3,221 respondents. Using the Green View Index (Li et al., 2015 ) (GVI), a method that measures the index of green in the streets trough SVI, we tested if our aesthetic values (Elo scores) are related to the index of green trough Person’s correlation, showing there are is a very small relationship between them, but mostly a spatial mismatch between aesthetic and GVI, with a R² of just 0.153 between the mean values of the variables in a 50 x50 m grid for the entire city (Fig. 2 ). To understand the determinants of aesthetic value, we extracted for each 420 SVI features (1) from the landscape - above and (2) from the image - front (see Fig. 1 and Methods). Using a dredge procedure with Akaike Information Criteria (AIC) model selection (see dredge results at Fig. S12 and Fig. S13), we obtained two models, one with the features from above, one with features from the front. Combining both models, we obtained a final model of the impact of the images and landscape features on aesthetics. The best model from the landscape features - above - presents six of 17 variables tested, with tree volume having the most important (positive) on aesthetic value (Fig. 2 A). The proportion of buildings coverage, the number of building patches and the building's edge density were following, all with negative effects (Fig. 3 A). Analyzing the best model from image features (from the front) of 12 variables used, five were selected (Fig. 3 B), with image shape index (here called image complexity) having the positive strongest effect, followed by fraction of blue in the image and color heterogeneity in the image. The image brightness and the PCA3, that is a mixture between gray and blue in the image (see supplementary Fig. S14 - S16), were following with both negative effects (Fig. 3 B) Pulling both models together, we obtained the final model comparing the relative coefficient estimate of each variable, where the most significant variable acting positively on aesthetic scores were the image complexity and the tree volume (Fig. 3 C). The building coverage, image brightness, followed by buildings edge density were acting negatively (Fig. 3 C). The analysis by Local Climate Zone (LCZ) also revealed significant differences between LCZ zones and the aesthetic values (Fig. 4 ). Zones with open arrangement of mid-rise or low-rise buildings (e.g. LCZ5 and LCZ6), which provide space between buildings for vegetation and tree cover, had the highest aesthetic values. The lowest street aesthetic values were given to LCZ’s dominated by buildings: the LCZ8 (low-big) and the LCZ3 (low compact). Using convolutional neural networks (CCN’s) we accurately predicted through deep learning the aesthetic values of images used in our survey. The model showed high R² values on all three split sets (training = 0.92; validation = 0.83; test = 0.91; full results and values are available on supplementary Fig. S22-S23). With this model in hand, we predicted the aesthetic value for 355,168 SVI of the entire city of São Paulo, representing a spatial resolution of 50 m, that is one SVI each 50m for all the streets mapped in the city (Fig. 5 A). This map presents, in an unprecedented manner, the provision of scenic beauty ecosystem services at the scale of a megacity. By juxtaposing this with population density (Fig. 5 B), often considered a proxy for service demand (Baró et al., 2017 ), we can discern various combinations of supply and demand within the city of São Paulo, particularly highlighting areas of mismatch into the city (Fig. 5 C). Most of the areas of high aesthetic scores have low population density, corresponding to the city's most affluent and upscale neighborhoods (Fig. 5 D), but some central areas, with taller buildings and the occurrence of green areas between buildings or on the streets, present high demand and high supply (Fig. 5 E). Several areas of the city have high demand but low local supply of the ecosystem service (brightest fuchsia in Fig. 5 C), which generally corresponds to the city's peripheral and less valued neighborhoods (Fig. 5 F). DISCUSSION Our findings contribute to interdisciplinary understandings of urban aesthetics by demonstrating that perceived beauty extends beyond mere green quantity to encompass the spatial complexity and configuration of vegetation and built-up areas in the streetscape ecosystem. These results align with neurocognitive theories that suggest humans prefer heterogeneous environments due to evolutionary traits (Kaplan, 1987a ; Richardson et al., 2017 ). By linking visual aesthetic preferences to established landscape ecology metrics, this study underscores the need for multidimensional approaches in assessing cultural ecosystem services in highly anthropic environments that are part of daily landscapes . Despite green coverage not being selected in our models, the volume of vegetation emerged as one crucial factor in shaping street aesthetics. Conversely, complexity observed from a frontal perspective – street level view - played a significant role in explaining aesthetic preferences, reflecting the intricate arrangement of vegetation and built-up areas in triggering aesthetic experience within urban landscapes. Images showcasing diverse shapes consistently received higher aesthetic ratings from respondents (Fig. 6 ). Upon closer examination of the importance of vegetation, it becomes evident that the effect of greenery involves a crucial three-dimensional aspect. It is not solely the horizontal spread of tree coverage that influences perceived beauty; rather, it is the combined horizontal and vertical extent of greenery. Higher densities of trees have been particularly influential in shaping aesthetic perception. This means that in megacities like São Paulo, street trees formations with larger volumes are highly valued. Consequently, older and taller trees, characterized by greater height or canopy area, and thus, larger volume, are particularly prized (Suchocka et al., 2022 ). This underscores the importance of preserving older and bigger trees to maintain aesthetic appeal (Blicharska & Mikusiński, 2014 ; Suchocka et al., 2022 ). From the perspective of ecosystem services, this finding carries significant implications for street tree management. It highlights that the services provided by large, old trees cannot simply be replaced by planting young trees. Therefore, it underscores the necessity of conserving and nurturing healthy mature trees in urban environments. This is in line with other ecosystem services provided by urban trees in cities, such as microclimate regulation(Kong et al., 2017 ); pollutant control (Grote et al., 2016 ); noise alleviation (Xu et al., 2022 ) among several others (Salmond et al., 2016 ). Moreover, the comparative analysis between the scenic beauty of streets and vegetation cover indicates that approximately 84% of the variance in scenic service is not explained by the Green View Index, highlighting that urban features beyond seen greenery from street level play a significant role in aesthetic appreciation. The shape complexity of the images was the most important variable to understanding the aesthetic scores, affecting them positively. This variable consists in the shape index of the image seen from the front and is linked to the format and shape that the vegetation appears on the streets, with several leaves, trunks and branches forming intricate and complex figures. This landscape complexity could be more attractive for people because of neurobiological preferences to some visual features (Pearce et al., 2016 ). According to neurocognitive theories people like heterogeneous environments, and the preference for complexity could be an evolutionary characteristic to select places with availability of resources (Kaplan, 1987b ). Image complexity can also be associated with the heterogeneity or fragmentation of the urban landscape. Specifically, aerial factors like the presence of built-up edges and the number of buildings patches measured from above were found to negatively impact landscape beauty. This means that areas with lower number of buildings, with space and green cover in between them could be considered more visually pleasant (Fig. 6 A, B) than those areas with dense construction attached to each other (Fig. 6 C, D). While previous studies have recognized the importance of buildings in shaping the streetscape of cities (Chen et al., 2016a ; Weber et al., 2008 ), it is becoming evident that the quantity and arrangement of buildings also play critical roles. The number of buildings and their placement in relation to green areas within the landscape significantly impact aesthetic evaluations. Surprisingly, socioeconomic profiles appear to have no discernible impact on the perceived aesthetic quality of streets. Some research has shown that the general “aesthetic judgment” can vary according to age, gender, culture and social characteristics (Kalivoda et al., 2014 ; Pugach et al., 2017 ; Zhan et al., 2021 ) although our investigation into the association between conflicting profiles revealed that socioeconomic status does not influence preferences for street images (Figs. S1 – S11). This finding aligns with some existing literature on landscape aesthetics, which suggests that when evaluating the beauty of the environment, consensus often emerges across diverse social groups (Hagerhall, 2001 ) or different regions (Chen et al., 2016b). In our study, we observed a similar pattern, indicating that regardless of origin, ethnicity, education level, or gender, individuals tend to converge on shared notions of beauty in urban landscapes, in other words the elements contribute for streetscapes aesthetic CES supply are consensual among dwellers with diverse backgrounds. The uneven distribution of aesthetic service throughout the city highlights social disparities, with areas of high demand and low supply consistently located in less privileged neighborhoods (Fig. 5 ). When we extrapolated the aesthetic scores to encompass the entire city, wealthier areas within the city consistently exhibited elevated aesthetic values. The zones with the higher aesthetic values are neighborhoods with buildings of middle-rise, sparse on space, with other uses, as green areas, in between the buildings (LCZ 5 and 6). These zones, with high aesthetic service supply and low demand, represent wealthier areas of the city, with low human density, big private houses and gardens and a low number of building patches, similarly to a land-sharing arrangement of the city (Cirino et al., 2022 ). Conversely, large regions of the city characterized by higher population densities and lower socioeconomic indicators registered lower aesthetic scores. This pattern is particularly pronounced in densely populated areas, such as urban slums and unplanned occupations in peripheral regions (LCZ 2 and 3) where the high density and high number of buildings and the lack of vegetation result in lower aesthetic ratings. This disparity underscores a significant mismatch between the supply and demand of ecosystem services in these areas (Dobbs et al., 2018 , 2019 ), thereby amplifying indications of a “luxury effect” in the provision of such services (Hope et al., 2003 ; Leong et al., 2018 ). These disparities not only highlight socio-economic inequities but also underscore the need for targeted interventions to bridge the gap in access to aesthetically pleasing urban environments. From a broader perspective, these results illustrate how urban aesthetic experiences operate at the interface between CES and NCP, reflecting both material and relational values (Chan et al., 2016 ; Díaz et al., 2018 ). While the cultural service framing traditionally emphasizes the measurable provision of benefits such as scenic enjoyment, the NCP framework acknowledges the relational dimension — the ways people establish meaningful connections with their surroundings through daily encounters with urban nature. In this sense, the observed inequalities in the spatial distribution of aesthetic opportunities are not only ecological mismatches but also relational injustices, as they limit the capacity of certain groups to build emotional, cultural, and identity-based ties with nature in their everyday environments – despite they agree about which streetscapes are more beautiful. By integrating perception-based mapping with landscape structure, our study bridges these perspectives, demonstrating that the beauty of urban streets is a tangible manifestation of how ecosystems contribute to human well-being, both through biophysical characteristics and through the relational experiences they enable. The map showing the balance of supply and demand can guide urban planning in different ways. In particular, areas characterized by high demand and low to medium supply are prime candidates for urban interventions, as they can positively impact a larger number of city inhabitants and enhance the provision of ecosystem services increasing the contribution of nature-based solutions. Alternatively, considering that urban densification offers significant advantages, including more efficient utilization of urban infrastructure (Bergesen et al., 2017 ; Pelczynski & Tomkowicz, 2019 ), it is interesting to consider as reference areas of the city that exhibit both high supply and high demand for ecosystem services (Fig. 5 E). These areas typically belong to the middle and high-middle class segments, located in verticalized sections of the city, predominantly represented by LCZ 4 (high-compact) but also LCZ 1 (high-open). Despite their high population density, these neighborhoods are often well-planned, situated in the central parts of the city, and adorned with trees—particularly those with substantial volume—lining the streets and interspersed among high-rise buildings and condominiums. This scenario illustrates the possibility of reconciling the supply and demand of aesthetic ecosystem services within the city, facilitated by ample sidewalk space and effective urban planning. This scenario exemplifies cities that have successfully integrated dense populations with ample green spaces, mirroring the model seen in Singapore and Hong Kong, among others (Wu et al., 2019 ; Xue et al., 2017 ). More generally our approach holds potential not only for São Paulo but also for cities worldwide, particularly megacities grappling with complex urban challenges. The consistent correlation between scenic beauty evaluations and the LCZ indicates that these zones can serve as a basis for extrapolating the results or gaining a deeper understanding of the urban conditions influencing the appreciation of scenic beauty on urban streets. Moreover, São Paulo, being a pronounced diverse city, with areas representing different urban settings, architecture diversification and social realities, fitted as a great trial for assessing perception and landscape patterns on the supply of aesthetic CES. By harnessing our methodology, which integrates cutting-edge technology with landscape analysis, cities globally can gain unprecedented insights into the factors influencing the aesthetic quality of their urban streets. This presents a groundbreaking opportunity for urban planners and policymakers to make informed decisions aimed at enhancing the beauty and equity of their cities. By enhancing the aesthetic appeal of streets, a range of co-benefits is likely to emerge, including increased walkability (Lwin & Murayama, 2011 ), improved health and social cohesion (De Vries et al., 2013 ), heightened social interaction, a stronger sense of place (Semenza et al., 2007 ) and safety (De Nadai et al., 2016 ), as well as greater use of public spaces for physical and recreational activities, all of which contribute to the overall well-being of residents (Bratman et al., 2015 , 2019 ; Dai et al., 2021 ; De Vries et al., 2013 ). Our findings highlight the universality of certain principles governing urban aesthetics, suggesting that lessons learned in São Paulo can be extrapolated to other global cities. By embracing this approach, cities across the world can leverage data-driven strategies to address urban social justice, promote public health, and create more livable and sustainable urban environments. Ultimately, our study reveals a pioneering tool for urban planning, with the potential to transform the way cities envision and design their streetscapes with more equity considering the relational dimension between people and their daily environment, treating the streets as part of a complex urban ecosystem that can avoid the extinction of human-nature experience. METHODS Study area and landscape/images sampling São Paulo City is the biggest megacity of Brazil and the biggest in south hemisphere, with more than 12 million people in the municipality and more than 21 million in the metropolitan region. São Paulo presents high social inequality and diverse morphology in terms of building densification and urban forestry, but also socio-cultural identity. The urban area of the city is more than 914.5 km², and 48.18% of the municipality area is covered by vegetation (São Paulo, 2020 ). Most of the vegetation is represented by secondary and semi-natural Atlantic rainforest, concentrated in two big state parks, one in the north and another in the south of the city (Fig. 1 A), making most of this vegetation isolated from the local population. A stratified sampling approach was employed to ensure representation across São Paulo's seven urban morphologies, categorized by Local Climate Zones (LCZ) (Stewart and Oke, 2012; Ferreira et al., 2017 ). This method enables robust extrapolation of aesthetic perceptions to diverse urban contexts while maintaining ecological relevance. Each sampled Street View Image (SVI) represents typical streetscapes for its LCZ, providing a systematic basis for integrating landscape-level and street-level features (Ferreira et al., 2017 ). We randomly choose 60 street points in each LCZ (Fig. 1 A) and collected for each one the image on Google Street View with an angle of view of 180° in relation to the street line, to ensure that the images face from the front of each street. For each collected image, by drawing polygons with trapeze shape, we estimated the landscape correspondent seen from above (Figs. 1 B; 1 C). Each polygon has variable sizes, depending on the reach of view of each image seen from the front. For each landscape we estimated the composition of green and buildings (proportion on the landscape), based on the map of vegetation cover of São Paulo municipality (São Paulo, 2020 ). The data consist of a vegetation classification in the resolution of 1:1000 inside the city, and 1:5000 within protected areas, based on orthophotos with 0.12 m of resolution, from 2017, obtained by the municipal environment office. Additionally, we used LiDAR data on the resolution of 10 points per m² and precision of 0.1m to estimate the area occupied by buildings (Gomes, 2022 ). Using the LiDAR data, we estimated the height and volume density of trees and buildings. Additionally, we also evaluated the configuration of the urban landscape (buildings and different types of vegetation), through: (1) large patch index; (2) edge density; and (3) number of patches (McGarigal, 2001 ), using the R package landscapemetrics v2.1.4. In that way we obtained three subsets of landscape measurements, (1) the composition, (2) the configuration and (3) the three-dimensional morphology (Table 1 ). For each image collected we extracted image features that may influence human aesthetic perception. We used 500x500 pixel images at 96 dpi. (1) Color heterogeneity was measured following the procedure used in Langlois et al. (Langlois et al., 2022 ): the K-means clustering algorithm was used to separate each pixel of an image in the CIELAB color space (we used 9 clusters). The mean distance between the 9 cluster centers was used as a measure of color heterogeneity. (2) Color saturation and (3) Brightness of each image were measured using the HSV (Hue, Saturation, Value) color space that differentiates saturation (S) from perceptual lightness (V). Images (4) Contrast and (5) Fractal dimension (self_similarity) were measured using the function img_contrast() and img_self_similarity() of the R package imagefluency v.0.2.5 . The proportion of (6) Green , (7) Blue and (8) Gray in each image was computed with the function countColors() of the R package countcolors v.0.9.1 . Finally the R package colordistance v.1.1.2 was used to provide a more integrated measure of color distance between images. The functions getLabHistList() and getColorDistanceMatrix() were used to obtain color distance matrix between all images which was then reduced using a PCA analysis (function dudi.pca () of the package ade4 v.1.7–22 ). The three first axes of the PCA accounted for most of the variance (respectively 51.3%, 29,7% and 8.6%) and could be used to classify the images along meaningful axes of dominant color and/or color mixtures (Fig. S14 – Fig. S16). Online survey and Elo’s scores We conducted an online survey, available to the general public between May 01st and September 30th, 2023, presenting to each respondent 30 pairs of images randomly selected from the pool of 420 SVI. For each pair, respondents were asked to click on the image they thought show the most beautiful street. Each respondent then was redirected to a socioeconomic and profile survey to collect information on sociocultural backgrounds (gender; ethnicity; age; scholarity; social class; type/size of the city where they live and grew up; Fig. S18). All the responses were anonymous to protect the identity of the respondents. The online survey was available in Portuguese on a dedicated website ( https://www.biodiful.org/ ) and was answered by 3,221 respondents. The survey was advertised by e-mails list for universities and research centers and asked to be publicly disclosed. A publication was also advertised on Instagram ; the target audience was any Brazilian adult over 18 years old. Note that participants were asked if they had color perception deficiency, and, if yes, that we removed their answer from the final dataset analyzed. To account for sociocultural background on the statistical analysis, we employed a generalized linear mixed model (GLMM) with a binomial error structure. This analysis was conducted using the glmer() function from the R package lme4 v1.1–26 . In this model, the image was treated as a random effect variable, allowing us to assess the individual impact of sociocultural variables on the response variable and effectively order them based on their influence. This analysis showed no effect of any of the sociocultural variables which allowed us to pool all the respondent answers to compute a global Elo scores analysis. The Elo scoring method(Elo, 2008 ) was employed to compare images as paired competitors, ultimately assigning final images scores based on the cumulative comparisons within our dataset. We used the R package EloChoice v0.29.4 (Clark et al., 2018 ) with 1,000 bootstrapings. This methodological approach enabled a nuanced examination of perceived image beauty, offering insights into the aesthetic inclinations within our sampled image population. Image Elo scores will be referred to as “aesthetic value” hereafter. Statistical analysis – understanding front and above features Aiming to understand the factors associated to the preference of people by the images, we divided the statistical analysis into two components: one of predictive variables based on the landscape – from above ; and another of predictive variables based on the image’s features – from the front (street-level view). For each set of variables, we constructed a full model with all variables in the same Linear Model (LM), the predictive variables are described in Table 1 , they represent all the variables tested in the study. Some of them were log-transformed to adjust to the linear model. The response variable was the aesthetic values of each of the 420 SVI used in our online survey. Table 1 Variables used in the full model, divided according to the group : from above (landscape) and from the front (SVI’s features). All the following variables were tested. The ones signed with * where log transformed. Variable Group/Type Composition Proportion of buildings coverage above Proportion of tree coverage* above Proportion of open vegetation coverage* above Proportion of total vegetation coverage* above Configuration Large patch index of total vegetation* above Edge density of total vegetation above Number of patches of total vegetation* above Large patch index of buildings above Edge density of buildings above Number of patches of buildings above Large patch index of trees* above Edge density of trees above Number of patches of trees* above Height & Volume Mean height of trees above Volume of trees* above Mean height of buildings* above Volume of buildings* above Images features Color heterogeneity front Complexity/shape index front Saturation front Brightness front PCA 1 - Gray and darkness/shadows mixture front PCA 2 – Green and shadows mixture front PCA 3 - Gray and blue mixture front Contrast front Fractal/Self similarity front Green proportion front Gray proportion front Blue proportion front After generating the full models, we run a dredge process (R, package MuMln v1.47.5 ), that consists in generating all the possible combinations between one and the total number of variables adjusted to the response variable – aesthetics values. After generating all models, the dredge ranks the best models by Akaike Information Criteria (AIC), searching for the model that best adjusts to our response. We ran two dredge processes, one for the variables from the front and another to above. For each dredge process we kept just the variables present in the best model, with the lowest AIC value. On sequence we tested for Variance Inflation factor using the vif() function from the R package car v4.3.3 , to detect multicollinearity in the linear regression. Multicollinearity occurs when independent variables in a regression model are highly correlated with each other, if multicollinearity is detected (VIF values are higher than 4) we investigated the Pearson correlation of the variables in order to exclude the highly correlated ones (more than 0.6 of correlation). Given that all SVI have an explicitly spatial component, we tested the best models of each set of variables for Spatial Autocorrelation. We used Moran’s I test to check autocorrelation of the residuals given the coordinates of each sample (R package DHARMa v0.4.6 ). The test pointed to spatial autocorrelation for both models – see supplementary text – for that reason we opt to use a modeling process that considers the space as a mixed part of the model. We ran a model with the variables selected by AIC using the FitMe() function (Fitting function for fixed and mixed-effect models with GLM response of the R package spaMM) . This function allows fitting a Generalized Linear Model (GLM) with a mixed part of the model with non-gaussian random effects, that is the case of the spatial information – latitude and longitude of the samples. With that, we could consider as part of the model the spatial distribution of our samples, avoiding model inflation by spatial autocorrelation. In addition to space, we added as a random variable in the model the Local Climate Zone (LCZ) where each of the 420 images were collected, in order to control the random effects of the morphology of the city in the statistical analysis. We analyzed the fixed part of each model, given by the coefficient estimates of each variable in relation to the response variable (Fig. 3 ; Table S1 to Table S3). With this process we obtained three final models controlling the space, one for the variables from the front, another for the variables from above and finally, for the third model, we gathered the variables of both previous models in a new FitMe model, combining the effect of both the front and above variables to see the relative estimate of each variable in relation to each other. Moreover, we ran an aditional model considering as predictive variables the LCZ of the city, that are a simplified way to categorize the morphology of the city, and are common for several cities across the world. We ran an ANOVA analysis with Tukey test to compare pair to pair of each LCZ, searching for significant difference (p – value < 0.005) between LCZ categories (results are presented in Table S4). Prediction of aesthetic values with a CNN model The 420 images evaluated in the online survey and their aesthetic values were used as a training dataset for a deep learning algorithm. We used a Convolutional Neural Network (CNN) deep learning algorithm that involved fine-tuning a CNN pre-trained on ImageNet(Deng et al., 2010 ) to facilitate transfer learning. Our dataset was partitioned into training (314 images), validation (53 images), and testing (53 images) sets. Data augmentation, fine-tuning of the models, and prediction of the aesthetic scores were carried out using Python 3.7, Pytorch 1.4.0 , and torchvision 0.5.0 . We used the ResNet50 architecture initialized with weights pre-trained on the ImageNet dataset, available via the torchvision.models module from PyTorch , and fine-tuned it using our training set. The model was trained with a learning rate of 1e-2, a batch size of 16, and for up to 350 epochs, with additional regularization techniques such as dropout and weight decay to mitigate overfitting. The final performance of the model was estimated using several evaluation metrics, including the mean squared error (MSE), mean absolute error (MAE), median absolute error (MedAE), mean absolute percentage error (MAPE), mean error (ME), and the coefficient of determination (R²). These metrics compared the values predicted by the model to the values in the testing set, which was not used during training. The results revealed a remarkably accurate prediction of aesthetic values (e.g., R² = 0.91 on the testing set), demonstrating the model’s ability to estimate human-perceived aesthetic value of the images included in our survey (see supplementary Fig. S22-S23 for a complete visualization of the metrics used). The trained CNN model was then used to predict the aesthetic values for images downloaded on google street view for the whole city of São Paulo. We used the shapefile of lines containing all the streets, roads and avenues of the city, provided by the city-hall, and created a point each 50m in all the public streets of the city, totalizing 382,760 points, from which we extracted the coordinates. The 382,760 locations were used to download corresponding street view images with the function google_streetview() of the R package googleway v2.7.8 using the same angle view – 180° in relation to the street - that the 420 images collected for our online survey. Filtering Street View Images out of the range The download of the SVI resulted in a set of 355,801 images (the function returned no image for 27,013 locations). We used a trained image filter to exclude the sample images that correspond to wrong angles of view – facing isolated objects or buildings facades – and images that deviate from the expected pattern – as blurred images or images taken from inside of tunnels. We ended up with a dataset of 355,168 images for which we used our trained CNN model to predict the aesthetic values and produce the map of the aesthetic ecosystem service for the city. For these 355,168 images, we also extracted the same visual features used in the original 420 images from the perception survey: color heterogeneity, color saturation, brightness, perceptual lightness, contrast, fractal dimension (self-similarity), and the proportions of green, blue, and gray. We conducted a Principal Component Analysis (PCA) using the original 420 images and projected the 355,168 images into this PCA space (Fig. S19). Analyzing the first two principal components, we found that 94.2% of the projected images fell within the convex hull area defined by the original 420 training images. To further refine our prediction set and ensure representativeness, we retained only those images falling within the hull area plus a 10% margin. This resulted in a final dataset comprising 96.9% of the full image set, demonstrating that the original survey-based sample is highly representative of the broader image dataset across the city. These steps confirm the robustness and reliability of our deep learning model for extrapolating aesthetic values in São Paulo’s urban streetscapes (Fig. S20). Declarations Acknowledgments We also thank Artur Lupinetti-Cunha for the data on population and Gabriel Garcia for the design of some figures. We thank professors Denise Duarte, Renata Pardini and Vitor Vasconselos for their comments in early draft of this manuscript. We thank L. Roman Carrasco for his valuable conceptual insights and discussions during DWC’s research stay at the National University of Singapore, which helped shape the theoretical framing of this study Finally, we especially thank all the 3,221 respondents of the survey. Ethics statement and consent All methods were carried out in accordance with the relevant institutional guidelines and regulations. The study protocol, including the online survey in which participants rated the aesthetic quality of urban street images, was reviewed and approved by the Ethics Committee for Research with Human Subjects of the Institute of Biosciences, University of São Paulo (Comitê de Ética em Pesquisa com Seres Humanos do Instituto de Biociências da Universidade de São Paulo; approval no. 57.763.208). All participants were adults (≥18 years old) and provided informed consent prior to participating in the survey. No personally identifiable information was collected. Funding: São Paulo Research Foundation - FAPESP n. process: 2020/15785-7 (DWC) São Paulo Research Foundation - FAPESP n. process: 2020/06694-8 (JPM) Author contributions: Conceptualization: DWC, NM, JPM Methodology: DWC, NM, JPM Investigation: DWC, NM Visualization: DWC, NM Supervision: JPM Writing—original draft: DWC Writing—review & editing: DWC, NM, JPM Competing interests: Authors declare that they have no competing interests. Data and materials availability All data, code, and materials used in the analyses is available at: https://github.com/DougCirino/AESTHETIC_STREETS DOI: 10.5281/zenodo.13318006 References Baró, F., Gómez-Baggethun, E., & Haase, D. (2017). Ecosystem service bundles along the urban-rural gradient: Insights for landscape planning and management. Ecosystem Services , 24 , 147–159. https://doi.org/10.1016/J.ECOSER.2017.02.021 Beil, K., & Hanes, D. (2013). The influence of urban natural and built environments on physiological and psychological measures of stress- A pilot study. International Journal of Environmental Research and Public Health , 10 (4), 1250–1267. https://doi.org/10.3390/ijerph10041250 Bergesen, J. D., Suh, S., Baynes, T. M., & Musango, J. K. (2017). Environmental and natural resource implications of sustainable urban infrastructure systems. Environmental Research Letters , 12 (12), 125009. https://doi.org/10.1088/1748-9326/AA98CA Blicharska, M., & Mikusiński, G. (2014). Incorporating Social and Cultural Significance of Large Old Trees in Conservation Policy. Conservation Biology , 28 (6), 1558–1567. https://doi.org/10.1111/COBI.12341 Bratman, G. N., Anderson, C. B., Berman, M. G., Cochran, B., de Vries, S., Flanders, J., Folke, C., Frumkin, H., Gross, J. J., Hartig, T., Kahn, P. H., Kuo, M., Lawler, J. J., Levin, P. S., Lindahl, T., Meyer-Lindenberg, A., Mitchell, R., Ouyang, Z., Roe, J., … Daily, G. C. (2019). Nature and mental health: An ecosystem service perspective. Science Advances , 5 (7), 903–927. https://doi.org/10.1126/SCIADV.AAX0903/SUPPL_FILE/AAX0903_SM.PDF Bratman, G. N., Hamilton, J. P., Hahn, K. S., Daily, G. C., & Gross, J. J. (2015). Nature experience reduces rumination and subgenual prefrontal cortex activation. Proceedings of the National Academy of Sciences of the United States of America , 112 (28), 8567–8572. https://doi.org/10.1073/pnas.1510459112 Cabana, D., Ryfield, F., Crowe, T. P., & Brannigan, J. (2020). Evaluating and communicating cultural ecosystem services. Ecosystem Services , 42 , 101085. https://doi.org/10.1016/j.ecoser.2020.101085 Chan, K. M. A., Balvanera, P., Benessaiah, K., Chapman, M., Díaz, S., Gómez-Baggethun, E., Gould, R., Hannahs, N., Jax, K., Klain, S., Luck, G. W., Martín-López, B., Muraca, B., Norton, B., Ott, K., Pascual, U., Satterfield, T., Tadaki, M., Taggart, J., & Turner, N. (2016). Why protect nature? Rethinking values and the environment. Proceedings of the National Academy of Sciences of the United States of America , 113 (6), 1462–1465. https://doi.org/10.1073/PNAS.1525002113/ASSET/6CB06F81-F2E6-43FD-90F6-81CFCEB8FBBE/ASSETS/GRAPHIC/PNAS.1525002113FIG02.JPEG Chen, Z., Xu, B., & Devereux, B. (2016a). Assessing public aesthetic preferences towards some urban landscape patterns: the case study of two different geographic groups. Environmental Monitoring and Assessment , 188 (1), 1–17. https://doi.org/10.1007/S10661-015-5007-3/METRICS Cirino, D. W., Tambosi, L. R., Mauad, T., de Freitas, S. R., & Metzger, J. P. (2022). Balanced spatial distribution of green areas creates healthier urban landscapes. Journal of Applied Ecology , 59 (7), 1884–1896. https://doi.org/10.1111/1365-2664.14195 Clark, A. P., Howard, K. L., Woods, A. T., Penton-Voak, I. S., & Neumann, C. (2018). Why rate when you could compare? Using the “EloChoice” package to assess pairwise comparisons of perceived physical strength. PLOS ONE , 13 (1), e0190393. https://doi.org/10.1371/JOURNAL.PONE.0190393 Dai, L., Zheng, C., Dong, Z., Yao, Y., Wang, R., Zhang, X., Ren, S., Zhang, J., Song, X., & Guan, Q. (2021). Analyzing the correlation between visual space and residents’ psychology in Wuhan, China using street-view images and deep-learning technique. City and Environment Interactions , 11 , 100069. https://doi.org/10.1016/J.CACINT.2021.100069 Daniel, T. C., Muhar, A., Arnberger, A., Aznar, O., Boyd, J. W., Chan, K. M. A., Costanza, R., Elmqvist, T., Flint, C. G., Gobster, P. H., Grêt-Regamey, A., Lave, R., Muhar, S., Penker, M., Ribe, R. G., Schauppenlehner, T., Sikor, T., Soloviy, I., Spierenburg, M., … Von Der Dunk, A. (2012). Contributions of cultural services to the ecosystem services agenda. Proceedings of the National Academy of Sciences of the United States of America , 109 (23), 8812–8819. https://doi.org/10.1073/PNAS.1114773109/ASSET/BB9209BA-B18B-4515-B848-50D3A3011DBB/ASSETS/GRAPHIC/PNAS.1114773109FIG01.JPEG De Nadai, M., Vieriu, R. L., Zen, G., Dragicevic, S., Naik, N., Caraviello, M., Hidalgo, C. A., Sebe, N., & Lepri, B. (2016). Are safer looking neighborhoods more lively? A multimodal investigation into urban life. MM 2016 - Proceedings of the 2016 ACM Multimedia Conference , 1127–1135. https://doi.org/10.1145/2964284.2964312 De Vries, S., van Dillen, S. M. E., Groenewegen, P. P., & Spreeuwenberg, P. (2013). Streetscape greenery and health: Stress, social cohesion and physical activity as mediators. Social Science & Medicine , 94 , 26–33. https://doi.org/10.1016/J.SOCSCIMED.2013.06.030 Deng, J., Dong, W., Socher, R., Li, L.-J., Kai Li, & Li Fei-Fei. (2010). ImageNet: A large-scale hierarchical image database . 248–255. https://doi.org/10.1109/CVPR.2009.5206848 Díaz, S., Pascual, U., Stenseke, M., Martín-López, B., Watson, R. T., Molnár, Z., Hill, R., Chan, K. M. A., Baste, I. A., Brauman, K. A., Polasky, S., Church, A., Lonsdale, M., Larigauderie, A., Leadley, P. W., Van Oudenhoven, A. P. E., Van Der Plaat, F., Schröter, M., Lavorel, S., … Shirayama, Y. (2018). Assessing nature’s contributions to people: Recognizing culture, and diverse sources of knowledge, can improve assessments. Science , 359 (6373), 270–272. https://doi.org/10.1126/SCIENCE.AAP8826/SUPPL_FILE/AAP8826-DIAZ-SM.PDF Dobbs, C., Escobedo, F. J., Clerici, N., de la Barrera, F., Eleuterio, A. A., MacGregor-Fors, I., Reyes-Paecke, S., Vásquez, A., Camaño, J. D. Z., & Hernández, H. J. (2019). Urban ecosystem Services in Latin America: mismatch between global concepts and regional realities? Urban Ecosystems , 22 (1), 173–187. https://doi.org/10.1007/s11252-018-0805-3 Dobbs, C., Escobedo, F. J., & Zipperer, W. C. (2011). A framework for developing urban forest ecosystem services and goods indicators. Landscape and Urban Planning , 99 (3–4), 196–206. https://doi.org/10.1016/J.LANDURBPLAN.2010.11.004 Dobbs, C., Hernández-Moreno, Á., Reyes-Paecke, S., & Miranda, M. D. (2018). Exploring temporal dynamics of urban ecosystem services in Latin America: The case of Bogota (Colombia) and Santiago (Chile). Ecological Indicators , 85 , 1068–1080. https://doi.org/10.1016/j.ecolind.2017.11.062 Dubey, A., Naik, N., Parikh, D., Raskar, R., & Hidalgo, C. A. (2016). Deep learning the city: Quantifying urban perception at a global scale. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) , 9905 LNCS , 196–212. https://doi.org/10.1007/978-3-319-46448-0_12/FIGURES/4 Elo, A. E. (2008). The rating of chessplayers, past and present . 208. https://books.google.com/books/about/The_Rating_of_Chessplayers.html?hl=pt-BR&id=syjcPQAACAAJ Ferreira, L. S., Prata-Shimomura, A., Ferreira, A., Helena, D., & Duarte, S. (2017). Mapping Local Climate Zones for São Paulo Metropolitan Region: a comparison between the local climate zone map and two other local maps . https://www.researchgate.net/publication/318404791 Gomes, F. (2022). Building and Vegetation Height Model of São Paulo . https://doi.org/10.34740/KAGGLE/DSV/4044806 Grote, R., Samson, R., Alonso, R., Amorim, J. H., Cariñanos, P., Churkina, G., Fares, S., Thiec, D. Le, Niinemets, Ü., Mikkelsen, T. N., Paoletti, E., Tiwary, A., & Calfapietra, C. (2016). Functional traits of urban trees: air pollution mitigation potential. Frontiers in Ecology and the Environment , 14 (10), 543–550. https://doi.org/10.1002/FEE.1426 Hagerhall, C. M. (2001). CONSENSUS IN LANDSCAPE PREFERENCE JUDGEMENTS. Journal of Environmental Psychology , 21 (1), 83–92. https://doi.org/10.1006/JEVP.2000.0186 Hartig, T., & Kahn, P. H. (2016). Living in cities, naturally. Science , 352 (6288), 938–940. https://doi.org/10.1126/SCIENCE.AAF3759 Henderson, H., Child, S., Moore, S., Moore, J. B., & Kaczynski, A. T. (2016). The Influence of Neighborhood Aesthetics, Safety, and Social Cohesion on Perceived Stress in Disadvantaged Communities. American Journal of Community Psychology , 58 (1–2), 80–88. https://doi.org/10.1002/AJCP.12081 Hope, D., Gries, C., Zhu, W., Fagan, W. F., Redman, C. L., Grimm, N. B., Nelson, A. L., Martin, C., & Kinzig, A. (2003). Socioeconomics drive urban plant diversity. Proceedings of the National Academy of Sciences of the United States of America , 100 (15), 8788–8792. https://doi.org/10.1073/PNAS.1537557100/ASSET/9F5B8293-C983-4B01-B5B3-90DE21ECFEE5/ASSETS/GRAPHIC/PQ1537557002.JPEG Kalivoda, O., Vojar, J., Skřivanová, Z., & Zahradník, D. (2014). Consensus in landscape preference judgments: The effects of landscape visual aesthetic quality and respondents’ characteristics. Journal of Environmental Management , 137 , 36–44. https://doi.org/10.1016/J.JENVMAN.2014.02.009 Kaplan, S. (1987a). Aesthetics, Affect, and Cognition. Environment and Behavior , 19 (1), 3–32. https://doi.org/10.1177/0013916587191001 Kaplan, S. (1987b). Aesthetics, Affect, and Cognition. Http://Dx.Doi.Org/10.1177/0013916587191001 , 19 (1), 3–32. https://doi.org/10.1177/0013916587191001 Knez, I., Sang, A. O., Gunnarsson, B., & Hedblom, M. (2018). Wellbeing in urban greenery: The role of naturalness and place identity. Frontiers in Psychology , 9 (APR), 344587. https://doi.org/10.3389/FPSYG.2018.00491/BIBTEX Kong, L., Lau, K. K. L., Yuan, C., Chen, Y., Xu, Y., Ren, C., & Ng, E. (2017). Regulation of outdoor thermal comfort by trees in Hong Kong. Sustainable Cities and Society , 31 , 12–25. https://doi.org/10.1016/J.SCS.2017.01.018 Langlois, J., Guilhaumon, F., Baletaud, F., Casajus, N., De Almeida Braga, C., Fleuré, V., Kulbicki, M., Loiseau, N., Mouillot, D., Renoult, J. P., Stahl, A., Stuart Smith, R. D., Tribot, A. S., & Mouquet, N. (2022). The aesthetic value of reef fishes is globally mismatched to their conservation priorities. PLOS Biology , 20 (6), e3001640. https://doi.org/10.1371/JOURNAL.PBIO.3001640 Lenda, M. L., Skórka, P., Jaźwa, M., Lin, H.-Y., Nęcka, E., Tryjanowski, P., Moroń, D., Knops, J. M. H., & Possingham, H. P. (2023). Recognizing the importance of near-home contact with nature for mental well-being based on the COVID-19 lockdown experience. Ecology and Society , 28 (3). https://doi.org/10.5751/ES-14374-280313 Leong, M., Dunn, R. R., & Trautwein, M. D. (2018). Biodiversity and socioeconomics in the city: a review of the luxury effect. Biology Letters , 14 (5). https://doi.org/10.1098/RSBL.2018.0082 Li, X., Zhang, C., Li, W., Ricard, R., Meng, Q., & Zhang, W. (2015). Assessing street-level urban greenery using Google Street View and a modified green view index. Urban Forestry and Urban Greening , 14 (3), 675–685. https://doi.org/10.1016/j.ufug.2015.06.006 Lwin, K. K., & Murayama, Y. (2011). Modelling of urban green space walkability: Eco-friendly walk score calculator. Computers, Environment and Urban Systems , 35 (5), 408–420. https://doi.org/10.1016/J.COMPENVURBSYS.2011.05.002 McGarigal, K. (2001). Landscape Pattern Metrics. Encyclopedia of Environmetrics . https://doi.org/10.1002/9780470057339.VAL006 Pearce, M. T., Zaidel, D. W., Vartanian, O., Skov, M., Leder, H., Chatterjee, A., & Nadal, M. (2016). Neuroaesthetics: The Cognitive Neuroscience of Aesthetic Experience. Perspectives on Psychological Science , 11 (2), 265–279. https://doi.org/10.1177/1745691615621274/ASSET/IMAGES/LARGE/10.1177_1745691615621274-FIG1.JPEG Pelczynski, J., & Tomkowicz, B. (2019). Densification of cities as a method of sustainable development. IOP Conference Series: Earth and Environmental Science , 362 (1), 012106. https://doi.org/10.1088/1755-1315/362/1/012106 Pugach, C., Leder, H., & Graham, D. J. (2017). How stable are human aesthetic preferences across the lifespan? Frontiers in Human Neuroscience , 11 , 246626. https://doi.org/10.3389/FNHUM.2017.00289/BIBTEX Richardson, E. A., Pearce, J., Shortt, N. K., & Mitchell, R. (2017). The role of public and private natural space in children’s social, emotional and behavioural development in Scotland: A longitudinal study. Environmental Research , 158 , 729–736. https://doi.org/10.1016/j.envres.2017.07.038 Rossetti, T., Lobel, H., Rocco, V., & Hurtubia, R. (2019). Explaining subjective perceptions of public spaces as a function of the built environment: A massive data approach. Landscape and Urban Planning , 181 , 169–178. https://doi.org/10.1016/J.LANDURBPLAN.2018.09.020 Sabbion, P. (2018). Green Streets Social and Aesthetic Aspects. Nature Based Strategies for Urban and Building Sustainability , 283–290. https://doi.org/10.1016/B978-0-12-812150-4.00026-4 Salmond, J. A., Tadaki, M., Vardoulakis, S., Arbuthnott, K., Coutts, A., Demuzere, M., Dirks, K. N., Heaviside, C., Lim, S., MacIntyre, H., McInnes, R. N., & Wheeler, B. W. (2016). Health and climate related ecosystem services provided by street trees in the urban environment. Environmental Health : A Global Access Science Source , 15 Suppl 1 (Suppl 1). https://doi.org/10.1186/S12940-016-0103-6 Sandifer, P. A., Sutton-Grier, A. E., & Ward, B. P. (2015). Exploring connections among nature, biodiversity, ecosystem services, and human health and well-being: Opportunities to enhance health and biodiversity conservation. In Ecosystem Services (Vol. 12, pp. 1–15). Elsevier B.V. https://doi.org/10.1016/j.ecoser.2014.12.007 São Paulo, S. do V. e M. A. (SVMA). (2020). Mapeamento Digital da Cobertura Vegetal do Município de São Paulo. Relatório Final . https://www.prefeitura.sp.gov.br/cidade/secretarias/upload/meio_ambiente/RelCobVeg2020_vFINAL_compressed(1).pdf Semenza, J. C., March, T. L., & Bontempo, B. D. (2007). Community-initiated urban development: An ecological intervention. Journal of Urban Health , 84 (1), 8–20. https://doi.org/10.1007/s11524-006-9124-8 Soga, M., & Gaston, K. J. (2016). Extinction of experience: The loss of human-nature interactions. In Frontiers in Ecology and the Environment (Vol. 14, Issue 2, pp. 94–101). Ecological Society of America. https://doi.org/10.1002/fee.1225 Suchocka, M., Wojnowska-Heciak, M., Błaszczyk, M., Gawłowska, A., Ciemniewska, J., Jarska, A., Heciak, J., & Pachnowska, B. (2022). Old trees are perceived as a valuable element of the municipal forest landscape. PeerJ , 10 . https://doi.org/10.7717/PEERJ.12700/SUPP-6 Teixeira, F. Z., Bachi, L., Blanco, J., Zimmermann, I., Welle, I., & Carvalho-Ribeiro, S. M. (2019). Perceived ecosystem services (ES) and ecosystem disservices (EDS) from trees: insights from three case studies in Brazil and France. Landscape Ecology , 34 (7). https://doi.org/10.1007/s10980-019-00778-y Weber, R., Schnier, J., & Jacobsen, T. (2008). Aesthetics of Streetscapes: Influence of Fundamental Properties on Aesthetic Judgments of Urban Space. Perceptual and Motor Skills , 106 (1), 128–146. https://doi.org/10.2466/PMS.106.1.128-146 Wu, Z., Chen, R., Meadows, M. E., Sengupta, D., & Xu, D. (2019). Changing urban green spaces in Shanghai: trends, drivers and policy implications. Land Use Policy , 87 , 104080. https://doi.org/10.1016/J.LANDUSEPOL.2019.104080 Xu, C., Han, B., Lu, F., & Wu, T. (2022). Assessing the Traffic Noise Reduction Effect of Roadside Green Space Using LiDAR Point Cloud Data in Shenzhen, China. Forests 2022, Vol. 13, Page 765 , 13 (5), 765. https://doi.org/10.3390/F13050765 Xue, F., Gou, Z., & Lau, S. S. Y. (2017). Green open space in high-dense Asian cities: Site configurations, microclimates and users’ perceptions. Sustainable Cities and Society , 34 , 114–125. https://doi.org/10.1016/J.SCS.2017.06.014 Zhan, J., Liu, M., Garrod, O. G. B., Daube, C., Ince, R. A. A., Jack, R. E., & Schyns, P. G. (2021). Modeling individual preferences reveals that face beauty is not universally perceived across cultures. Current Biology , 31 (10), 2243-2252.e6. https://doi.org/10.1016/J.CUB.2021.03.013 Zhang, F., Zhou, B., Liu, L., Liu, Y., Fung, H. H., Lin, H., & Ratti, C. (2018). Measuring human perceptions of a large-scale urban region using machine learning. Landscape and Urban Planning , 180 , 148–160. https://doi.org/10.1016/J.LANDURBPLAN.2018.08.020 Additional Declarations No competing interests reported. Supplementary Files Cirinoetalsuppmaterial.docx Cite Share Download PDF Status: Under Review Version 3 posted Editorial decision: Revision requested 02 Mar, 2026 Reviews received at journal 19 Feb, 2026 Reviews received at journal 17 Feb, 2026 Reviewers agreed at journal 08 Feb, 2026 Reviewers agreed at journal 08 Feb, 2026 Reviews received at journal 21 Nov, 2025 Reviewers agreed at journal 12 Nov, 2025 Reviewers invited by journal 09 Nov, 2025 Editor assigned by journal 09 Nov, 2025 Editor invited by journal 28 Oct, 2025 Submission checks completed at journal 27 Oct, 2025 First submitted to journal 27 Oct, 2025 You are reading this latest preprint version Show more versions Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5109482","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[{"code":1,"date":"2024-12-03 02:36:51","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}},{"code":2,"date":"2025-04-23 06:04:14","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"articleType":"Article","associatedPublications":[],"authors":[{"id":542719257,"identity":"59b7047a-6208-4a03-9966-13137034cfb1","order_by":0,"name":"Douglas W. Cirino","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA40lEQVRIiWNgGAWjYDCCA0D8AMRgBrNtQEw2wloSEFrSSNECAYcJa+E73mP8IqGmjoG/nffh4YKK84nzZzewPa7Ao0XyzBkzi4RjhxkkDrMbHJ5x5nbihjsH2A3P4NFicCPHzCCB7QCDATMbw2HeNqAWiQQ2yQaCWv7VwbScS5w/g7AW4weJbcwwLQcSG24Q0CJ55lgZQ2LfYR6Jw0AtPGeSjTfcOdhuiE8L3/HmzR8+fKuT4+8/xvyZp8JOdv7s5mMP8WkBAjYJIMGD4EswEtAAjPYPqHwJQhpGwSgYBaNgpAEALs1PjXYiupwAAAAASUVORK5CYII=","orcid":"","institution":"Univ of São Paulo (USP) – Institute of Biosciences","correspondingAuthor":true,"prefix":"","firstName":"Douglas","middleName":"W.","lastName":"Cirino","suffix":""},{"id":542719258,"identity":"7508d5a2-9483-4711-8414-9344de9f82e1","order_by":1,"name":"Nicolas Mouquet","email":"","orcid":"","institution":"MARBEC, Univ Montpellier, CNRS, Ifremer, IRD","correspondingAuthor":false,"prefix":"","firstName":"Nicolas","middleName":"","lastName":"Mouquet","suffix":""},{"id":542719259,"identity":"e7ddf2ab-ec28-4891-90ba-cea77032474e","order_by":2,"name":"Jean Paul Metzger","email":"","orcid":"","institution":"Univ of São Paulo (USP) – Institute of Biosciences","correspondingAuthor":false,"prefix":"","firstName":"Jean","middleName":"Paul","lastName":"Metzger","suffix":""}],"badges":[],"createdAt":"2024-09-18 11:02:48","currentVersionCode":3,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-5109482/v3","doiUrl":"https://doi.org/10.21203/rs.3.rs-5109482/v3","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95670623,"identity":"d68cfc31-9b70-4334-884b-b6fa81dcb9b5","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11286966,"visible":true,"origin":"","legend":"","description":"","filename":"Cirinoetalmsversion2.docx","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/3702de8533b34e7bb9cbfff4.docx"},{"id":95799128,"identity":"ebc7dee3-7085-4947-a10e-ce8c4098ad6e","added_by":"auto","created_at":"2025-11-13 08:18:51","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6292,"visible":true,"origin":"","legend":"","description":"","filename":"80a1536eb61b43d4a330d5ec2abfd6f9.json","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/e64d7383b3001955299f3a4e.json"},{"id":95670628,"identity":"c73ffe7c-e0a8-4e61-a4eb-ce0640e70894","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19648492,"visible":true,"origin":"","legend":"","description":"","filename":"Cirinoetalsuppmaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/15bda21349965c42b8633df6.docx"},{"id":95797253,"identity":"4950b428-68ce-43e3-b5a6-605030c096d8","added_by":"auto","created_at":"2025-11-13 08:02:31","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":154598,"visible":true,"origin":"","legend":"","description":"","filename":"80a1536eb61b43d4a330d5ec2abfd6f91enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/0fa2aabacd6b7eba30e458c9.xml"},{"id":95798358,"identity":"a7665817-91c0-445f-a17a-7b4873dd385d","added_by":"auto","created_at":"2025-11-13 08:16:32","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2853302,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/fd062b08f2ff826bafbb0d87.png"},{"id":95670626,"identity":"add40d5c-794c-4df6-b166-a62fcba916c5","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":433299,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/34bb6e574f24b4b5f580e0e9.jpeg"},{"id":95797445,"identity":"43df8437-dba1-4151-8d0b-fd74adae993e","added_by":"auto","created_at":"2025-11-13 08:05:13","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":798386,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/b1b83018876e08a7f7a4bb78.png"},{"id":95670621,"identity":"1356bc8d-7272-4eaf-8911-84a41b6ba536","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":126052,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/daf214c991cba52f2bf3cdbd.png"},{"id":95798282,"identity":"74369b45-a800-471f-9829-4eddf5762110","added_by":"auto","created_at":"2025-11-13 08:16:20","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4869783,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/0bdaea56bb56ebd9c7c8f662.png"},{"id":95799700,"identity":"836f26ac-a514-4211-b21f-251a13958a0d","added_by":"auto","created_at":"2025-11-13 08:20:39","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1806119,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/bebd6532145147bf1a86f04b.png"},{"id":95799418,"identity":"2c79a9ef-574f-4fc7-85f8-dcc94594add7","added_by":"auto","created_at":"2025-11-13 08:19:54","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":549979,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/03f2a9923eada8d957c82e2b.png"},{"id":95799097,"identity":"2964bf1f-e9d8-4e25-a5a4-c888aca6bdd9","added_by":"auto","created_at":"2025-11-13 08:18:45","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":230313,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/d43ad7c71c937efee762d909.png"},{"id":95670634,"identity":"8c04b6b2-f127-44c1-8b4b-bbcc966eab28","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":138529,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/5cfddda2c543ec2750ada1a5.png"},{"id":95670632,"identity":"5c26c3e7-4df6-4844-bba5-f5d9a06058d2","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":23595,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/6b0deea48111fc8a6934d850.png"},{"id":95670633,"identity":"9ac24616-ed6d-4b3a-a2b3-c72df7421f88","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":996870,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/216dc329e5734479e9576367.png"},{"id":95670625,"identity":"64764236-51dd-4bff-8af4-45afe34dd5d4","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":345661,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/7f62af92c81306efd1844d49.png"},{"id":95670636,"identity":"acfb2393-c0b6-4e1e-9a40-4646a2f2508a","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"xml","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":151384,"visible":true,"origin":"","legend":"","description":"","filename":"80a1536eb61b43d4a330d5ec2abfd6f91structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/36bdc9c60087ea45a226ab6e.xml"},{"id":95670637,"identity":"cb008c20-6bd3-4d55-9081-8b8312961fe8","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"html","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":164880,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/93654ccaf3629b9c9ae862cb.html"},{"id":95670615,"identity":"d0d1fe88-97b9-4b09-bf76-66df0b5dc55e","added_by":"auto","created_at":"2025-11-11 17:21:38","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":426975,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eProcedure for sampling and collecting image data. \u003c/strong\u003e\u0026nbsp;(A) the city of São Paulo divided per Local Climate Zones (LCZ; a classification of the urban tissue according to the type of use and buildings) and the distribution of the 420 Google Street View Images (SVI) sampled; (B) an example of an SVI seen from the front; (C) an example of a landscape delimited as a polygon of an SVI – seen from above; and (D) an example of the layers utilized to extract the variables seen from above.\u003cstrong\u003e \u003c/strong\u003eImagery © 2024 Google.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/bf6533ecb564199011f99d23.png"},{"id":95670617,"identity":"3e2b57c8-1adc-4a45-8f75-3d8c024a3f4b","added_by":"auto","created_at":"2025-11-11 17:21:38","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":586451,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMismatch between aesthetic value and green view index (GVI).\u003c/strong\u003e (A) Pearson correlation between aesthetic score and Green View Index showing low (R² 0.153) correlation. (B) Example of area with high aesthetic and high GVI (C) Biplot image showing the combination of aesthetic service scores and Green View Index, and examples of contrasting situations: (D) low GVI and high aesthetic score; (E) high GVI and low aesthetic score. Imagery © 2024 Google.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/a064059e6878f7929a1eb7a6.png"},{"id":95797769,"identity":"854425d0-9c58-46fe-bf1c-205ffb75f621","added_by":"auto","created_at":"2025-11-13 08:10:25","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":325255,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCoefficient estimate of the models explaining street aesthetics values.\u003c/strong\u003e On the Y axis the variable in the X axis is the relative coefficient estimates of the variable. (A) Best model considering variables from landscape (above) as predictive variables; (B) Best model considering variables from image’s features (from the front) as predictive variables; (C) Best model combining A and B. The response variable is the aesthetic value of the 420 SVI. N.P. means number of patches; E.D. means edge density. PCA3 is the axis 3 of the Principal Component Analysis performed with the figure features. (D) Examples of different aesthetic values on sequence, from the higher (left) to the lower (right). Imagery © 2024 Google.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/e3509f3815677bef1afc4642.png"},{"id":95670618,"identity":"2a26e536-3bdc-4bc5-9df8-63b91ceed482","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":102822,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDifferences in the evaluated aesthetic scores according to urban morphology types. \u003c/strong\u003eDistribution of aesthetics scores according to different Local Climate Zones (LCZ). Dots represent the mean and whiskers the standard deviation. Schemes modified from (\u003cem\u003e19\u003c/em\u003e).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/a6f7b6f3cf4710c785a25f99.png"},{"id":95799110,"identity":"098e07d6-802e-4c72-9f3c-3a19c7676a7b","added_by":"auto","created_at":"2025-11-13 08:18:49","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":716415,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSupply-Demand maps and examples.\u003c/strong\u003e (A) the aesthetic valuesof the streets of the city of São Paulo (supply of the cultural service of “aesthetic view”), (B) the population density (a proxy of the people who demand this aesthetic service); (C) the combination (biplot image) of aesthetic service supply and demand. Illustrative Google Street view images are shown to exemplify situations of, (D) high supply and high demand (E) high supply and low demand; and (F) low supply and high demand. The map was made assuming a local flow for this type of service between supply and demand in a square grid of 100 m. The resolution used is a grid of 100x100m. Imagery © 2024 Google.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/1b6ca79b5301acb07f26ea35.png"},{"id":95804602,"identity":"8aeebc27-1fd0-4c7a-bded-c319a806a437","added_by":"auto","created_at":"2025-11-13 08:38:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3044435,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/f9d7324c-a6be-4fae-af35-85878759aafd.pdf"},{"id":95670638,"identity":"6b312cad-f39c-4f6a-881d-95d6a47b88b7","added_by":"auto","created_at":"2025-11-11 17:21:39","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":19648492,"visible":true,"origin":"","legend":"","description":"","filename":"Cirinoetalsuppmaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-5109482/v3/e252753dc90c3572c6bfee5b.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Bridging perception and landscape structure: mapping urban streetscape aesthetics as nature’s contributions to people","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe environment in which people reside and their perception of it are intimately intertwined with human well-being and health. This connection gives rise to a spectrum of effects, ranging from beneficial to detrimental (Beil \u0026amp; Hanes, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Cirino et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Dobbs et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Teixeira et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). An integral aspect of this perception lies in the appreciation of the \"environmental scenic beauty\", reflecting individuals' personal comfort and sense of connection to their surroundings (Knez et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Often regarded as a cultural ecosystem service (CES), this aesthetical non-material perception, linked to a sense of place, social cohesion, and spirituality (Cabana et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), is related to relational values (Chan et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) and ultimately is a nature contribution to people (NCP) ((D\u0026iacute;az et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). This value is frequently sought in natural settings or scenic locations, such as in natural parks, wilderness areas, and other remote spots offering beautiful views (Sandifer et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), but regardless of the contribution of those \u003cem\u003eonce in a time landscape experiences\u003c/em\u003e, this aesthetic service plays a vital role in our daily lives, particularly within urban landscapes (Lenda et al., \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) where most individuals spend most of their time.\u003c/p\u003e\u003cp\u003eIn urban environments, people are less prone to having daily contact with natural and semi-natural environments, leading to an extinction of the human-nature experience (Soga \u0026amp; Gaston, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). While greenery plays a central role in shaping the aesthetic appeal of urban streets, the spatial configuration of vegetation, built infrastructure, and their interplay are critical but underexplored dimensions in urban ecosystem service provision (Chen et al., 2016; Weber et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). This study builds on landscape ecology principles to analyze how heterogeneity and spatial arrangement contribute to the aesthetic cultural ecosystem service, expanding beyond the traditional focus on green cover alone(Chen et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2016a\u003c/span\u003e; Weber et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). The streetscape aesthetics are part of the artificial, natural and semi-natural aspects of the urban ecosystem, and are referred here as a Cultural Ecosystem Service. Particularly, the interspersing of green and built-up areas, the heterogeneity of the landscape or the diversity of green or blue areas should affect the perception of urban aesthetics (Daniel et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Sabbion, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe beauty of public spaces, such as streets, should play an important role in people's well-being (Bratman et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Streets, being primarily for public use, serve as open pathways for all citizens and visitors and are the main commuting location for urban populations. The appearance of streets influences more than just their visual appeal\u0026mdash;it also affects community pride, feelings of safety, ease of walking, and the expression of culture (Henderson et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Therefore, streets provide a shared, public aesthetic and cultural experience.\u003c/p\u003e\u003cp\u003eDespite the considerable impact of urban environments, especially streets, on human quality of life (Hartig \u0026amp; Kahn, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), there has been a scarcity of studies examining how people perceive the aesthetics of city streets. Most of those studies are small-scale investigations that offer valuable insights into citizen preferences concerning urban environments, but with limited ability to extrapolate or make assumptions on larger scales (Chen et al., 2016; Weber et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2008\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eOne notable initiative that attempted to overcome this limitation is the \u003cem\u003ePlace Pulse Project\u003c/em\u003e (Dubey et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), which applied deep learning to predict perceptions of safety, liveliness, boredom, wealth, depression, and beauty across 56 cities using Google Street View imagery. \u003cem\u003ePlace Pulse\u003c/em\u003e was a groundbreaking effort that has since been adapted for studies assessing urban perception in various cities (Rossetti et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Zhang et al., \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). However, while these models successfully predicted how people perceive certain attributes of streets, they did not explicitly analyze the underlying \u003cb\u003elandscape features\u003c/b\u003e that drive these perceptions. In other words, they could classify \"beautiful\" streets but could not determine which urban and ecological characteristics contribute to this beauty, they used extensive survey and cutting-edge technology, but did not connect the human perception to real landscape and ecological attributes.\u003c/p\u003e\u003cp\u003eHere, we deepen this approach by integrating landscape ecology principles into urban perception studies, analyzing landscape features related to human perception. Our study bridges the perspectives of urban dwellers (seeing from the front) and landscape ecologists and urban planners (seeing from above) by analyzing both image-level features (composition, color, complexity) and landscape-level metrics (patch configuration, spatial heterogeneity, three dimensionality and urban form). Here, we examine how people perceive the beauty of streetscapes \u0026mdash; the \u003cem\u003e'daily landscapes'\u003c/em\u003e \u0026mdash; using S\u0026atilde;o Paulo, the largest megacity in the Southern Hemisphere, as our model system. Beyond that, the understanding of how streets are essentially part of the \u003cb\u003eurban ecosystem\u003c/b\u003e is limited, and it is urgent to treat those human-made environments under an ecological lens. Our objective is to overcome this narrow and dichotomous view \u0026ndash; between the natural and the build environment \u0026ndash; to a more holistic and relational approach between the city\u0026rsquo;s semi-natural daily landscapes and their dwellers. By integrating the perception of the landscape to the interplay between build and biological elements of this ecosystem we can overcome the limitations on understanding the provision of CES in everyday life. Although highly anthropogenic, urban streetscapes sustain ecological processes making them legitimate units for analyzing nature\u0026rsquo;s aesthetic contributions to people. To operationalize this perspective, we developed an integrated approach combining perceptual data, landscape metrics, and deep learning to connect what people see at street level with what ecologists and planners measure from above (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eWe conducted an extensive online survey with over 3,000 participants encompassing diverse social and demographic profiles. Each participant evaluated pairs of Google Street View images (SVIs) depicting S\u0026atilde;o Paulo\u0026rsquo;s streetscapes, selecting which street they perceived as more beautiful. These pairwise comparisons were then translated into continuous aesthetic values through Elo score computation (Elo, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). The resulting scores provided a robust, perception-based measure of urban beauty across 420 sampled images, representing the socio-morphological diversity of S\u0026atilde;o Paulo\u0026mdash;the largest megacity in the Southern Hemisphere.\u003c/p\u003e\u003cp\u003eAiming to understand the landscape factors shaping these perceptions, we analyzed each georeferenced image within its surrounding urban fabric using high-resolution satellite data (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Landscape metrics were extracted to quantify vegetation composition, spatial heterogeneity, patch configuration, and three-dimensional urban structure, encompassing both green and built components. This multiscale integration allowed us to test whether beauty is primarily driven by vegetation amount or by the interplay between green and gray elements\u0026mdash;how buildings, trees, and open spaces coexist within the urban mosaic.\u003c/p\u003e\u003cp\u003eFinally, using all the responses of the survey, we trained a deep learning model to generalize these relationships, predicting aesthetic scores for over 350,000 street-level images across the entire city. The resulting map, at 50 m resolution, provides the first comprehensive visualization of S\u0026atilde;o Paulo\u0026rsquo;s aesthetic ecosystem service distribution. Together, these analyses reveal how urban form, ecological structure, and spatial configuration jointly influence people\u0026rsquo;s aesthetic experiences, offering a framework to link perception, landscape ecology, and urban sustainability in an empirically grounded way.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"RESULTS","content":"\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eWe tested the effect of personal characteristics of the respondents as potential effects on the selection of the images in the survey, using generalized linear mixed models to compare different socioeconomic aspects between them. We did not find any significant differences between the individual characteristics (such as age, gender, culture and social level) and their aesthetic preference (examples provided in Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e \u0026ndash; S11), we thus computed the Elo scores by pooling across the 3,221 respondents.\u003c/p\u003e\u003cp\u003eUsing the Green View Index (Li et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) (GVI), a method that measures the index of green in the streets trough SVI, we tested if our aesthetic values (Elo scores) are related to the index of green trough Person\u0026rsquo;s correlation, showing there are is a very small relationship between them, but mostly a spatial mismatch between aesthetic and GVI, with a R\u0026sup2; of just 0.153 between the mean values of the variables in a 50 x50 m grid for the entire city (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eTo understand the determinants of aesthetic value, we extracted for each 420 SVI features (1) from the landscape - above and (2) from the image - front (see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Methods). Using a dredge procedure with Akaike Information Criteria (AIC) model selection (see dredge results at Fig. S12 and Fig. S13), we obtained two models, one with the features from above, one with features from the front. Combining both models, we obtained a final model of the impact of the images and landscape features on aesthetics. The best model from the landscape features - above - presents six of 17 variables tested, with tree volume having the most important (positive) on aesthetic value (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). The proportion of buildings coverage, the number of building patches and the building's edge density were following, all with negative effects (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Analyzing the best model from image features (from the front) of 12 variables used, five were selected (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB), with image shape index (here called image complexity) having the positive strongest effect, followed by fraction of blue in the image and color heterogeneity in the image. The image brightness and the PCA3, that is a mixture between gray and blue in the image (see supplementary Fig. S14 - S16), were following with both negative effects (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB)\u003c/p\u003e\u003cp\u003ePulling both models together, we obtained the final model comparing the relative coefficient estimate of each variable, where the most significant variable acting positively on aesthetic scores were the image complexity and the tree volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). The building coverage, image brightness, followed by buildings edge density were acting negatively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe analysis by Local Climate Zone (LCZ) also revealed significant differences between LCZ zones and the aesthetic values (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Zones with open arrangement of mid-rise or low-rise buildings (e.g. LCZ5 and LCZ6), which provide space between buildings for vegetation and tree cover, had the highest aesthetic values. The lowest street aesthetic values were given to LCZ\u0026rsquo;s dominated by buildings: the LCZ8 (low-big) and the LCZ3 (low compact).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eUsing convolutional neural networks (CCN\u0026rsquo;s) we accurately predicted through deep learning the aesthetic values of images used in our survey. The model showed high R\u0026sup2; values on all three split sets (training\u0026thinsp;=\u0026thinsp;0.92; validation\u0026thinsp;=\u0026thinsp;0.83; test\u0026thinsp;=\u0026thinsp;0.91; full results and values are available on supplementary Fig. S22-S23). With this model in hand, we predicted the aesthetic value for 355,168 SVI of the entire city of S\u0026atilde;o Paulo, representing a spatial resolution of 50 m, that is one SVI each 50m for all the streets mapped in the city (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). This map presents, in an unprecedented manner, the provision of scenic beauty ecosystem services at the scale of a megacity. By juxtaposing this with population density (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB), often considered a proxy for service demand (Bar\u0026oacute; et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), we can discern various combinations of supply and demand within the city of S\u0026atilde;o Paulo, particularly highlighting areas of mismatch into the city (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC). Most of the areas of high aesthetic scores have low population density, corresponding to the city's most affluent and upscale neighborhoods (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD), but some central areas, with taller buildings and the occurrence of green areas between buildings or on the streets, present high demand and high supply (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE). Several areas of the city have high demand but low local supply of the ecosystem service (brightest fuchsia in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC), which generally corresponds to the city's peripheral and less valued neighborhoods (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eOur findings contribute to interdisciplinary understandings of urban aesthetics by demonstrating that perceived beauty extends beyond mere green quantity to encompass the spatial complexity and configuration of vegetation and built-up areas in the streetscape ecosystem. These results align with neurocognitive theories that suggest humans prefer heterogeneous environments due to evolutionary traits (Kaplan, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e1987a\u003c/span\u003e; Richardson et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). By linking visual aesthetic preferences to established landscape ecology metrics, this study underscores the need for multidimensional approaches in assessing cultural ecosystem services in highly anthropic environments that are part of \u003cem\u003edaily landscapes\u003c/em\u003e. Despite green coverage not being selected in our models, the volume of vegetation emerged as one crucial factor in shaping street aesthetics. Conversely, complexity observed from a frontal perspective \u0026ndash; street level view - played a significant role in explaining aesthetic preferences, reflecting the intricate arrangement of vegetation and built-up areas in triggering aesthetic experience within urban landscapes. Images showcasing diverse shapes consistently received higher aesthetic ratings from respondents (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eUpon closer examination of the importance of vegetation, it becomes evident that the effect of greenery involves a crucial three-dimensional aspect. It is not solely the horizontal spread of tree coverage that influences perceived beauty; rather, it is the combined horizontal and vertical extent of greenery. Higher densities of trees have been particularly influential in shaping aesthetic perception. This means that in megacities like S\u0026atilde;o Paulo, street trees formations with larger volumes are highly valued. Consequently, older and taller trees, characterized by greater height or canopy area, and thus, larger volume, are particularly prized (Suchocka et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This underscores the importance of preserving older and bigger trees to maintain aesthetic appeal (Blicharska \u0026amp; Mikusiński, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Suchocka et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). From the perspective of ecosystem services, this finding carries significant implications for street tree management. It highlights that the services provided by large, old trees cannot simply be replaced by planting young trees. Therefore, it underscores the necessity of conserving and nurturing healthy mature trees in urban environments. This is in line with other ecosystem services provided by urban trees in cities, such as microclimate regulation(Kong et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2017\u003c/span\u003e); pollutant control (Grote et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e); noise alleviation (Xu et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) among several others (Salmond et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eMoreover, the comparative analysis between the scenic beauty of streets and vegetation cover indicates that approximately 84% of the variance in scenic service is not explained by the Green View Index, highlighting that urban features beyond seen greenery from street level play a significant role in aesthetic appreciation. The shape complexity of the images was the most important variable to understanding the aesthetic scores, affecting them positively. This variable consists in the shape index of the image seen from the front and is linked to the format and shape that the vegetation appears on the streets, with several leaves, trunks and branches forming intricate and complex figures. This landscape complexity could be more attractive for people because of neurobiological preferences to some visual features (Pearce et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). According to neurocognitive theories people like heterogeneous environments, and the preference for complexity could be an evolutionary characteristic to select places with availability of resources (Kaplan, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e1987b\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eImage complexity can also be associated with the heterogeneity or fragmentation of the urban landscape. Specifically, aerial factors like the presence of built-up edges and the number of buildings patches measured from above were found to negatively impact landscape beauty. This means that areas with lower number of buildings, with space and green cover in between them could be considered more visually pleasant (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA, B) than those areas with dense construction attached to each other (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eC, D). While previous studies have recognized the importance of buildings in shaping the streetscape of cities (Chen et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2016a\u003c/span\u003e; Weber et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), it is becoming evident that the quantity and arrangement of buildings also play critical roles. The number of buildings and their placement in relation to green areas within the landscape significantly impact aesthetic evaluations.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eSurprisingly, socioeconomic profiles appear to have no discernible impact on the perceived aesthetic quality of streets. Some research has shown that the general \u0026ldquo;aesthetic judgment\u0026rdquo; can vary according to age, gender, culture and social characteristics (Kalivoda et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Pugach et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Zhan et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) although our investigation into the association between conflicting profiles revealed that socioeconomic status does not influence preferences for street images (Figs. S1 \u0026ndash; S11). This finding aligns with some existing literature on landscape aesthetics, which suggests that when evaluating the beauty of the environment, consensus often emerges across diverse social groups (Hagerhall, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2001\u003c/span\u003e) or different regions (Chen et al., 2016b). In our study, we observed a similar pattern, indicating that regardless of origin, ethnicity, education level, or gender, individuals tend to converge on shared notions of beauty in urban landscapes, in other words the elements contribute for streetscapes aesthetic CES supply are consensual among dwellers with diverse backgrounds.\u003c/p\u003e\u003cp\u003eThe uneven distribution of aesthetic service throughout the city highlights social disparities, with areas of high demand and low supply consistently located in less privileged neighborhoods (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). When we extrapolated the aesthetic scores to encompass the entire city, wealthier areas within the city consistently exhibited elevated aesthetic values. The zones with the higher aesthetic values are neighborhoods with buildings of middle-rise, sparse on space, with other uses, as green areas, in between the buildings (LCZ 5 and 6). These zones, with high aesthetic service supply and low demand, represent wealthier areas of the city, with low human density, big private houses and gardens and a low number of building patches, similarly to a land-sharing arrangement of the city (Cirino et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Conversely, large regions of the city characterized by higher population densities and lower socioeconomic indicators registered lower aesthetic scores. This pattern is particularly pronounced in densely populated areas, such as urban slums and unplanned occupations in peripheral regions (LCZ 2 and 3) where the high density and high number of buildings and the lack of vegetation result in lower aesthetic ratings. This disparity underscores a significant mismatch between the supply and demand of ecosystem services in these areas (Dobbs et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2018\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), thereby amplifying indications of a \u0026ldquo;luxury effect\u0026rdquo; in the provision of such services (Hope et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Leong et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). These disparities not only highlight socio-economic inequities but also underscore the need for targeted interventions to bridge the gap in access to aesthetically pleasing urban environments.\u003c/p\u003e\u003cp\u003eFrom a broader perspective, these results illustrate how urban aesthetic experiences operate at the interface between CES and NCP, reflecting both material and relational values (Chan et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; D\u0026iacute;az et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). While the cultural service framing traditionally emphasizes the measurable provision of benefits such as scenic enjoyment, the NCP framework acknowledges the relational dimension \u0026mdash; the ways people establish meaningful connections with their surroundings through daily encounters with urban nature. In this sense, the observed inequalities in the spatial distribution of aesthetic opportunities are not only ecological mismatches but also relational injustices, as they limit the capacity of certain groups to build emotional, cultural, and identity-based ties with nature in their everyday environments \u0026ndash; despite they agree about which streetscapes are more beautiful. By integrating perception-based mapping with landscape structure, our study bridges these perspectives, demonstrating that the beauty of urban streets is a tangible manifestation of how ecosystems contribute to human well-being, both through biophysical characteristics and through the relational experiences they enable.\u003c/p\u003e\u003cp\u003eThe map showing the balance of supply and demand can guide urban planning in different ways. In particular, areas characterized by high demand and low to medium supply are prime candidates for urban interventions, as they can positively impact a larger number of city inhabitants and enhance the provision of ecosystem services increasing the contribution of nature-based solutions. Alternatively, considering that urban densification offers significant advantages, including more efficient utilization of urban infrastructure (Bergesen et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Pelczynski \u0026amp; Tomkowicz, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), it is interesting to consider as reference areas of the city that exhibit both high supply and high demand for ecosystem services (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE). These areas typically belong to the middle and high-middle class segments, located in verticalized sections of the city, predominantly represented by LCZ 4 (high-compact) but also LCZ 1 (high-open). Despite their high population density, these neighborhoods are often well-planned, situated in the central parts of the city, and adorned with trees\u0026mdash;particularly those with substantial volume\u0026mdash;lining the streets and interspersed among high-rise buildings and condominiums. This scenario illustrates the possibility of reconciling the supply and demand of aesthetic ecosystem services within the city, facilitated by ample sidewalk space and effective urban planning. This scenario exemplifies cities that have successfully integrated dense populations with ample green spaces, mirroring the model seen in Singapore and Hong Kong, among others (Wu et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Xue et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eMore generally our approach holds potential not only for S\u0026atilde;o Paulo but also for cities worldwide, particularly megacities grappling with complex urban challenges. The consistent correlation between scenic beauty evaluations and the LCZ indicates that these zones can serve as a basis for extrapolating the results or gaining a deeper understanding of the urban conditions influencing the appreciation of scenic beauty on urban streets. Moreover, S\u0026atilde;o Paulo, being a pronounced diverse city, with areas representing different urban settings, architecture diversification and social realities, fitted as a great trial for assessing perception and landscape patterns on the supply of aesthetic CES.\u003c/p\u003e\u003cp\u003eBy harnessing our methodology, which integrates cutting-edge technology with landscape analysis, cities globally can gain unprecedented insights into the factors influencing the aesthetic quality of their urban streets. This presents a groundbreaking opportunity for urban planners and policymakers to make informed decisions aimed at enhancing the beauty and equity of their cities. By enhancing the aesthetic appeal of streets, a range of co-benefits is likely to emerge, including increased walkability (Lwin \u0026amp; Murayama, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2011\u003c/span\u003e), improved health and social cohesion (De Vries et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), heightened social interaction, a stronger sense of place (Semenza et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) and safety (De Nadai et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), as well as greater use of public spaces for physical and recreational activities, all of which contribute to the overall well-being of residents (Bratman et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Dai et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; De Vries et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2013\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eOur findings highlight the universality of certain principles governing urban aesthetics, suggesting that lessons learned in S\u0026atilde;o Paulo can be extrapolated to other global cities. By embracing this approach, cities across the world can leverage data-driven strategies to address urban social justice, promote public health, and create more livable and sustainable urban environments. Ultimately, our study reveals a pioneering tool for urban planning, with the potential to transform the way cities envision and design their streetscapes with more equity considering the relational dimension between people and their daily environment, treating the streets as part of a complex urban ecosystem that can avoid the extinction of human-nature experience.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"METHODS","content":"\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003eStudy area and landscape/images sampling\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eS\u0026atilde;o Paulo City is the biggest megacity of Brazil and the biggest in south hemisphere, with more than 12\u0026nbsp;million people in the municipality and more than 21\u0026nbsp;million in the metropolitan region. S\u0026atilde;o Paulo presents high social inequality and diverse morphology in terms of building densification and urban forestry, but also socio-cultural identity. The urban area of the city is more than 914.5 km\u0026sup2;, and 48.18% of the municipality area is covered by vegetation (S\u0026atilde;o Paulo, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Most of the vegetation is represented by secondary and semi-natural Atlantic rainforest, concentrated in two big state parks, one in the north and another in the south of the city (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA), making most of this vegetation isolated from the local population.\u003c/p\u003e\u003cp\u003eA stratified sampling approach was employed to ensure representation across S\u0026atilde;o Paulo's seven urban morphologies, categorized by Local Climate Zones (LCZ) (Stewart and Oke, 2012; Ferreira et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). This method enables robust extrapolation of aesthetic perceptions to diverse urban contexts while maintaining ecological relevance. Each sampled Street View Image (SVI) represents typical streetscapes for its LCZ, providing a systematic basis for integrating landscape-level and street-level features (Ferreira et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). We randomly choose 60 street points in each LCZ (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA) and collected for each one the image on Google Street View with an angle of view of 180\u0026deg; in relation to the street line, to ensure that the images face from the front of each street.\u003c/p\u003e\u003cp\u003eFor each collected image, by drawing polygons with trapeze shape, we estimated the landscape correspondent seen from above (Figs.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB; \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC). Each polygon has variable sizes, depending on the reach of view of each image seen from the front. For each landscape we estimated the composition of green and buildings (proportion on the landscape), based on the map of vegetation cover of S\u0026atilde;o Paulo municipality (S\u0026atilde;o Paulo, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The data consist of a vegetation classification in the resolution of 1:1000 inside the city, and 1:5000 within protected areas, based on orthophotos with 0.12 m of resolution, from 2017, obtained by the municipal environment office. Additionally, we used LiDAR data on the resolution of 10 points per m\u0026sup2; and precision of 0.1m to estimate the area occupied by buildings (Gomes, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eUsing the LiDAR data, we estimated the height and volume density of trees and buildings. Additionally, we also evaluated the configuration of the urban landscape (buildings and different types of vegetation), through: (1) large patch index; (2) edge density; and (3) number of patches (McGarigal, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2001\u003c/span\u003e), using the R package \u003cem\u003elandscapemetrics v2.1.4.\u003c/em\u003e In that way we obtained three subsets of landscape measurements, (1) the composition, (2) the configuration and (3) the three-dimensional morphology (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFor each image collected we extracted image features that may influence human aesthetic perception. We used 500x500 pixel images at 96 dpi. (1) \u003cem\u003eColor heterogeneity\u003c/em\u003e was measured following the procedure used in Langlois et al. (Langlois et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2022\u003c/span\u003e): the K-means clustering algorithm was used to separate each pixel of an image in the CIELAB color space (we used 9 clusters). The mean distance between the 9 cluster centers was used as a measure of color heterogeneity. (2) \u003cem\u003eColor saturation\u003c/em\u003e and (3) \u003cem\u003eBrightness\u003c/em\u003e of each image were measured using the HSV (Hue, Saturation, Value) color space that differentiates saturation (S) from perceptual lightness (V). Images (4) \u003cem\u003eContrast\u003c/em\u003e and (5) \u003cem\u003eFractal dimension\u003c/em\u003e (self_similarity) were measured using the function \u003cem\u003eimg_contrast()\u003c/em\u003e and \u003cem\u003eimg_self_similarity()\u003c/em\u003e of the R package \u003cem\u003eimagefluency v.0.2.5\u003c/em\u003e. The proportion of (6) \u003cem\u003eGreen\u003c/em\u003e, (7) \u003cem\u003eBlue\u003c/em\u003e and (8) \u003cem\u003eGray\u003c/em\u003e in each image was computed with the function \u003cem\u003ecountColors()\u003c/em\u003e of the R package \u003cem\u003ecountcolors v.0.9.1\u003c/em\u003e. Finally the R package \u003cem\u003ecolordistance v.1.1.2\u003c/em\u003e was used to provide a more integrated measure of color distance between images. The functions \u003cem\u003egetLabHistList()\u003c/em\u003e and \u003cem\u003egetColorDistanceMatrix()\u003c/em\u003e were used to obtain color distance matrix between all images which was then reduced using a PCA analysis (function \u003cem\u003edudi.pca\u003c/em\u003e() of the package \u003cem\u003eade4 v.1.7\u0026ndash;22\u003c/em\u003e). The three first axes of the PCA accounted for most of the variance (respectively 51.3%, 29,7% and 8.6%) and could be used to classify the images along meaningful axes of dominant color and/or color mixtures (Fig. S14 \u0026ndash; Fig. S16).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eOnline survey and Elo’s scores\u003c/h3\u003e\n\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eWe conducted an online survey, available to the general public between May 01st and September 30th, 2023, presenting to each respondent 30 pairs of images randomly selected from the pool of 420 SVI. For each pair, respondents were asked to click on the image they thought show the most beautiful street. Each respondent then was redirected to a socioeconomic and profile survey to collect information on sociocultural backgrounds (gender; ethnicity; age; scholarity; social class; type/size of the city where they live and grew up; Fig. S18). All the responses were anonymous to protect the identity of the respondents.\u003c/p\u003e\u003cp\u003eThe online survey was available in Portuguese on a dedicated website (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.biodiful.org/\u003c/span\u003e\u003cspan address=\"https://www.biodiful.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e)\u003c/span\u003e and was answered by 3,221 respondents. The survey was advertised by e-mails list for universities and research centers and asked to be publicly disclosed. A publication was also advertised on \u003cem\u003eInstagram\u003c/em\u003e; the target audience was any Brazilian adult over 18 years old. Note that participants were asked if they had color perception deficiency, and, if yes, that we removed their answer from the final dataset analyzed.\u003c/p\u003e\u003cp\u003eTo account for sociocultural background on the statistical analysis, we employed a generalized linear mixed model (GLMM) with a binomial error structure. This analysis was conducted using the \u003cem\u003eglmer()\u003c/em\u003e function from the R package \u003cem\u003elme4 v1.1\u0026ndash;26\u003c/em\u003e. In this model, the image was treated as a random effect variable, allowing us to assess the individual impact of sociocultural variables on the response variable and effectively order them based on their influence. This analysis showed no effect of any of the sociocultural variables which allowed us to pool all the respondent answers to compute a global Elo scores analysis.\u003c/p\u003e\u003cp\u003eThe Elo scoring method(Elo, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2008\u003c/span\u003e) was employed to compare images as paired competitors, ultimately assigning final images scores based on the cumulative comparisons within our dataset. We used the R package \u003cem\u003eEloChoice v0.29.4\u003c/em\u003e(Clark et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) with 1,000 bootstrapings. This methodological approach enabled a nuanced examination of perceived image beauty, offering insights into the aesthetic inclinations within our sampled image population. Image Elo scores will be referred to as \u0026ldquo;aesthetic value\u0026rdquo; hereafter.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003eStatistical analysis – understanding front and above features\u003c/h3\u003e\n\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eAiming to understand the factors associated to the preference of people by the images, we divided the statistical analysis into two components: one of predictive variables based on the landscape \u0026ndash; from \u003cem\u003eabove\u003c/em\u003e; and another of predictive variables based on the image\u0026rsquo;s features \u0026ndash; from the \u003cem\u003efront\u003c/em\u003e (street-level view). For each set of variables, we constructed a full model with all variables in the same Linear Model (LM), the predictive variables are described in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, they represent all the variables tested in the study. Some of them were log-transformed to adjust to the linear model. The response variable was the aesthetic values of each of the 420 SVI used in our online survey.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003e\u003cb\u003eVariables used in the full model, divided according to the group\u003c/b\u003e: from above (landscape) and from the front (SVI\u0026rsquo;s features). All the following variables were tested. The ones signed with * where log transformed.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eVariable\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eGroup/Type\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003e\u003cb\u003eComposition\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eProportion of buildings coverage\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eProportion of tree coverage*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eProportion of open vegetation coverage*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eProportion of total vegetation coverage*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"8\" rowspan=\"9\"\u003e\u003cp\u003e\u003cb\u003eConfiguration\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLarge patch index of total vegetation*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEdge density of total vegetation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNumber of patches of total vegetation*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLarge patch index of buildings\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEdge density of buildings\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNumber of patches of buildings\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLarge patch index of trees*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEdge density of trees\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNumber of patches of trees*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003e\u003cb\u003eHeight \u0026amp; Volume\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMean height of trees\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eVolume of trees*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMean height of buildings*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eVolume of buildings*\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eabove\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"11\" rowspan=\"12\"\u003e\u003cp\u003e\u003cb\u003eImages features\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eColor heterogeneity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eComplexity/shape index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSaturation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBrightness\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePCA 1 - Gray and darkness/shadows mixture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePCA 2 \u0026ndash; Green and shadows mixture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePCA 3 - Gray and blue mixture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eContrast\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFractal/Self similarity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGreen proportion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGray proportion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBlue proportion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efront\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eAfter generating the full models, we run a dredge process (R, package \u003cem\u003eMuMln v1.47.5\u003c/em\u003e), that consists in generating all the possible combinations between one and the total number of variables adjusted to the response variable \u0026ndash; aesthetics values. After generating all models, the dredge ranks the best models by Akaike Information Criteria (AIC), searching for the model that best adjusts to our response. We ran two dredge processes, one for the variables from the front and another to above. For each dredge process we kept just the variables present in the best model, with the lowest AIC value. On sequence we tested for Variance Inflation factor using the \u003cem\u003evif()\u003c/em\u003e function from the R package \u003cem\u003ecar v4.3.3\u003c/em\u003e, to detect multicollinearity in the linear regression. Multicollinearity occurs when independent variables in a regression model are highly correlated with each other, if multicollinearity is detected (VIF values are higher than 4) we investigated the \u003cem\u003ePearson\u003c/em\u003e correlation of the variables in order to exclude the highly correlated ones (more than 0.6 of correlation).\u003c/p\u003e\u003cp\u003eGiven that all SVI have an explicitly spatial component, we tested the best models of each set of variables for Spatial Autocorrelation. We used Moran\u0026rsquo;s I test to check autocorrelation of the residuals given the coordinates of each sample (R package \u003cem\u003eDHARMa v0.4.6\u003c/em\u003e). The test pointed to spatial autocorrelation for both models \u0026ndash; see supplementary text \u0026ndash; for that reason we opt to use a modeling process that considers the space as a mixed part of the model. We ran a model with the variables selected by AIC using the \u003cem\u003eFitMe()\u003c/em\u003e function (Fitting function for fixed and mixed-effect models with GLM response of the R package \u003cem\u003espaMM)\u003c/em\u003e. This function allows fitting a Generalized Linear Model (GLM) with a mixed part of the model with non-gaussian random effects, that is the case of the spatial information \u0026ndash; latitude and longitude of the samples. With that, we could consider as part of the model the spatial distribution of our samples, avoiding model inflation by spatial autocorrelation. In addition to space, we added as a random variable in the model the Local Climate Zone (LCZ) where each of the 420 images were collected, in order to control the random effects of the morphology of the city in the statistical analysis. We analyzed the fixed part of each model, given by the coefficient estimates of each variable in relation to the response variable (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e to Table S3).\u003c/p\u003e\u003cp\u003eWith this process we obtained three final models controlling the space, one for the variables from the front, another for the variables from above and finally, for the third model, we gathered the variables of both previous models in a new \u003cem\u003eFitMe\u003c/em\u003e model, combining the effect of both the front and above variables to see the relative estimate of each variable in relation to each other. Moreover, we ran an aditional model considering as predictive variables the LCZ of the city, that are a simplified way to categorize the morphology of the city, and are common for several cities across the world. We ran an ANOVA analysis with Tukey test to compare pair to pair of each LCZ, searching for significant difference (p \u0026ndash; value\u0026thinsp;\u0026lt;\u0026thinsp;0.005) between LCZ categories (results are presented in Table S4).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003ePrediction of aesthetic values with a CNN model\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe 420 images evaluated in the online survey and their aesthetic values were used as a training dataset for a deep learning algorithm. We used a Convolutional Neural Network (CNN) deep learning algorithm that involved fine-tuning a CNN pre-trained on ImageNet(Deng et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) to facilitate transfer learning. Our dataset was partitioned into training (314 images), validation (53 images), and testing (53 images) sets. Data augmentation, fine-tuning of the models, and prediction of the aesthetic scores were carried out using Python 3.7, \u003cem\u003ePytorch 1.4.0\u003c/em\u003e, and \u003cem\u003etorchvision 0.5.0\u003c/em\u003e. We used the ResNet50 architecture initialized with weights pre-trained on the ImageNet dataset, available via the \u003cem\u003etorchvision.models\u003c/em\u003e module from \u003cem\u003ePyTorch\u003c/em\u003e, and fine-tuned it using our training set. The model was trained with a learning rate of 1e-2, a batch size of 16, and for up to 350 epochs, with additional regularization techniques such as dropout and weight decay to mitigate overfitting. The final performance of the model was estimated using several evaluation metrics, including the mean squared error (MSE), mean absolute error (MAE), median absolute error (MedAE), mean absolute percentage error (MAPE), mean error (ME), and the coefficient of determination (R\u0026sup2;). These metrics compared the values predicted by the model to the values in the testing set, which was not used during training. The results revealed a remarkably accurate prediction of aesthetic values (e.g., R\u0026sup2; = 0.91 on the testing set), demonstrating the model\u0026rsquo;s ability to estimate human-perceived aesthetic value of the images included in our survey (see supplementary Fig. S22-S23 for a complete visualization of the metrics used).\u003c/p\u003e\u003cp\u003eThe trained CNN model was then used to predict the aesthetic values for images downloaded on google street view for the whole city of S\u0026atilde;o Paulo. We used the shapefile of lines containing all the streets, roads and avenues of the city, provided by the city-hall, and created a point each 50m in all the public streets of the city, totalizing 382,760 points, from which we extracted the coordinates. The 382,760 locations were used to download corresponding street view images with the function \u003cem\u003egoogle_streetview()\u003c/em\u003e of the R package \u003cem\u003egoogleway v2.7.8\u003c/em\u003e using the same angle view \u0026ndash; 180\u0026deg; in relation to the street - that the 420 images collected for our online survey.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eFiltering Street View Images out of the range\u003c/h3\u003e\n\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe download of the SVI resulted in a set of 355,801 images (the function returned no image for 27,013 locations). We used a trained image filter to exclude the sample images that correspond to wrong angles of view \u0026ndash; facing isolated objects or buildings facades \u0026ndash; and images that deviate from the expected pattern \u0026ndash; as blurred images or images taken from inside of tunnels. We ended up with a dataset of 355,168 images for which we used our trained CNN model to predict the aesthetic values and produce the map of the aesthetic ecosystem service for the city.\u003c/p\u003e\u003cp\u003eFor these 355,168 images, we also extracted the same visual features used in the original 420 images from the perception survey: color heterogeneity, color saturation, brightness, perceptual lightness, contrast, fractal dimension (self-similarity), and the proportions of green, blue, and gray. We conducted a Principal Component Analysis (PCA) using the original 420 images and projected the 355,168 images into this PCA space (Fig. S19). Analyzing the first two principal components, we found that 94.2% of the projected images fell within the convex hull area defined by the original 420 training images.\u003c/p\u003e\u003cp\u003eTo further refine our prediction set and ensure representativeness, we retained only those images falling within the hull area plus a 10% margin. This resulted in a final dataset comprising 96.9% of the full image set, demonstrating that the original survey-based sample is highly representative of the broader image dataset across the city. These steps confirm the robustness and reliability of our deep learning model for extrapolating aesthetic values in S\u0026atilde;o Paulo\u0026rsquo;s urban streetscapes (Fig. S20).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe also thank Artur Lupinetti-Cunha for the data on population and Gabriel Garcia for the design of some figures. We thank professors Denise Duarte, Renata Pardini and Vitor Vasconselos for their comments in early draft of this manuscript. We thank L. Roman Carrasco for his valuable conceptual insights and discussions during DWC\u0026rsquo;s research stay at the National University of Singapore, which helped shape the theoretical framing of this study Finally, we especially thank all the 3,221 respondents of the survey.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics statement and consent\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll methods were carried out in accordance with the relevant institutional guidelines and regulations. The study protocol, including the online survey in which participants rated the aesthetic quality of urban street images, was reviewed and approved by the Ethics Committee for Research with Human Subjects of the Institute of Biosciences, University of S\u0026atilde;o Paulo (Comit\u0026ecirc; de \u0026Eacute;tica em Pesquisa com Seres Humanos do Instituto de Bioci\u0026ecirc;ncias da Universidade de S\u0026atilde;o Paulo; approval no. 57.763.208). All participants were adults (\u0026ge;18 years old) and provided informed consent prior to participating in the survey. No personally identifiable information was collected.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eS\u0026atilde;o Paulo Research Foundation - FAPESP n. process: 2020/15785-7 (DWC)\u003c/p\u003e\n\u003cp\u003eS\u0026atilde;o Paulo Research Foundation - FAPESP n. process: 2020/06694-8 (JPM)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Conceptualization: DWC, NM, JPM\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Methodology: DWC, NM, JPM\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Investigation: DWC, NM\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Visualization: DWC, NM\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Supervision: JPM\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Writing\u0026mdash;original draft: DWC\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Writing\u0026mdash;review \u0026amp; editing: DWC, NM, JPM\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e Authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData and materials availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data, code, and materials used in the analyses is available at: https://github.com/DougCirino/AESTHETIC_STREETS\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDOI: 10.5281/zenodo.13318006\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBar\u0026oacute;, F., G\u0026oacute;mez-Baggethun, E., \u0026amp; Haase, D. (2017). Ecosystem service bundles along the urban-rural gradient: Insights for landscape planning and management. \u003cem\u003eEcosystem Services\u003c/em\u003e, \u003cem\u003e24\u003c/em\u003e, 147\u0026ndash;159. https://doi.org/10.1016/J.ECOSER.2017.02.021\u003c/li\u003e\n\u003cli\u003eBeil, K., \u0026amp; Hanes, D. (2013). The influence of urban natural and built environments on physiological and psychological measures of stress- A pilot study. \u003cem\u003eInternational Journal of Environmental Research and Public Health\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(4), 1250\u0026ndash;1267. https://doi.org/10.3390/ijerph10041250\u003c/li\u003e\n\u003cli\u003eBergesen, J. D., Suh, S., Baynes, T. M., \u0026amp; Musango, J. K. (2017). Environmental and natural resource implications of sustainable urban infrastructure systems. \u003cem\u003eEnvironmental Research Letters\u003c/em\u003e, \u003cem\u003e12\u003c/em\u003e(12), 125009. https://doi.org/10.1088/1748-9326/AA98CA\u003c/li\u003e\n\u003cli\u003eBlicharska, M., \u0026amp; Mikusiński, G. (2014). Incorporating Social and Cultural Significance of Large Old Trees in Conservation Policy. \u003cem\u003eConservation Biology\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(6), 1558\u0026ndash;1567. https://doi.org/10.1111/COBI.12341\u003c/li\u003e\n\u003cli\u003eBratman, G. N., Anderson, C. B., Berman, M. G., Cochran, B., de Vries, S., Flanders, J., Folke, C., Frumkin, H., Gross, J. J., Hartig, T., Kahn, P. H., Kuo, M., Lawler, J. J., Levin, P. S., Lindahl, T., Meyer-Lindenberg, A., Mitchell, R., Ouyang, Z., Roe, J., \u0026hellip; Daily, G. C. (2019). Nature and mental health: An ecosystem service perspective. \u003cem\u003eScience Advances\u003c/em\u003e, \u003cem\u003e5\u003c/em\u003e(7), 903\u0026ndash;927. https://doi.org/10.1126/SCIADV.AAX0903/SUPPL_FILE/AAX0903_SM.PDF\u003c/li\u003e\n\u003cli\u003eBratman, G. N., Hamilton, J. P., Hahn, K. S., Daily, G. C., \u0026amp; Gross, J. J. (2015). Nature experience reduces rumination and subgenual prefrontal cortex activation. \u003cem\u003eProceedings of the National Academy of Sciences of the United States of America\u003c/em\u003e, \u003cem\u003e112\u003c/em\u003e(28), 8567\u0026ndash;8572. https://doi.org/10.1073/pnas.1510459112\u003c/li\u003e\n\u003cli\u003eCabana, D., Ryfield, F., Crowe, T. P., \u0026amp; Brannigan, J. (2020). Evaluating and communicating cultural ecosystem services. \u003cem\u003eEcosystem Services\u003c/em\u003e, \u003cem\u003e42\u003c/em\u003e, 101085. https://doi.org/10.1016/j.ecoser.2020.101085\u003c/li\u003e\n\u003cli\u003eChan, K. M. A., Balvanera, P., Benessaiah, K., Chapman, M., D\u0026iacute;az, S., G\u0026oacute;mez-Baggethun, E., Gould, R., Hannahs, N., Jax, K., Klain, S., Luck, G. W., Mart\u0026iacute;n-L\u0026oacute;pez, B., Muraca, B., Norton, B., Ott, K., Pascual, U., Satterfield, T., Tadaki, M., Taggart, J., \u0026amp; Turner, N. (2016). Why protect nature? Rethinking values and the environment. \u003cem\u003eProceedings of the National Academy of Sciences of the United States of America\u003c/em\u003e, \u003cem\u003e113\u003c/em\u003e(6), 1462\u0026ndash;1465. https://doi.org/10.1073/PNAS.1525002113/ASSET/6CB06F81-F2E6-43FD-90F6-81CFCEB8FBBE/ASSETS/GRAPHIC/PNAS.1525002113FIG02.JPEG\u003c/li\u003e\n\u003cli\u003eChen, Z., Xu, B., \u0026amp; Devereux, B. (2016a). Assessing public aesthetic preferences towards some urban landscape patterns: the case study of two different geographic groups. \u003cem\u003eEnvironmental Monitoring and Assessment\u003c/em\u003e, \u003cem\u003e188\u003c/em\u003e(1), 1\u0026ndash;17. https://doi.org/10.1007/S10661-015-5007-3/METRICS\u003c/li\u003e\n\u003cli\u003eCirino, D. W., Tambosi, L. R., Mauad, T., de Freitas, S. R., \u0026amp; Metzger, J. P. (2022). Balanced spatial distribution of green areas creates healthier urban landscapes. \u003cem\u003eJournal of Applied Ecology\u003c/em\u003e, \u003cem\u003e59\u003c/em\u003e(7), 1884\u0026ndash;1896. https://doi.org/10.1111/1365-2664.14195\u003c/li\u003e\n\u003cli\u003eClark, A. P., Howard, K. L., Woods, A. T., Penton-Voak, I. S., \u0026amp; Neumann, C. (2018). Why rate when you could compare? Using the \u0026ldquo;EloChoice\u0026rdquo; package to assess pairwise comparisons of perceived physical strength. \u003cem\u003ePLOS ONE\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(1), e0190393. https://doi.org/10.1371/JOURNAL.PONE.0190393\u003c/li\u003e\n\u003cli\u003eDai, L., Zheng, C., Dong, Z., Yao, Y., Wang, R., Zhang, X., Ren, S., Zhang, J., Song, X., \u0026amp; Guan, Q. (2021). Analyzing the correlation between visual space and residents\u0026rsquo; psychology in Wuhan, China using street-view images and deep-learning technique. \u003cem\u003eCity and Environment Interactions\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e, 100069. https://doi.org/10.1016/J.CACINT.2021.100069\u003c/li\u003e\n\u003cli\u003eDaniel, T. C., Muhar, A., Arnberger, A., Aznar, O., Boyd, J. W., Chan, K. M. A., Costanza, R., Elmqvist, T., Flint, C. G., Gobster, P. H., Gr\u0026ecirc;t-Regamey, A., Lave, R., Muhar, S., Penker, M., Ribe, R. G., Schauppenlehner, T., Sikor, T., Soloviy, I., Spierenburg, M., \u0026hellip; Von Der Dunk, A. (2012). Contributions of cultural services to the ecosystem services agenda. \u003cem\u003eProceedings of the National Academy of Sciences of the United States of America\u003c/em\u003e, \u003cem\u003e109\u003c/em\u003e(23), 8812\u0026ndash;8819. https://doi.org/10.1073/PNAS.1114773109/ASSET/BB9209BA-B18B-4515-B848-50D3A3011DBB/ASSETS/GRAPHIC/PNAS.1114773109FIG01.JPEG\u003c/li\u003e\n\u003cli\u003eDe Nadai, M., Vieriu, R. L., Zen, G., Dragicevic, S., Naik, N., Caraviello, M., Hidalgo, C. A., Sebe, N., \u0026amp; Lepri, B. (2016). Are safer looking neighborhoods more lively? A multimodal investigation into urban life. \u003cem\u003eMM 2016 - Proceedings of the 2016 ACM Multimedia Conference\u003c/em\u003e, 1127\u0026ndash;1135. https://doi.org/10.1145/2964284.2964312\u003c/li\u003e\n\u003cli\u003eDe Vries, S., van Dillen, S. M. E., Groenewegen, P. P., \u0026amp; Spreeuwenberg, P. (2013). Streetscape greenery and health: Stress, social cohesion and physical activity as mediators. \u003cem\u003eSocial Science \u0026amp; Medicine\u003c/em\u003e, \u003cem\u003e94\u003c/em\u003e, 26\u0026ndash;33. https://doi.org/10.1016/J.SOCSCIMED.2013.06.030\u003c/li\u003e\n\u003cli\u003eDeng, J., Dong, W., Socher, R., Li, L.-J., Kai Li, \u0026amp; Li Fei-Fei. (2010). \u003cem\u003eImageNet: A large-scale hierarchical image database\u003c/em\u003e. 248\u0026ndash;255. https://doi.org/10.1109/CVPR.2009.5206848\u003c/li\u003e\n\u003cli\u003eD\u0026iacute;az, S., Pascual, U., Stenseke, M., Mart\u0026iacute;n-L\u0026oacute;pez, B., Watson, R. T., Moln\u0026aacute;r, Z., Hill, R., Chan, K. M. A., Baste, I. A., Brauman, K. A., Polasky, S., Church, A., Lonsdale, M., Larigauderie, A., Leadley, P. W., Van Oudenhoven, A. P. E., Van Der Plaat, F., Schr\u0026ouml;ter, M., Lavorel, S., \u0026hellip; Shirayama, Y. (2018). Assessing nature\u0026rsquo;s contributions to people: Recognizing culture, and diverse sources of knowledge, can improve assessments. \u003cem\u003eScience\u003c/em\u003e, \u003cem\u003e359\u003c/em\u003e(6373), 270\u0026ndash;272. https://doi.org/10.1126/SCIENCE.AAP8826/SUPPL_FILE/AAP8826-DIAZ-SM.PDF\u003c/li\u003e\n\u003cli\u003eDobbs, C., Escobedo, F. J., Clerici, N., de la Barrera, F., Eleuterio, A. A., MacGregor-Fors, I., Reyes-Paecke, S., V\u0026aacute;squez, A., Cama\u0026ntilde;o, J. D. Z., \u0026amp; Hern\u0026aacute;ndez, H. J. (2019). Urban ecosystem Services in Latin America: mismatch between global concepts and regional realities? \u003cem\u003eUrban Ecosystems\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(1), 173\u0026ndash;187. https://doi.org/10.1007/s11252-018-0805-3\u003c/li\u003e\n\u003cli\u003eDobbs, C., Escobedo, F. J., \u0026amp; Zipperer, W. C. (2011). A framework for developing urban forest ecosystem services and goods indicators. \u003cem\u003eLandscape and Urban Planning\u003c/em\u003e, \u003cem\u003e99\u003c/em\u003e(3\u0026ndash;4), 196\u0026ndash;206. https://doi.org/10.1016/J.LANDURBPLAN.2010.11.004\u003c/li\u003e\n\u003cli\u003eDobbs, C., Hern\u0026aacute;ndez-Moreno, \u0026Aacute;., Reyes-Paecke, S., \u0026amp; Miranda, M. D. (2018). Exploring temporal dynamics of urban ecosystem services in Latin America: The case of Bogota (Colombia) and Santiago (Chile). \u003cem\u003eEcological Indicators\u003c/em\u003e, \u003cem\u003e85\u003c/em\u003e, 1068\u0026ndash;1080. https://doi.org/10.1016/j.ecolind.2017.11.062\u003c/li\u003e\n\u003cli\u003eDubey, A., Naik, N., Parikh, D., Raskar, R., \u0026amp; Hidalgo, C. A. (2016). Deep learning the city: Quantifying urban perception at a global scale. \u003cem\u003eLecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)\u003c/em\u003e, \u003cem\u003e9905 LNCS\u003c/em\u003e, 196\u0026ndash;212. https://doi.org/10.1007/978-3-319-46448-0_12/FIGURES/4\u003c/li\u003e\n\u003cli\u003eElo, A. E. (2008). \u003cem\u003eThe rating of chessplayers, past and present\u003c/em\u003e. 208. https://books.google.com/books/about/The_Rating_of_Chessplayers.html?hl=pt-BR\u0026amp;id=syjcPQAACAAJ\u003c/li\u003e\n\u003cli\u003eFerreira, L. S., Prata-Shimomura, A., Ferreira, A., Helena, D., \u0026amp; Duarte, S. (2017). \u003cem\u003eMapping Local Climate Zones for S\u0026atilde;o Paulo Metropolitan Region: a comparison between the local climate zone map and two other local maps\u003c/em\u003e. https://www.researchgate.net/publication/318404791\u003c/li\u003e\n\u003cli\u003eGomes, F. (2022). \u003cem\u003eBuilding and Vegetation Height Model of S\u0026atilde;o Paulo\u003c/em\u003e. https://doi.org/10.34740/KAGGLE/DSV/4044806\u003c/li\u003e\n\u003cli\u003eGrote, R., Samson, R., Alonso, R., Amorim, J. H., Cari\u0026ntilde;anos, P., Churkina, G., Fares, S., Thiec, D. Le, Niinemets, \u0026Uuml;., Mikkelsen, T. N., Paoletti, E., Tiwary, A., \u0026amp; Calfapietra, C. (2016). Functional traits of urban trees: air pollution mitigation potential. \u003cem\u003eFrontiers in Ecology and the Environment\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(10), 543\u0026ndash;550. https://doi.org/10.1002/FEE.1426\u003c/li\u003e\n\u003cli\u003eHagerhall, C. M. (2001). CONSENSUS IN LANDSCAPE PREFERENCE JUDGEMENTS. \u003cem\u003eJournal of Environmental Psychology\u003c/em\u003e, \u003cem\u003e21\u003c/em\u003e(1), 83\u0026ndash;92. https://doi.org/10.1006/JEVP.2000.0186\u003c/li\u003e\n\u003cli\u003eHartig, T., \u0026amp; Kahn, P. H. (2016). Living in cities, naturally. \u003cem\u003eScience\u003c/em\u003e, \u003cem\u003e352\u003c/em\u003e(6288), 938\u0026ndash;940. https://doi.org/10.1126/SCIENCE.AAF3759\u003c/li\u003e\n\u003cli\u003eHenderson, H., Child, S., Moore, S., Moore, J. B., \u0026amp; Kaczynski, A. T. (2016). The Influence of Neighborhood Aesthetics, Safety, and Social Cohesion on Perceived Stress in Disadvantaged Communities. \u003cem\u003eAmerican Journal of Community Psychology\u003c/em\u003e, \u003cem\u003e58\u003c/em\u003e(1\u0026ndash;2), 80\u0026ndash;88. https://doi.org/10.1002/AJCP.12081\u003c/li\u003e\n\u003cli\u003eHope, D., Gries, C., Zhu, W., Fagan, W. F., Redman, C. L., Grimm, N. B., Nelson, A. L., Martin, C., \u0026amp; Kinzig, A. (2003). Socioeconomics drive urban plant diversity. \u003cem\u003eProceedings of the National Academy of Sciences of the United States of America\u003c/em\u003e, \u003cem\u003e100\u003c/em\u003e(15), 8788\u0026ndash;8792. https://doi.org/10.1073/PNAS.1537557100/ASSET/9F5B8293-C983-4B01-B5B3-90DE21ECFEE5/ASSETS/GRAPHIC/PQ1537557002.JPEG\u003c/li\u003e\n\u003cli\u003eKalivoda, O., Vojar, J., Skřivanov\u0026aacute;, Z., \u0026amp; Zahradn\u0026iacute;k, D. (2014). Consensus in landscape preference judgments: The effects of landscape visual aesthetic quality and respondents\u0026rsquo; characteristics. \u003cem\u003eJournal of Environmental Management\u003c/em\u003e, \u003cem\u003e137\u003c/em\u003e, 36\u0026ndash;44. https://doi.org/10.1016/J.JENVMAN.2014.02.009\u003c/li\u003e\n\u003cli\u003eKaplan, S. (1987a). Aesthetics, Affect, and Cognition. \u003cem\u003eEnvironment and Behavior\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(1), 3\u0026ndash;32. https://doi.org/10.1177/0013916587191001\u003c/li\u003e\n\u003cli\u003eKaplan, S. (1987b). Aesthetics, Affect, and Cognition. \u003cem\u003eHttp://Dx.Doi.Org/10.1177/0013916587191001\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(1), 3\u0026ndash;32. https://doi.org/10.1177/0013916587191001\u003c/li\u003e\n\u003cli\u003eKnez, I., Sang, A. O., Gunnarsson, B., \u0026amp; Hedblom, M. (2018). Wellbeing in urban greenery: The role of naturalness and place identity. \u003cem\u003eFrontiers in Psychology\u003c/em\u003e, \u003cem\u003e9\u003c/em\u003e(APR), 344587. https://doi.org/10.3389/FPSYG.2018.00491/BIBTEX\u003c/li\u003e\n\u003cli\u003eKong, L., Lau, K. K. L., Yuan, C., Chen, Y., Xu, Y., Ren, C., \u0026amp; Ng, E. (2017). Regulation of outdoor thermal comfort by trees in Hong Kong. \u003cem\u003eSustainable Cities and Society\u003c/em\u003e, \u003cem\u003e31\u003c/em\u003e, 12\u0026ndash;25. https://doi.org/10.1016/J.SCS.2017.01.018\u003c/li\u003e\n\u003cli\u003eLanglois, J., Guilhaumon, F., Baletaud, F., Casajus, N., De Almeida Braga, C., Fleur\u0026eacute;, V., Kulbicki, M., Loiseau, N., Mouillot, D., Renoult, J. P., Stahl, A., Stuart Smith, R. D., Tribot, A. S., \u0026amp; Mouquet, N. (2022). The aesthetic value of reef fishes is globally mismatched to their conservation priorities. \u003cem\u003ePLOS Biology\u003c/em\u003e, \u003cem\u003e20\u003c/em\u003e(6), e3001640. https://doi.org/10.1371/JOURNAL.PBIO.3001640\u003c/li\u003e\n\u003cli\u003eLenda, M. L., Sk\u0026oacute;rka, P., Jaźwa, M., Lin, H.-Y., Nęcka, E., Tryjanowski, P., Moroń, D., Knops, J. M. H., \u0026amp; Possingham, H. P. (2023). Recognizing the importance of near-home contact with nature for mental well-being based on the COVID-19 lockdown experience. \u003cem\u003eEcology and Society\u003c/em\u003e, \u003cem\u003e28\u003c/em\u003e(3). https://doi.org/10.5751/ES-14374-280313\u003c/li\u003e\n\u003cli\u003eLeong, M., Dunn, R. R., \u0026amp; Trautwein, M. D. (2018). Biodiversity and socioeconomics in the city: a review of the luxury effect. \u003cem\u003eBiology Letters\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(5). https://doi.org/10.1098/RSBL.2018.0082\u003c/li\u003e\n\u003cli\u003eLi, X., Zhang, C., Li, W., Ricard, R., Meng, Q., \u0026amp; Zhang, W. (2015). Assessing street-level urban greenery using Google Street View and a modified green view index. \u003cem\u003eUrban Forestry and Urban Greening\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(3), 675\u0026ndash;685. https://doi.org/10.1016/j.ufug.2015.06.006\u003c/li\u003e\n\u003cli\u003eLwin, K. K., \u0026amp; Murayama, Y. (2011). Modelling of urban green space walkability: Eco-friendly walk score calculator. \u003cem\u003eComputers, Environment and Urban Systems\u003c/em\u003e, \u003cem\u003e35\u003c/em\u003e(5), 408\u0026ndash;420. https://doi.org/10.1016/J.COMPENVURBSYS.2011.05.002\u003c/li\u003e\n\u003cli\u003eMcGarigal, K. (2001). Landscape Pattern Metrics. \u003cem\u003eEncyclopedia of Environmetrics\u003c/em\u003e. https://doi.org/10.1002/9780470057339.VAL006\u003c/li\u003e\n\u003cli\u003ePearce, M. T., Zaidel, D. W., Vartanian, O., Skov, M., Leder, H., Chatterjee, A., \u0026amp; Nadal, M. (2016). Neuroaesthetics: The Cognitive Neuroscience of Aesthetic Experience. \u003cem\u003ePerspectives on Psychological Science\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e(2), 265\u0026ndash;279. https://doi.org/10.1177/1745691615621274/ASSET/IMAGES/LARGE/10.1177_1745691615621274-FIG1.JPEG\u003c/li\u003e\n\u003cli\u003ePelczynski, J., \u0026amp; Tomkowicz, B. (2019). Densification of cities as a method of sustainable development. \u003cem\u003eIOP Conference Series: Earth and Environmental Science\u003c/em\u003e, \u003cem\u003e362\u003c/em\u003e(1), 012106. https://doi.org/10.1088/1755-1315/362/1/012106\u003c/li\u003e\n\u003cli\u003ePugach, C., Leder, H., \u0026amp; Graham, D. J. (2017). How stable are human aesthetic preferences across the lifespan? \u003cem\u003eFrontiers in Human Neuroscience\u003c/em\u003e, \u003cem\u003e11\u003c/em\u003e, 246626. https://doi.org/10.3389/FNHUM.2017.00289/BIBTEX\u003c/li\u003e\n\u003cli\u003eRichardson, E. A., Pearce, J., Shortt, N. K., \u0026amp; Mitchell, R. (2017). The role of public and private natural space in children\u0026rsquo;s social, emotional and behavioural development in Scotland: A longitudinal study. \u003cem\u003eEnvironmental Research\u003c/em\u003e, \u003cem\u003e158\u003c/em\u003e, 729\u0026ndash;736. https://doi.org/10.1016/j.envres.2017.07.038\u003c/li\u003e\n\u003cli\u003eRossetti, T., Lobel, H., Rocco, V., \u0026amp; Hurtubia, R. (2019). Explaining subjective perceptions of public spaces as a function of the built environment: A massive data approach. \u003cem\u003eLandscape and Urban Planning\u003c/em\u003e, \u003cem\u003e181\u003c/em\u003e, 169\u0026ndash;178. https://doi.org/10.1016/J.LANDURBPLAN.2018.09.020\u003c/li\u003e\n\u003cli\u003eSabbion, P. (2018). Green Streets Social and Aesthetic Aspects. \u003cem\u003eNature Based Strategies for Urban and Building Sustainability\u003c/em\u003e, 283\u0026ndash;290. https://doi.org/10.1016/B978-0-12-812150-4.00026-4\u003c/li\u003e\n\u003cli\u003eSalmond, J. A., Tadaki, M., Vardoulakis, S., Arbuthnott, K., Coutts, A., Demuzere, M., Dirks, K. N., Heaviside, C., Lim, S., MacIntyre, H., McInnes, R. N., \u0026amp; Wheeler, B. W. (2016). Health and climate related ecosystem services provided by street trees in the urban environment. \u003cem\u003eEnvironmental Health : A Global Access Science Source\u003c/em\u003e, \u003cem\u003e15 Suppl 1\u003c/em\u003e(Suppl 1). https://doi.org/10.1186/S12940-016-0103-6\u003c/li\u003e\n\u003cli\u003eSandifer, P. A., Sutton-Grier, A. E., \u0026amp; Ward, B. P. (2015). Exploring connections among nature, biodiversity, ecosystem services, and human health and well-being: Opportunities to enhance health and biodiversity conservation. In \u003cem\u003eEcosystem Services\u003c/em\u003e (Vol. 12, pp. 1\u0026ndash;15). Elsevier B.V. https://doi.org/10.1016/j.ecoser.2014.12.007\u003c/li\u003e\n\u003cli\u003eS\u0026atilde;o Paulo, S. do V. e M. A. (SVMA). (2020). \u003cem\u003eMapeamento Digital da Cobertura Vegetal do Munic\u0026iacute;pio de S\u0026atilde;o Paulo. Relat\u0026oacute;rio Final\u003c/em\u003e. https://www.prefeitura.sp.gov.br/cidade/secretarias/upload/meio_ambiente/RelCobVeg2020_vFINAL_compressed(1).pdf\u003c/li\u003e\n\u003cli\u003eSemenza, J. C., March, T. L., \u0026amp; Bontempo, B. D. (2007). Community-initiated urban development: An ecological intervention. \u003cem\u003eJournal of Urban Health\u003c/em\u003e, \u003cem\u003e84\u003c/em\u003e(1), 8\u0026ndash;20. https://doi.org/10.1007/s11524-006-9124-8\u003c/li\u003e\n\u003cli\u003eSoga, M., \u0026amp; Gaston, K. J. (2016). Extinction of experience: The loss of human-nature interactions. In \u003cem\u003eFrontiers in Ecology and the Environment\u003c/em\u003e (Vol. 14, Issue 2, pp. 94\u0026ndash;101). Ecological Society of America. https://doi.org/10.1002/fee.1225\u003c/li\u003e\n\u003cli\u003eSuchocka, M., Wojnowska-Heciak, M., Błaszczyk, M., Gawłowska, A., Ciemniewska, J., Jarska, A., Heciak, J., \u0026amp; Pachnowska, B. (2022). Old trees are perceived as a valuable element of the municipal forest landscape. \u003cem\u003ePeerJ\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e. https://doi.org/10.7717/PEERJ.12700/SUPP-6\u003c/li\u003e\n\u003cli\u003eTeixeira, F. Z., Bachi, L., Blanco, J., Zimmermann, I., Welle, I., \u0026amp; Carvalho-Ribeiro, S. M. (2019). Perceived ecosystem services (ES) and ecosystem disservices (EDS) from trees: insights from three case studies in Brazil and France. \u003cem\u003eLandscape Ecology\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e(7). https://doi.org/10.1007/s10980-019-00778-y\u003c/li\u003e\n\u003cli\u003eWeber, R., Schnier, J., \u0026amp; Jacobsen, T. (2008). Aesthetics of Streetscapes: Influence of Fundamental Properties on Aesthetic Judgments of Urban Space. \u003cem\u003ePerceptual and Motor Skills\u003c/em\u003e, \u003cem\u003e106\u003c/em\u003e(1), 128\u0026ndash;146. https://doi.org/10.2466/PMS.106.1.128-146\u003c/li\u003e\n\u003cli\u003eWu, Z., Chen, R., Meadows, M. E., Sengupta, D., \u0026amp; Xu, D. (2019). Changing urban green spaces in Shanghai: trends, drivers and policy implications. \u003cem\u003eLand Use Policy\u003c/em\u003e, \u003cem\u003e87\u003c/em\u003e, 104080. https://doi.org/10.1016/J.LANDUSEPOL.2019.104080\u003c/li\u003e\n\u003cli\u003eXu, C., Han, B., Lu, F., \u0026amp; Wu, T. (2022). Assessing the Traffic Noise Reduction Effect of Roadside Green Space Using LiDAR Point Cloud Data in Shenzhen, China. \u003cem\u003eForests 2022, Vol. 13, Page 765\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(5), 765. https://doi.org/10.3390/F13050765\u003c/li\u003e\n\u003cli\u003eXue, F., Gou, Z., \u0026amp; Lau, S. S. Y. (2017). Green open space in high-dense Asian cities: Site configurations, microclimates and users\u0026rsquo; perceptions. \u003cem\u003eSustainable Cities and Society\u003c/em\u003e, \u003cem\u003e34\u003c/em\u003e, 114\u0026ndash;125. https://doi.org/10.1016/J.SCS.2017.06.014\u003c/li\u003e\n\u003cli\u003eZhan, J., Liu, M., Garrod, O. G. B., Daube, C., Ince, R. A. A., Jack, R. E., \u0026amp; Schyns, P. G. (2021). Modeling individual preferences reveals that face beauty is not universally perceived across cultures. \u003cem\u003eCurrent Biology\u003c/em\u003e, \u003cem\u003e31\u003c/em\u003e(10), 2243-2252.e6. https://doi.org/10.1016/J.CUB.2021.03.013\u003c/li\u003e\n\u003cli\u003eZhang, F., Zhou, B., Liu, L., Liu, Y., Fung, H. H., Lin, H., \u0026amp; Ratti, C. (2018). Measuring human perceptions of a large-scale urban region using machine learning. \u003cem\u003eLandscape and Urban Planning\u003c/em\u003e, \u003cem\u003e180\u003c/em\u003e, 148\u0026ndash;160. https://doi.org/10.1016/J.LANDURBPLAN.2018.08.020\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":true,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Urban Sustainability, Relational Values, Cultural Ecosystem Services, Landscape Heterogeneity, Environmental Justice, Urban Ecology, Megacities","lastPublishedDoi":"10.21203/rs.3.rs-5109482/v3","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5109482/v3","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eUrban streetscapes serve as essential public domains shaping human perception, sense of community identity and well-being. An integral aspect of this perception lies in the appreciation of the \"environmental scenic beauty\", reflecting individuals' personal comfort and sense of connection to their surroundings. Our study investigates the nuanced factors influencing the provision of urban aesthetic ecosystem services \u0026mdash; a key non-material benefit with implications for mental well-being and urban quality of life here framed as a Nature\u0026rsquo;s Contribution to People (NCP). Employing online surveys, deep learning analyses, and spatial modeling, we bridge ground-level perception with landscape-level features, exploring the intricate interplay between green and built areas in shaping aesthetic preferences in S\u0026atilde;o Paulo\u0026rsquo;s streets \u0026mdash; the largest megacity in the Southern Hemisphere and a highly diverse urban environment. We found that the perceived beauty of streets is positively affected by the heterogeneous arrangement of vegetation and built-up areas and by the three-dimensionality of trees \u0026mdash; and not solely by the quantity of greenery. Surprisingly, socioeconomic profiles of respondents exhibit no discernible impact on aesthetic evaluations, suggesting consensus across people with diverse social characteristics. Using convolutional neural networks trained on our survey, we predicted aesthetic scores for over 350,000 street images, yielding for the first time a map of the scenic beauty ecosystem service of an entire megacity. This aesthetic map uncovers significant mismatches between supply and demand for aesthetic services, exposing urban inequalities. By revealing these drivers and spatial patterns, our framework provides actionable insights for policymakers \u0026mdash; linking perception and landscape-level planning \u0026mdash; and offers a pathway to cultivate more socially equitable and aesthetically meaningful urban environments.\u003c/p\u003e","manuscriptTitle":"Bridging perception and landscape structure: mapping urban streetscape aesthetics as nature’s contributions to people","msid":"","msnumber":"","nonDraftVersions":[{"code":"","date":"2026-05-19 05:31:53","doi":"","editorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-19T05:31:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-14T23:50:47+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-14T07:19:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"166766505349401903613863274552447596844","date":"2026-05-13T14:44:31+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"253598969086158655925421741250135529764","date":"2026-05-13T08:14:02+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"96009563846808873367068915294795460920","date":"2026-05-12T06:31:11+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-11T03:06:16+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-25T14:28:55+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-04-24T18:54:09+00:00","index":"","fulltext":""},{"type":"notPreprinted","content":""}],"status":"timeline","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}},{"code":"","date":"2026-04-24 06:28:50","doi":"","editorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-04-24T06:28:50+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-14T00:50:50+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-13T10:12:56+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-09T15:28:22+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"96009563846808873367068915294795460920","date":"2026-04-07T07:31:05+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"253598969086158655925421741250135529764","date":"2026-04-05T16:40:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"166766505349401903613863274552447596844","date":"2026-04-04T14:22:34+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-03T12:19:14+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-25T13:32:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-03-24T19:23:54+00:00","index":"","fulltext":""},{"type":"notPreprinted","content":""}],"status":"timeline","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}},{"code":3,"date":"2025-11-11 17:21:34","doi":"10.21203/rs.3.rs-5109482/v3","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-02T10:10:28+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-19T15:33:39+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-17T22:06:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"166766505349401903613863274552447596844","date":"2026-02-09T04:51:19+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"253598969086158655925421741250135529764","date":"2026-02-08T17:40:10+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-21T07:48:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"96009563846808873367068915294795460920","date":"2025-11-12T06:12:50+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-09T18:03:36+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-11-09T18:00:50+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-10-28T18:22:05+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-27T12:55:02+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-10-27T12:51:32+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"bbacf4f9-bbb5-461e-b214-52d6e296382b","owner":[],"postedDate":"November 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":57734609,"name":"Earth and environmental sciences/Environmental social sciences"},{"id":57734610,"name":"Social science/Environmental studies"},{"id":57734611,"name":"Scientific community and society/Geography"},{"id":57734612,"name":"Social science/Geography"}],"tags":[],"updatedAt":"2026-05-21T22:38:09+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-11 17:21:34","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v3","identity":"rs-5109482","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5109482","identity":"rs-5109482","version":["v3"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}