Next Article in Journal
Multi-Source Data Integration for Sustainable Management Zone Delineation in Precision Agriculture
Previous Article in Journal
Social Nudging for Sustainable Electricity Use: Behavioral Interventions in Energy Conservation Policy
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Street Trees’ Obstruction of Retail Signage and Retail Rent: An Exploratory Scene Parsing Street View Analysis of Seoul’s Commercial Districts

Department of Urban Design and Planning, Hongik University, Seoul 04066, Republic of Korea
*
Author to whom correspondence should be addressed.
Sustainability 2025, 17(15), 6934; https://doi.org/10.3390/su17156934
Submission received: 10 June 2025 / Revised: 21 July 2025 / Accepted: 29 July 2025 / Published: 30 July 2025

Abstract

Urban greening initiatives, including the incorporation of street trees, have been widely recognized for a variety of environmental benefits. However, their economic impact on retail, in particular, the impact of street trees on the visibility of signs, has been underexplored. Street trees can obscure retail signs, potentially reducing customer engagement and discouraging retailers from paying higher rents for such locations. This paper investigates how the blocking of retail signage by street trees affects monthly rent in developed commercial districts in Seoul. It identifies, through Google Street View and state-of-the-art deep-learning-based semantic segmentation methods, environmental elements such as street trees, sidewalks, and buildings; quantifies their proportions; and analyzes their impact on rent using OLS regression, controlling for socio-economic variables. The results reveal that rents significantly diminish when street trees blocking views of retail signs increase. Our findings require more nuanced consideration by planners and policymakers in balancing both environmental and economic demands toward sustainable street design and planning.

1. Introduction

Incorporating greenery into retail spaces is becoming a crucial strategy to improve urban landscapes and make them more habitable and appealing as urban areas deal with the challenges of climate change [1,2]. However, retail greenery that utilizes plants, trees, and other natural elements in the layout and surroundings of retail stores demonstrates a conflict between the potential positive and negative effects of urban greening. In retail settings, the presence of greenery has been proven to improve consumer experiences [3], primarily because nature-inspired design features offer restorative benefits that alleviate stress and boost mood [4,5]. This enhancement in mood subsequently affects consumer behavior, since a better store environment has been associated with an increased chance of buying [6,7]. For example, the presence of plants at a mall promoted more social contacts in addition to increasing the possibility that customers would study things more closely [8]. Furthermore, the inclusion of trees significantly increased consumer preference ratings for photographs in a number of studies by Wolf [9,10,11,12]. Wolf’s research shows that environments having neat sidewalks and good-quality structures, but lacking trees, were rated poorly, whereas photos showcasing healthy, large trees—even if they covered other features like historic buildings—garnered the best ratings. Collectively, these results illustrate the substantial favorable influence of greenery on consumer experiences, bolstered by empirical studies and psychological advantages linked to nature in retail settings.
While urban greening efforts, including the incorporation of plants and trees, are generally recognized for enhancing consumer experiences, retail business owners may not always support having street trees positioned directly outside their shops. This stems from worries that street trees could block retail signs and it thus negatively impacts sales. Nasar [13] emphasizes this point, stating that “Using signs, each shop owner aiming to attract attention to their business pursues a unique sign to convey a favorable image and differentiate from the environment.” Every individual sign might convey a positive impression and draw interest, yet numerous such signs positioned next to each other could result in confusion. Although Nasar [13] emphasizes the necessity of managing urban signage in this depiction, it also indicates that retailers worry about their signs being obstructed. Wolf [9] supports retailers’ apprehensions through a survey analysis, showing that retailers place considerably less value on the advantages of street trees than consumers do and show a diminished preference for their existence. She determines that retailers give priority to the short-term effects of street trees, like blocked visibility of signs and storefronts, along with possible maintenance expenses due to tree debris, rather than the long-term advantages they offer, such as shade, better air quality, and their favorable influence on consumer emotions and actions.
These visibility issues are especially important when considered from an economic impact perspective, as sustainability entails reconciling conflicts among economic growth, environmental conservation, and fairness [14]. The visibility of signs is essential for luring customers to retail stores [13], and visual interference from street trees may decrease sales. Given that retailers operating small businesses often lack the resources for broad advertising campaigns, signage is a critical marketing tool that can substantially influence business success [13]. This indicates that retailers may be reluctant to pay higher rents for locations with street trees that obstruct their signs’ visibility, despite the broader benefits. While street trees have been well recognized for their roles in creating the aesthetic appeal of an area and driving higher property values [15,16,17], there is still a relative lack of research into street trees’ impact on retail rents via the view-obstruction of signs. In retail settings, vegetation affects consumption behavior as it improves ambiance and thus increases sales [3,4,5,6,7]. This suggests that retail areas with street trees could become more attractive for retail business owners, which, as a result, could lead to potentially higher demand and increasing rents. Conversely, the presence of street trees may obstruct retail signs, which are essential for attracting customers in commercial areas where visibility directly impacts traffic and sales [13]. Since signs represent the first point of contact between retailers and potential customers, their obstruction by street trees could impede customer engagement and reduce opportunities to attract potential customers. This potential decrease in visibility could discourage retailers from selecting locations with substantial tree cover. Additionally, while retailers usually favor working in green-certified structures, their readiness to lease declines if certification results in increased rental costs [18]. Likewise, any negative externalities associated with impaired visibility or additional upkeep requirements and expenses attributed to street trees would lead to decreased rents or even the renegotiation of rental contracts between property owners and retail business tenants. It could be questioned if the aesthetic value and perceived gains made by consumers outweigh the disadvantages brought by street trees such as obstructing storefronts and signage [3].
This paper fills this gap by leveraging an increasingly popular data source in urban studies, street view image data, which provides a fine-grained view of the streetscape from a pedestrian’s perspective. Studies by Kang et al. [19], Biljecki and Ito [20], and Yue et al. [21] are some examples. Our research investigates how the presence of street trees that visually block retail signs affects the monthly rent prices for retailers. Advances in deep-learning-based image processing techniques enable the extraction of large-scale environmental features from the image data, such as artificial structures, natural elements, and pedestrian densities [20,21]. By using street view imagery and a deep learning tool to process it, we analyze how the view-obstruction of retail signs by street trees negatively affects monthly rent. This would, in fact, be a very valuable piece of information for urban planners and policymakers in relation to street trees and their commercial viability. The question at hand emphasizes the need for an increasingly sensitive integration of principles into planning and design to further urban sustainability. Our findings indicate that the obstruction of retail signs by street trees significantly reduces retailers’ rents, and these findings suggest that the existing urban greening strategies should be reviewed to capture unexplored economic and environmental effects.

2. Data and Methods

2.1. Study Area

Retail business represents a large proportion not only of the economy in Seoul but also in the greater national economy of South Korea. The proportion of retail businesses is sixth among member countries in the OECD [22], showing how this industry has assumed great importance within the nation’s economic perspective. The role of retail businesses is more prominent in Seoul. Small businesses are the dominant ones at 69.8% of all enterprises in the city [23]. Among those small businesses, 45% are retail, accommodation, and food service industries that sell goods or services directly to consumers [24]. What is more, retail contributes 1,203,088 jobs out of the total in Seoul, a quarter of employment, as shown by Lee [24]. Thus, this evidence points out that, since there is such significant activity in this sector, this area contributes not just to the Seoul economy but to the state in general, too; Seoul contributes to the GDP by over fifty percent [25].
However, the survival rate of retail businesses remains fragile. According to 2023 data from the Seoul City Commercial District Analysis Service (SCCDAS) [26], 81.3% of retail businesses in Seoul remain operational 1 year after their establishment, 64.3% continue to operate after three3 years, and only 48.5% are still in business 5 years following their opening. The fact that about half of these businesses close after 5 years is not merely an issue for the business owners and their enterprises but can have widespread effects on the broader economy of Seoul. In particular, since self-employment absorbs unemployed individuals and those disadvantaged or underrepresented in the labor market [27,28,29], the closure of retail businesses not only leads to higher unemployment rates but can also negatively impact the local economy as unemployed workers lose their purchasing power [29]. Additionally, a significant number of retail business owners finance their initial operations through loans [24,30], so the high failure rate increases the likelihood of bad loans. This, in turn, can undermine the profitability of financial institutions and challenge the stability of the financial system [24,30]. Within the regional economy, retail businesses also play a role beyond simply connecting consumers and goods; they also facilitate economic circulation [29,31]. The increasing closure of small businesses and decline of local markets can have negative ripple effects on other nearby businesses, including producers, logistics providers, and service industries.

2.2. Data

This study utilized 2023 statistical data from the SCCDAS, which is produced annually by the Seoul Metropolitan Government. The city government established the SCCDAS to support micro enterprise that involves retail business. The SCCDAS provides detailed, data-driven insights into commercial districts, making this information accessible to not only current and prospective small business owners but also researchers [32].
The city government categorizes commercial districts into four types—alley commercial districts (neighborhood-scale districts), developed commercial districts, traditional market commercial districts, and tourist special zone commercial districts—but this study focuses exclusively on developed commercial districts to ensure the most controlled environment for isolating the impact of street trees on retail signs. Under the Distribution Industry Development Act in South Korea [33], an area is classified as a commercial street if it contains 50 or more stores within a 2000-square-meter radius. A developed commercial district is defined as a cluster of these commercial streets, densely packed and accessible primarily by foot [26]. Initial mapping of developed commercial districts began in 2009 with a 50-by-50-meter grid, later updated in 2012 to a block-based system that encompasses approximately 360,000 blocks as defined by the Korean Statistical Office, integrating factors such as enumeration districts and road networks [26]. In 2016, the Ministry of SMEs and Startups—a governmental agency supporting small and medium-sized enterprises in South Korea—refined this classification to 249 distinct commercial areas, removing overlaps with traditional markets [26]. Our analysis includes these 249 developed commercial districts, which have a total area of 49.65 km2, and an average size of approximately 0.2 km2 (20 hectares). In contrast, alley commercial districts cover 100.33 km2, representing 27% of Seoul’s urbanized area of 371.54 km2 (Figure 1). This area is quite extensive and often includes mixed land uses and residential zones, which may dilute the impact of view-obstruction by street trees on retail rent price due to the broader zoning diversity. Traditional market districts, meanwhile, often lack street trees altogether, making them unsuitable for assessing this study’s focus. Likewise, tourist-specialized zones cater to unique functions that create dynamics distinct from typical retail environments, such as language barriers limiting tourist engagement with local signage. These variances could distort the findings and hinder a precise assessment of street trees’ influence on retail signs. Therefore, We created a dataset from sources at the SCCDAS, Statistics Korea, Real Estate Big Data Platform, and the Korea Credit Guarantee Fund in order to analyze the effect of visual obstruction by street trees on retailers’ monthly rent (Table 1). All of these sources are Korean government agencies. The publicly available SCCDAS data include variables such as the total number of stores within the commercial district, annual rates of store openings and closures, total store revenue, total foot traffic, income group scores reflecting minimum and maximum income levels of district residents, total residential and workforce populations, and 5-year store survival rates. Considering that local Gross Regional Domestic Product (GRDP) has a significant effect on the revenues of commercial districts [34,35], we further supplemented the dataset with GRDP data from Statistics Korea, the official census bureau in Korea. Additionally, we obtained Point of Interest (POI) data—which provides location information on educational, medical, retail, transportation, and neighborhood facilities—from the Real Estate Big Data Platform operated by the Ministry of Land, Infrastructure, and Transport, given that POI density enhances the urban vitality of commercial areas [36]. In contrast to these open-source datasets, average monthly rent data were obtained from the Korea Credit Guarantee Fund, requiring specific access permissions as it is not publicly available.

2.3. Method—Street View Analysis

To capture the effect of the view-obstruction of retail signs by street trees, we analyzed Google Street View (GSV) images taken within the boundaries of developed commercial districts and measured various physical environmental factors, such as street trees, signs, buildings, and sidewalks. As shown in Figure 2, we extracted GSV images at 100 m intervals along the street network in the developed commercial districts as of 2018, given that the 100 m boundary represents a human-scale distance within which people feel more engaged with their surroundings [37]. However, it is noted that GSV images are taken by cameras mounted on vehicles that travel along the streets, which means that some streets may not have images available for analysis if the street view cars are unable to navigate them due to their narrow widths. The city of Seoul utilizes deciduous trees as street trees, which are affected by seasonal changes; therefore, images from December to February, when the leaves have fallen and only branches remain, were excluded from the analysis. GSV images are stored as panoramic images, which visually distort some of the environmental factors captured at the edges [38]. As a result, sections with significant distortion outside the human field of view were removed (Figure 3a). Specifically, 270° on each side, 80° from the top, and 70° from the bottom—corresponding to a pedestrian’s natural eye level—were cropped out [39,40]. Additionally, due to the structure of human vision, an extra 10° was cropped from the top compared to the bottom [40].
After we extracted 5504 GSV images, the built environment factors on the cropped images were categorized through the semantic segmentation method. We employed the ADE20K (annotated dataset for a type of computer vision task aimed at understanding the entire structure of a scene by segmenting and identifying all the objects) as the foundation for this segmentation task, as it provides pixel-level annotations across 150 object categories relevant to urban and natural environments. ADE20K’s diversity in both indoor and outdoor scenes, including common urban features like roads, buildings, vehicles, and vegetation, aligns well with the elements typically observed in GSV images. Based on this dataset, we applied a deep learning model pre-trained on ADE20K, specifically the Swin Transformer Large (Swin-L), to leverage its effectiveness in capturing both local and global context within scenes.
Swin-L, like other Swin Transformers, uses a unique approach called the shifted window strategy that substantially enhances the efficiency and scalability of the deep learning model [41]. Instead of processing the entire image globally, the Swin Transformer divides the image into smaller windows and processes them hierarchically (Figure 4). The shifted part of the term means that it changes the windows’ positions across layers to capture relationships between neighboring regions [42]. This approach helps the model capture both local details and long-range dependencies, which is valuable for complex scenes like those in ADE20K [41]. In other words, efficiency is achieved by restricting attention to localized windows, reducing computational demands, while the scalability arises from the shifted windows that extend the receptive field without incurring prohibitive memory or computation costs. Together, these innovations make Swin Transformers well-suited for applications requiring high-resolution processing, such as image and video analysis, where both computational efficiency and scalability are essential. We used Swin-L in conjunction with OneFormer, as this combination enables detailed semantic and instance segmentation. This makes it easier to identify finer elements in the original image, such as small objects like road cracks, signs, and streetlights [43]. Swin models exhibit better performance features compared to the traditional models of Convolutional Neural Networks [41], and the combination with OneFormer shows better performance given Panoptic Quality, Average Precision, and Mean Intersection over Union [43].
Before running the model on our GSV images, each image was preprocessed by resizing it to 512 × 512 to align with the standard input size and normalized to ensure consistency. During the inference stage, the model segmented each image into discrete categories, identifying built environment features in a systematic and reproducible manner. This semantic segmentation enabled us to classify and quantify relevant built environment factors accurately. However, upon reviewing the segmented images, we observed that some had significantly incorrect segmentations, and thereby we removed 197 images that were excluded from the raw dataset of 5504 images to prevent them from introducing inaccuracies in the analysis. This left a final sample of 5307 images for developed commercial districts. Similarly, for comparison with all commercial districts, we processed 17,213 images. Using the correctly segmented images, proportions for each built environment factor such as sidewalks, signs, street trees, etc., were calculated through Python 3.10.12. These proportions were derived from the segmentation outputs and provide quantitative measures for regression analysis on the relationship between built environment factors and monthly rent of developed commercial districts in Seoul.

2.4. Method—Analytical Method to Investigate Monthly Rent

We model linear relationships between monthly rent and various built environment factors to examine the impact of signs’ visual obstruction. Two sets of predictor variables are employed. The first set represents indicators of economic activity within developed commercial districts. Variables such as total shop count, the proportion of annual shop openings, and revenue capture business vibrancy, reflecting the economic dynamism within these areas [26]. To account for income levels, which can significantly influence commercial district vibrancy [44], we include two income indicators: (1) the income level of residents within the commercial districts and (2) the Gross Regional Domestic Product (GRDP) per capita at the ‘gu’ level. The ‘gu’ level, a broader administrative division encompassing the commercial districts, provides a wider socio-economic context (Seoul comprises 25 districts, each with an average population of approximately 380,000 as of 2024). Additionally, in line with findings that factors such as foot traffic, 5-year shop survival rate, daytime population, and POI data substantially impact commercial vibrancy [36,40,44], we incorporate these three variables within the economic activity set. Together, these variables enable a comprehensive analysis of the economic environment’s influence on rental prices.
The built environment factors include the visual proportions of street trees, signs, sidewalks, sky, roads, and building structures as observed in a GSV image. To investigate the relationship between the visual obstruction of retail signs by street trees and retail rents, we used the visual proportion of street trees in street view images as a proxy for the degree of view-obstruction. While this variable may not directly measure the obstruction of retail signs, we justified this approach for the following reasons. First, analysis of retail signs’ visual proportions in street view imagery revealed that they ranged narrowly between 0 and 0.0239, which indicates that retail signs are heavily occluded, predominantly by street trees and other urban elements, in the observed dataset. Second, a negative and statistically significant correlation (−0.322, p < 0.01) was observed between the visual proportions of street trees and retail signs, suggesting that higher visual proportions of street trees are associated with reduced visibility of retail signs. Also, to empirically validate the assumption that the ‘Street Tree View’ proportion serves as a reliable proxy for signage obstruction, we conducted a manual audit on a random subset of 63 images from our final dataset, for which the sample size of 63 was randomly chosen from a range of 50 to 100. A trained researcher, who was blind to the computer-generated ‘Street Tree View’ values, coded each image based on a binary classification: ‘Yes’ if trees were judged to be causing significant visual obstruction of retail signage, and ‘No’ if there was minimal or no such obstruction. The results of this audit provide strong quantitative support for our assumption. Of the 63 sample images, 33 were classified as ‘Yes’ (obstruction present), and the remaining 30 were classified as ‘No’ (obstruction not present). The group of images coded as ‘Yes’ (obstruction present) exhibited an average ‘Street Tree View’ proportion of 19.6%. In stark contrast, images coded as ‘No’ (obstruction not present) had an average proportion of only 4.3%. An independent samples t-test confirmed that this large difference between the two groups is statistically significant (p < 0.001). This validation process confirms that our ‘Street Tree View’ variable is a robust proxy for the degree of signage obstruction by street trees, thereby strengthening the foundation of our subsequent regression analysis.
Lastly, the semantic segmentation techniques applied to detect objects such as signs and trees in street view images are not perfectly accurate, especially in urban settings with complex overlapping elements. Despite our efforts to review the segmentation output for misclassifications, small and incomplete detection persisted (Figure 3b). Given the limited range of retail sign visual proportions (0 to 0.0239), directly measuring the obstruction of retail signs remains unreliable, whereas the visual proportion of street trees is more consistently measurable and interpretable. For these reasons, we conclude that the visual proportion of street trees can serve as a practical indicator of potential sign obstruction.
The variables of sidewalks, sky, roads, and building structures all contribute to the sense of enclosure. Spreiregen [45] and Kim and Kim [46] demonstrate that people feel more stable when there is a balanced sense of enclosure. In retail environments, spaces that balance openness—such as access to natural light and views of the sky—with a sense of security and containment are likely to be perceived as more attractive to both consumers and business owners. Given that an enhanced psychological state can positively influence consumer behavior [6,7], a strong sense of enclosure may foster a sense of community or exclusivity, which can attract higher foot traffic and, subsequently, increase sales potential. Therefore, we posit that retail owners’ willingness to pay for rent is influenced by the interplay between architectural features that create a sense of enclosure and the psychological benefits they provide to consumers.
The relationship between the dependent variable, namely the average monthly rent in a developed commercial district, and the specified independent variables was evaluated by applying a multiple linear regression model through ordinary least squares (OLS) estimation. The model was organized based on two sets of variables representing economic activity and built environment factors, allowing for a comprehensive examination of the factors influencing commercial rent. We conducted the regression analysis using R, with support from the ‘lmtest’ and ‘car’ packages to manage diagnostics and assumptions. To improve the model’s accuracy and ensure robust results, a backward stepwise procedure was employed to refine the set of independent variables. In this procedure, the full model, including all candidate predictors, was initially fit to the data. Variables were then sequentially removed based on their statistical insignificance, typically determined by p-values above a specified threshold (commonly 0.05). In each step, the variable with the highest p-value was eliminated first, assuming it added the least explanatory power to the model. This iterative process continued until all remaining predictors showed statistically significant contributions to the model. Throughout this procedure, multicollinearity diagnostics were also monitored to ensure that highly correlated predictors did not distort the model’s estimates. Removing any variable having a high variance inflation factor (VIF) helped maintain the independence of predictors, further strengthening the model’s reliability. The final model, thus, included only those variables that had significant explanatory value, improving interpretability and ensuring that the model was both parsimonious and robust.
Besides the model of developed commercial districts, we consider a model that includes all the commercial districts in Seoul as well to set our findings in a wider contextual framework. Such a comparison enables us to see whether the effect of view-obstruction of signs on rent is context-dependent in developed commercial districts. Highly developed commercial districts are characterized by high retail density, heavy pedestrian traffic, and intense reliance on visibility to ensure profitability of business enterprises. These characteristics make retail rents particularly sensitive to visual obstructions, such as street trees blocking signs. In contrast, the whole-districts model involves a wider range of commercial contexts, including areas having mixed land uses and varied economic circumstances. These settings have a mix of other characteristics that would weaken or even turn around the impact of visual obstruction on retail rents, because, in mixed or mainly residential settings, the overall street tree benefits of air quality improvements, shade, mitigation of the urban heat, and aesthetics at a better neighborhood level typically outweigh any negative impacts associated with the view-obstruction effect on retail rents. This greenery can make the environment more appealing to residents, who are a considerable percentage of the population in these districts, and can thus raise residential and commercial rents alike. Since all commercial districts taken together account for about 44% of the urbanized area in Seoul, any positive amenities from street trees in these broader settings would likely outweigh concerns about the visibility of signs. This dynamic provides some explanation for the diminished or even reversed relationship between the visual obstruction caused by street trees and rent prices seen in the whole-districts model.

3. Results

As shown in Table 2, the initial model for the developed commercial districts produced a Multiple R squared of 0.3945, and it indicates that roughly 39.5% of the variance in rent was explained by the included predictors. Significant variables in the 95% confidence level at this stage included both the economic activity variable set and the physical environment variable set, each contributing meaningfully to the model’s explanatory power. However, several other predictors—such as revenue, income, and sidewalks—were not statistically significant and did not contribute meaningfully to the model. To refine the model, a backward stepwise approach was used to sequentially remove non-significant predictors, optimizing for model parsimony and reliability. As seven variables were excluded (revenue, income_max, poi, pop_travel, pop_work, surv_rate5, and sidewalk), the refined model included only those predictors that demonstrated statistical significance and meaningful associations with rent levels. Due to the redistribution of explained variance and reduced multicollinearity, coefficients for the remaining predictors are enlarged as they account for both their own effects and some residual shared effects that were initially distributed among all variables. The refined model demonstrated an improved fit, with an F-statistic of 20.99 and a p-value < 2.2 × 10−16, indicating strong overall significance. The Multiple R squared for the refined model was 0.3847, suggesting that approximately 38.5% of the variance in rent is explained by these selected predictors. To validate the OLS model’s assumptions, diagnostic tests were conducted to assess the homoscedasticity of the residuals and multicollinearity among the predictor variables. The Breusch–Pagan Test and the Non-Constant Variance Score Test both indicated no heteroscedasticity, with p-values of 0.4304 and 0.98277, respectively. These results suggest homogeneity in the residual variance. Multicollinearity was also assessed using the VIF. The VIF values measure the degree of multicollinearity for each predictor. A VIF greater than 10 is typically considered indicative of problematic multicollinearity, as it suggests that a predictor is highly correlated with one or more other predictors in the model. All the VIF values in the refined model were below 5, indicating an acceptable level of multicollinearity among the predictors.
In the set of economic activity variables, the number of shops in a developed commercial district (shop_cnt_y), the annual shop opening rate of a district (shop_open_y), and the per capita gross regional domestic product (grdp_cap) are all positively correlated with the district’s average monthly rent. On the other hand, when examining the physical environmental factors, the visual proportions of street trees (Street Tree View), sky (sky), and buildings (structure) in street view images are negatively correlated with the dependent variable, while the proportion of signs in street view shows a positive correlation that is not statistically significant.
By comparison with the whole commercial districts regression results (Table 3), the developed commercial districts model presents significant differences in a few variables concerning significance and magnitude, though there are some similarities. The positive and highly significant relationship of the number of retail shops (shop_cnt_y) with average monthly rent is consistent in both models, and it emphasizes the role of retail density in driving rental values. Second, gross regional domestic product per capita (grdp_cap) shows significance and a positive relationship in both models and hence also reflects a specific key economic driver of rent level determination. However, the remaining variables showcase different patterns across the two models. For example, the visual proportion of street trees (Street Tree View) has a different effect. It is negatively related to rents in developed districts, where its view-obstruction effects may dominate over its aesthetic benefits, while in the whole-districts model, its significance in the reduced model suggests that its positive environmental and aesthetic contributions play a more prominent role. While the other variables in both contexts bear consistent signs, the impact of the visual proportion of sky (sky) has a stronger effect in the developed districts, possibly reflecting preferences for more enclosed environments in areas with high retail density. Variables of shop opening rate of a district (shop_open_y) and the total revenue of a district (revenue) also exhibit differences in significance. They are more robustly significant in the developed-districts model, reflecting the greater influence of commercial activity on rental values in these areas. In contrast, in the whole-districts model, the impact of commercial activity is diluted by the inclusion of mixed-use or predominantly residential areas, where non-commercial factors may play a larger role in rent determination.

4. Discussion

4.1. Economic Factors and Monthly Rent

The findings from this regression analysis highlight a critical gap in the data provided by the SCCDAS for understanding the driving factors behind commercial rent in Seoul’s retail districts. The SCCDAS aims to support current and potential retail business owners through its data, but the analysis reveals that many of its key indicators, such as local resident income, foot traffic, and 5-year retail shop survival rates, lack significant explanatory power for the variation in monthly rents. Among the variables analyzed, the number of retail shops in a developed commercial district stands out as the only SCCDAS-provided metric with significant explanatory power. GRDP, sourced from the Korean Census, exhibits robust statistical significance at the 99% confidence level and plays a pivotal role in determining monthly rent. These findings raise important questions about the adequacy of the data that the SCCDAS provides aimed at supporting retail business owners. Given the average retail shop size of 60.2 m2, the average monthly rent in Seoul’s developed commercial districts is 4,500,000 won [47]. This rent constitutes approximately 16% of the average monthly sales revenue of 27,870,000 won. Choi [48] conducted a simulation using an expected income model, which assumed that earned income, such as wages from employment, is converted into business income generated through retail operations. The findings indicated that for retail business owners in the food service industry to achieve financial viability, the rent for commercial spaces in Seoul must decrease by a minimum of 2.2% and up to 15%, depending on specific conditions. High rents are a defining feature of Seoul’s retail business [48], and this significantly affects the sustainability and profitability of retail businesses, which constitute a crucial part of the city’s economic fabric. Yet, the lack of robust data on rent-driving factors or business performance metrics beyond basic survival rates or revenue limits the ability of potential and current business owners to make informed decisions. For instance, more detailed information about the determinants of rent or the relationship between revenue and long-term profitability could shed more light on the financial dynamics involved. Moreover, the SCCDAS might consider integrating datasets having proven explanatory power, such as GRDP, or even conduct specific studies to investigate these critical relationships further. Providing a more comprehensive dataset and actionable insights based on research would empower retail business owners to navigate the complex economic environment of Seoul’s commercial districts more effectively. This would hopefully mitigate the negative impact of soaring rents on retailers and contribute toward creating a more sustainable retail ecosystem for the benefit of business communities and the wider economy of the city. In this regard, the SCCDAS may become more influential if it can supply data but also develop further analyses and recommendations that address the challenges Seoul’s retail business owners are facing.

4.2. A Tension in Sustainability

Our empirical analysis, which shows a direct link between the visual proportion of street trees and decreased retail rents, underscores the practical consequences of street greening that urban planners often overlook. This study illustrates how urban greening can create unintended economic challenges for retail businesses, even though it contributes to broader environmental goals. Although greenery in commercial districts provides significant psychological and behavioral benefits to consumers [3,4,5,12], by improving mood, reducing stress, fostering social interaction, and enhancing the overall shopping experience, the visibility issues caused by trees could make retail locations with significant tree cover less desirable for business owners. As Wolf [9] indicates, retailers often value the direct and immediate impacts, such as visibility and rent, more than the broader benefits that include shade or improved air quality, and this could make them less inclined to embrace street trees or other forms of greenery. This dual impact for both customers and retail business owners reminds planners of the inherent tension within sustainability between economic development and environmental protection.
The conflicting views of consumers and retail business owners on the benefits of street trees show how difficult it is for urban planners to achieve sustainability in planning and design. Urban greening initiatives are well recognized for their positive contributions to environmental protection, climate resilience, and social well-being. Strategies like tree planting and greening may help reduce the urban heat island effect, increase biodiversity, and create a healthier and more inclusive environment. The result is in harmony with better living quality for residents while ecological balance is achieved for sustainable urban development. With respect to the economic aspects of sustainability, any perceived barrier to brand visibility can also impact the business case for retail. This trade-off between achieving consumer welfare and economic profitability puts planners in a difficult position when seeking to balance environmental objectives with the economic dynamics of urban space. This tension necessitates a delicate approach to the planning of urban spaces that integrates both ecological and economic considerations.
Recent sustainable street designs aim to develop inclusive, safe, and sustainable streets that lessen reliance on cars [49]. These layouts focus on serving all types of users—pedestrians, cyclists, transit passengers, and motorists—while integrating green features to enhance visual appeal and offer ecological advantages such as improved air and water quality [49,50]. While these principles promote a holistic approach to sustainable urban spaces, they do not provide detailed guidance on how to sustain the economic activity of local businesses. For instance, the City of Los Angeles Complete Streets design guide [50] suggests smaller trees for commercial streets due to narrow sidewalk width and limited root zones next to multistory buildings. Yet it does not consider that tree placement would be blocking the stores’ signs and thus influencing visibility and the impact on their businesses. Urban planners might work out flexible tree-planting policies in order to balance environmental with economic needs. To maintain visibility, trees might be intentionally positioned so their canopies do not block sightlines. Additionally, the business owners’ involvement in the design process would allow planning goals to become more concomitant with the economic considerations necessary to provide a successful retail establishment. This would provide sustainable streets that could support both ecological concerns and successful economic performance in a commercial district.

4.3. Making Sense of Enclosure

These findings give significant insights into the role of the physical environment, inclusive of how the sense of enclosure influences economic outcomes in urban space. Negative significant relationships of the proportion of visible sky, road, and structure were found with monthly rent. In particular, higher shares of sky and road were associated with lower rents, β = −3981, p < 0.01, and β = −1939, p = 0.05, respectively, while higher shares of structure also had a negative association, β = −2067, p < 0.05. These results again point to the fact that a higher degree of enclosure, usually balanced among buildings and open space, might produce positive impacts on rent and consequently affect perceived space, the pedestrian experience, and economic activities.
The proportion of visible sky reflects spatial openness and influences pedestrian comfort and sense of place [51]. Densely built environments, with reduced sky visibility, are often vibrant urban centers that attract retail businesses due to high foot traffic, despite potential discomfort from enclosure. Similarly, the negative relationship between structure proportion and rent indicates that denser areas with high visibility of buildings are associated with higher rents, driven by their appeal to retailers seeking prime locations. Conversely, road-dominant street views may undermine pedestrian-friendly environments, detracting from the walkability that is a critical factor in retail success [44]. Excessive road visibility signals vehicle-oriented spaces, discouraging foot traffic and consumer spending. To enhance commercial district appeal, balancing structural density, road visibility, and sky openness is crucial. Overcrowded areas with inadequate openness can reduce attractiveness, while well-balanced environments foster pedestrian comfort, urban vibrancy, and higher rents.
In addition to the results of street trees’ proportion, these findings could serve as a cornerstone for advancing design strategies that enhance the sustainability of commercial districts. Efforts to optimize the sense of enclosure, by striking the right balance between visible sky, road, and structure, and strategically adjusting street trees’ locations to avoid obstructing store signs, can create spaces that are both environmentally sustainable and economically viable. However, achieving this balance requires intentional and nuanced design efforts that integrate ecological goals with the economic needs of retailers and consumers alike. By refining urban design strategies, planners can foster environments that improve the customer experience while addressing retailers’ concerns about visibility and accessibility. Along with using tree placements strategies that do not compromise the economic activity of businesses, planners also need to consider incorporating design elements that promote walkability and encourage longer customer visits, such as shaded pathways, seating areas, and aesthetically pleasing green features, which can contribute to vibrant, attractive commercial districts. This approach puts the accent on collaborative, adaptive design as the proper way to realize vivid commercial districts that will have a positive effect on all stakeholders. Taking into consideration a customer-centric perspective while solving the problems of retailers and maintaining ecological balance, an urban planner will create an area that will be not only beautiful and ecologically friendly but economically successful. Such design efforts represent the way forward in creating dynamic, thriving commercial areas that are in line with the principles of sustainable urban development. Our research provides new insights that add depth and richness to these balanced design efforts, pointing the way toward more sustainable urban spaces.

4.4. Limitations and Future Directions

While providing valuable insights, this study has a limitation that opens avenues for future research. The analysis treats the retail stores within a developed commercial district as a homogeneous group. However, the degree of reliance on physical signage as a marketing tool likely differs significantly between independent, self-employed retailers and large corporate franchises. In other words, small and independent businesses often depend heavily on the visibility of their storefronts and signs to attract foot traffic and build a local customer base because of their relatively limited resources. In contrast, major franchises can leverage national brand recognition and extensive marketing campaigns, making them less reliant on the visibility of a single location’s sign. Considering this, the negative economic impact of sign obstruction by street trees might be disproportionately felt by small business owners. A more severe impact on these smaller retailers could translate into greater downward pressure on their willingness to pay rent. However, the aggregated district-level data available for this study did not allow for a distinction between individual store types, such as independent and franchise. The Seoul City Commercial District Analysis Service (SCCDAS) does not provide point-level data for individual stores. Therefore, future research using store-level data would be highly valuable. Such an analysis could empirically test the hypothesis that the negative effects of sign obstruction are more pronounced for small and independent retailers.
Additionally, this study did not test for potential spatial autocorrelation among the commercial districts, a notable limitation given the spatial nature of the dataset. As illustrated in Figure 1 and Figure 2, the developed commercial districts are both geographically dispersed and contiguous, which makes constructing a methodologically sound spatial weights matrix challenging. Future research should aim to develop an appropriate spatial weights matrix for these unique spatial units and employ spatial regression models to produce more robust estimates that account for spatial dependencies.

5. Conclusions

This study provides critical insights into the relationship between urban greening and commercial rent dynamics in Seoul’s retail districts. By utilizing advanced image processing techniques on street view data, the analysis uncovered that the visual obstruction of retail signs by street trees and the physical environment factors have a significant negative effect on monthly rents. These results highlight the importance of reconciling environmental objectives with economic considerations in urban planning and design, particularly in commercial areas where visibility could be a crucial factor in business operations. The results also demonstrate the broader implications of spatial design, such as enclosure, in affecting economic activities. A balance between openness and enclosure, mediated by the proportions of visible sky, sidewalks, and buildings, would appear to create environments both more appealing to pedestrians and propitious for higher rental values. Too much of one or the other might discourage pedestrian traffic or drain the life from urban space, underscoring the need for subtle design tactics that balance aesthetic concerns with economic return. In addition, the findings show the insufficiencies in the existing datasets organized by the SCCDAS. While GRDP proved to be a sound proxy for rental prices, other variables such as local income and pedestrian traffic from the SCCDAS demonstrated little explanatory power. The result brings into question the adequacy of the data that currently exists to properly aid retail business owners. Therefore, expanding datasets and research to include more detailed variables like the economic benefits of visibility or the relationships between revenue and rent costs may yield more valuable information for stakeholders. This would better position business owners and policymakers to address the challenges created by rising rents and complex urban dynamics. Moreover, this study contributes to the ongoing conversation about the tensions within sustainability. Although urban greening provides substantial environmental and social advantages, including the reduction of urban heat and enhancement of quality of life, these initiatives may unintentionally pose economic difficulties for businesses. This situation illustrates a more extensive dilemma in urban planning: the necessity to reconcile environmental objectives with the economic conditions prevalent in commercial areas. For example, adjusting street tree location or size to reduce their impact on the visibility of retail signs can help resolve these issues without sacrificing environmental aims. The message from this research is the need for integrated, evidence-based approaches to urban design that take into account the difficult dynamics of commercial precincts. By integrating environmental sustainability with a thorough assessment of economic factors, urban planners can develop areas that promote thriving retail ecosystems and, at the same time, contribute to the attainment of broader goals in sustainable urban development. Such findings enhance our understanding of the economic impacts associated with urban greening and set a foundation for more informed and just urban planning approaches in the future. However, we emphasize that these conclusions are context-specific as the analysis is limited to the developed commercial districts in Seoul, which indicates that the economic tension arising from the obstruction of retail signs by street trees may not apply in other urban contexts. For instance, this tension may be diminished or even reversed in residential areas or other cities where the environmental and aesthetic benefits of urban greening might be valued in different ways. This underscores that urban greening strategies, aimed at achieving urban sustainability, should be developed through nuanced and integrated design that reflects the unique economic and social context of each community or commercial district, rather than being uniformly applied across the city.

Author Contributions

Conceptualization, M.P., B.Y. and J.L.; methodology, M.P., J.W. and D.P.; validation, B.Y. and J.L.; formal analysis, M.P. and J.W.; investigation, J.W. and B.Y.; resources, J.L.; data curation, M.P., J.W. and D.P.; writing—original draft preparation, M.P. and J.W.; writing—review and editing, B.Y. and J.L.; visualization, J.W.; supervision, J.L.; project administration, J.L.; funding acquisition, J.L. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2024S1A5A2A03038101).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The original data presented in the study are openly available in the data sources specified in the manuscript.

Acknowledgments

We’ve confirmed that all individuals explicitly named in the Acknowledgements section have provided their consent for inclusion.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Daniels, P.; El Baghdadi, O.; Desha, C.; Matthews, T. Evaluating net community benefits of integrating nature within cities. Sustain. Earth 2020, 3, 12. [Google Scholar] [CrossRef]
  2. Grey, T.; Xidous, D.; O’Neill, D.; Collier, M. Growing Older Urbanism: Exploring the nexus between ageing, the built environment, and urban ecosystems. Urban Transform. 2023, 5, 8. [Google Scholar] [CrossRef]
  3. Joye, Y.; Willems, K.; Brengman, M.; Wolf, K. The effects of urban retail greenery on consumer experience: Reviewing the evidence from a restorative perspective. Urban For. Urban Green. 2010, 9, 57–64. [Google Scholar] [CrossRef]
  4. Joye, Y. Architectural lessons from environmental psychology: The case of biophilic architecture. Rev. Gen. Psychol. 2007, 11, 305–328. [Google Scholar] [CrossRef]
  5. Kellert, S.R.; Heerwagen, J.; Mador, M. Biophilic Design: The Theory, Science and Practice of Bringing Buildings to Life; John Wiley & Sons: Hoboken, NJ, USA, 2011. [Google Scholar]
  6. Turley, L.W.; Milliman, R.E. Atmospheric effects on shopping behavior: A review of the experimental evidence. J. Bus. Res. 2000, 49, 193–211. [Google Scholar] [CrossRef]
  7. Maxwell, S.; Kover, A. Negative affect: The dark side of retailing. J. Bus. Res. 2003, 56, 553–559. [Google Scholar] [CrossRef]
  8. Luczaj, K. Evolutionary Aesthetics and Print Advertising: Empirical Evidence from Poland. SAGE Open 2015, 5, 2158244015592165. [Google Scholar] [CrossRef]
  9. Wolf, K.L. Nature in the retail environment: Comparing consumer and business response to urban forest conditions. Landsc. J. 2004, 23, 40–51. [Google Scholar] [CrossRef]
  10. Wolf, K.L. Trees in the small city retail business district: Comparing resident and visitor perceptions. J. For. 2005, 103, 390–395. [Google Scholar] [CrossRef]
  11. Wolf, K.L. Assessing public response to freeway roadsides: Urban forestry and context-sensitive solutions. Transp. Res. Rec. 2006, 1984, 102–111. [Google Scholar] [CrossRef]
  12. Wolf, K.L. Community context and strip mall retail: Public response to the roadside landscape. Transp. Res. Rec. 2008, 2060, 95–103. [Google Scholar] [CrossRef]
  13. Nasar, J.L. The effect of sign complexity and coherence on the perceived quality of retail scenes. J. Am. Plan. Assoc. 1987, 53, 499–509. [Google Scholar] [CrossRef]
  14. Campbell, S.D. The planner’s triangle revisited: Sustainability and the evolution of a planning ideal that can’t stand still. J. Am. Plan. Assoc. 2016, 82, 388–397. [Google Scholar] [CrossRef]
  15. Donovan, G.H.; Butry, D.T. The effect of urban trees on the rental price of single-family homes in Portland, Oregon. Urban For. Urban Green. 2011, 10, 163–168. [Google Scholar] [CrossRef]
  16. Sander, H.A.; Zhao, C. Urban green and blue: Who values what and where? Land Use Policy 2015, 42, 194–209. [Google Scholar] [CrossRef]
  17. Lin, J.; Huang, B.; Wang, Q.; Chen, M.; Lee, H.F.; Kwan, M.P. Impacts of street tree abundance, greenery, structure and management on residential house prices in New York City. Urban For. Urban Green. 2024, 94, 128288. [Google Scholar] [CrossRef]
  18. Jang, D.C.; Kim, B.; Kim, S.H. The effect of green building certification on potential tenants’ willingness to rent space in a building. J. Clean. Prod. 2018, 194, 645–655. [Google Scholar] [CrossRef]
  19. Kang, Y.; Zhang, F.; Gao, S.; Lin, H.; Liu, Y. A review of urban physical environment sensing using street view imagery in public health studies. Ann. GIS 2020, 26, 261–275. [Google Scholar] [CrossRef]
  20. Biljecki, F.; Ito, K. Street view imagery in urban analytics and GIS: A review. Landsc. Urban Plan. 2021, 215, 104217. [Google Scholar] [CrossRef]
  21. Yue, H.; Liu, L.; Xiao, L. Investigating the effect of people on the street and streetscape physical environment on the location choice of street theft crime offenders using street view images and a discrete spatial choice model. Appl. Geogr. 2023, 157, 103025. [Google Scholar] [CrossRef]
  22. OECD. Self-Employment Rate. Available online: https://www.oecd.org/en/data/indicators/self-employment-rate.html (accessed on 28 October 2022).
  23. Park, H.-S. Self-Employed in Seoul: Characteristics and Policy Support Direction. 2011. Available online: https://www.si.re.kr/node/42560 (accessed on 28 October 2024).
  24. Lee, J.Y. A Study on the Effect of Internal and External Factors of Self-Employed Business in Seoul on the Default of Credit Guarantee: Focused on the Start-Up Guarantee of the Seoul Credit Guarantee Foundation. Master’s Thesis, Yonsei University, Seoul, Republic of Korea, 2021. [Google Scholar]
  25. Lee, S.; Ishiro, T. Regional economic analysis of major areas in South Korea: Using 2005–2010–2015 multi-regional input–output tables. J. Econ. Struct. 2023, 12, 12. [Google Scholar] [CrossRef]
  26. Seoul City Commercial District Analysis Service (SCCDA). Available online: https://golmok.seoul.go.kr/introduce.do (accessed on 30 October 2024).
  27. Fondeville, N.; Ozdemir, E.; Lelkes, O.; Ward, T.; Zolyomi, E. Recent changes in self-employment and entrepreneurship across the EU. Res. Note 2015, 6, 2015. [Google Scholar]
  28. Chung, S.; Kim, E.K.; Kim, W.-Y. Does increase in self-employment affect economic growth and unemployment rate? J. Ind. Econ. Bus. 2020, 33, 1937–1959. [Google Scholar] [CrossRef]
  29. Lee, J.; Han, J.; Kim, J.; Oh, Y.; Kim, M.; Bae, J. Study on self-employment and policy suggestions. KDI Res. Monogr. 2020, 6. [Google Scholar] [CrossRef]
  30. Yun, S. A study on the effect of local commercial characteristic information on closure rate of small business. J. Corp. Innov. 2019, 42, 21–39. [Google Scholar] [CrossRef]
  31. Williams, C.C. Rethinking the role of the retail sector in economic development. Serv. Ind. J. 1997, 17, 205–220. [Google Scholar] [CrossRef]
  32. Smart City Korea. Seoul City, Reorganization of Commercial District Analysis Service, More Convenient Big Data-Based Commercial District Information at a Glance. Ministry of Land, Infrastructure and Transport. 25 April 2022. Available online: https://smartcity.go.kr/en/%EC%A0%95%EC%B1%85/%EB%B3%B4%EB%8F%84%EC%9E%90%EB%A3%8C/?_sf_s=%EC%84%9C%EC%9A%B8%EC%8B%9C%20%EC%83%81%EA%B6%8C%EB%B6%84%EC%84%9D (accessed on 28 October 2024).
  33. Distribution Industry Development Act. Chapter 1, Articles 1–5. 2020. Available online: https://www.law.go.kr/LSW/eng/engLsSc.do?menuId=2&section=lawNm&query=%EC%9C%A0%ED%86%B5%EC%82%B0%EC%97%85%EB%B0%9C%EC%A0%84%EB%B2%95&x=50&y=36#liBgcolor0 (accessed on 30 October 2024).
  34. Lee, M.H.; Hong, S.; Shin, D.B. A study on application of spatial weighted regression for analyzing the influential factors of sales in sub-region commercial area of Seoul: Focusing on geo-spatial big data utilization. Geogr. J. Korea 2019, 53, 21–224. [Google Scholar]
  35. Woo, J.G. Analysis on the Influence of Changes in the Composition of Commercial District in Itaewon: Focusing on the Relationship Between Floating Population and Sale. Master’s Thesis, Seoul National University, Seoul, Republic of Korea, 2021. Available online: https://dcollection.snu.ac.kr/handle/10371/000000165474 (accessed on 20 May 2025).
  36. Cao, Y.; Lee, S. Analysis of urban vitality and its determinant factors using POI big data in Seoul, Korea. J. Korea Plan. Assoc. 2021, 56, 87–102. [Google Scholar] [CrossRef]
  37. Gehl, J. Cities for People; Island Press: Washington, DC, USA, 2013. [Google Scholar]
  38. Yin, L.; Wang, Z. Measuring visual enclosure for street walkability: Using machine learning algorithms and Google Street View imagery. Appl. Geogr. 2016, 76, 147–153. [Google Scholar] [CrossRef]
  39. Knapp, A. An introduction to clinical perimetry. Arch. Ophthalmol. 1938, 20, 1116–1117. [Google Scholar] [CrossRef]
  40. Kim, S.; Woo, A. Streetscape and business survival: Examining the impact of walkable environments on the survival of restaurant businesses in commercial areas based on street view images. J. Transp. Geogr. 2022, 105, 103480. [Google Scholar] [CrossRef]
  41. Wang, C.; Endo, T.; Hirofuchi, T.; Ikegami, T. Pyramid Swin Transformer for multi-task: Expanding to more computer vision tasks. In Proceedings of the International Conference on Advanced Concepts for Intelligent Vision Systems, Cham, Switzerland, 21–23 August 2023; Springer Nature: Cham, Switzerland, 2023; pp. 53–65. [Google Scholar]
  42. Liu, Z.; Lin, Y.; Cao, Y.; Hu, H.; Wei, Y.; Zhang, Z.; Lin, S.; Guo, B. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, QC, Canada, 10–17 October 2021; pp. 10012–10022. [Google Scholar]
  43. Jain, J.; Li, J.; Chiu, M.T.; Hassani, A.; Orlov, N.; Shi, H. Oneformer: One transformer to rule universal image segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, BC, Canada, 19–24 June 2023; pp. 2989–2998. [Google Scholar]
  44. Park, J. The Effect of Street Environments on Commercial Sales in Seoul: Using Street Big Data and Deep Learning Technologies. Master’s Thesis, Hanyang University, Seoul, Republic of Korea, 2023. Available online: http://www.dcollection.net/handler/hanyang/200000651849 (accessed on 25 November 2024).
  45. Spreiregen, P.D. Urban Design: The Architecture of Towns and Cities; McGraw-Hill: New York, NY, USA, 1965; p. 45. [Google Scholar]
  46. Kim, J.; Kim, S. Finding the optimal D/H ratio for an enclosed urban square: Testing an urban design principle using immersive virtual reality simulation techniques. Int. J. Environ. Res. Public Health 2019, 16, 865. [Google Scholar] [CrossRef]
  47. Kim, J.-Y. Myeong-Dong Is No Longer Seoul’s Priciest District for Store Rent. Korea JoongAng Daily. 8 March 2024. Available online: https://koreajoongangdaily.joins.com/news/2024-03-08/business/economy/Myeongdong-is-no-longer-Seouls-priciest-district-for-store-rent/1997702 (accessed on 25 November 2024).
  48. Choi, J.-Y. A Study on the Appropriate Rent and Asset Value of Franchise Lease Small Commercial Real Estate. Master’s Thesis, Dong-Eui University, Busan, Republic of Korea, 2021. [Google Scholar]
  49. Montella, A.; Chiaradonna, S.; De Sanctis Mihiel, A.C.; Lovegrove, G.; Nunziante, P.; Rella Riccardi, M. Sustainable complete streets design criteria and case study in Naples, Italy. Sustainability 2022, 14, 13142. [Google Scholar] [CrossRef]
  50. City of Los Angeles. Complete Streets Design Guide. Available online: https://planning.lacity.gov/odocument/c9596f05-0f3a-4ada-93aa-e70bbde68b0b/Complete_Street_Design_Guide.pdf (accessed on 26 November 2024).
  51. Ewing, R.; Handy, S. Measuring the unmeasurable: Urban design qualities related to walkability. J. Urban Des. 2009, 14, 65–84. [Google Scholar] [CrossRef]
Figure 1. Distribution of commercial districts in Seoul.
Figure 1. Distribution of commercial districts in Seoul.
Sustainability 17 06934 g001
Figure 2. Developed commercial districts and sample GSV image collection locations.
Figure 2. Developed commercial districts and sample GSV image collection locations.
Sustainability 17 06934 g002
Figure 3. (a) GSV image crop. Yellow polygons indicate ground-truth locations of 20 retail signs manually annotated. (b) Semantic segmentation output generated by Swin-L + OneFormer; blue outlines mark the areas the model labelled as retail signs. The model located 9 of the 20 annotated signs, but some were captured only partially rather than in their full extent. Street trees obstructed 10 signs, and these occlusions accounted for most of the missed detections.
Figure 3. (a) GSV image crop. Yellow polygons indicate ground-truth locations of 20 retail signs manually annotated. (b) Semantic segmentation output generated by Swin-L + OneFormer; blue outlines mark the areas the model labelled as retail signs. The model located 9 of the 20 annotated signs, but some were captured only partially rather than in their full extent. Street trees obstructed 10 signs, and these occlusions accounted for most of the missed detections.
Sustainability 17 06934 g003
Figure 4. Feature extraction in the OneFormer semantic segmentation framework with Swin-Transformer backbone: (a) demonstrates extracting and grid-dividing patches from panoramic images. This prepares the diverse views for the Swin L Transformer backbone; (b) shows a panoramic input image and its pixel-level semantic segmentation prediction. OneFormer, powered by Swin L, aims to perform such segmentation universally across tasks; (c) outlines the initial steps and the core mechanism of the encoder part within the OneFormer’s encoder-decoder structure. The RGB image is divided into non-overlapping patches. Each initial patch contains 16 pixels, which corresponds to a small image of size 4 × 4. Since each pixel has 3 RGB values, a single patch can be represented as having a size of 4 × 4 × 3. These divided patches are treated analogously to tokens in natural language processing (NLP). The total number of patches is H/4 × W/4, determined by the dimensions of the input image. The raw patches are then passed through a linear embedding layer, which transforms each patch into a size of 4 × 4 × (C/16), effectively increasing the channel dimension. Subsequently, the patches are processed by the Swin Transformer. The embedded patches are processed by the Swin Transformer, a hierarchical vision transformer. The Swin Transformer consists of multiple stages, where patches are progressively grouped and merged using shifted window attention. This hierarchical structure allows the network to capture long-range dependencies in the image. The final feature maps from the Swin Transformer can be used for various image processing tasks, such as classification, segmentation, and object detection.
Figure 4. Feature extraction in the OneFormer semantic segmentation framework with Swin-Transformer backbone: (a) demonstrates extracting and grid-dividing patches from panoramic images. This prepares the diverse views for the Swin L Transformer backbone; (b) shows a panoramic input image and its pixel-level semantic segmentation prediction. OneFormer, powered by Swin L, aims to perform such segmentation universally across tasks; (c) outlines the initial steps and the core mechanism of the encoder part within the OneFormer’s encoder-decoder structure. The RGB image is divided into non-overlapping patches. Each initial patch contains 16 pixels, which corresponds to a small image of size 4 × 4. Since each pixel has 3 RGB values, a single patch can be represented as having a size of 4 × 4 × 3. These divided patches are treated analogously to tokens in natural language processing (NLP). The total number of patches is H/4 × W/4, determined by the dimensions of the input image. The raw patches are then passed through a linear embedding layer, which transforms each patch into a size of 4 × 4 × (C/16), effectively increasing the channel dimension. Subsequently, the patches are processed by the Swin Transformer. The embedded patches are processed by the Swin Transformer, a hierarchical vision transformer. The Swin Transformer consists of multiple stages, where patches are progressively grouped and merged using shifted window attention. This hierarchical structure allows the network to capture long-range dependencies in the image. The final feature maps from the Swin Transformer can be used for various image processing tasks, such as classification, segmentation, and object detection.
Sustainability 17 06934 g004
Table 1. Dataset for analysis.
Table 1. Dataset for analysis.
VariableDescriptionSource
Predicted VariableThe average monthly rent in a developed commercial district per square meterKorea Credit Guarantee Fund
https://www.kodit.co.kr/koditEng/main.do
(accessed on 30 October 2024).
Rent
Predictor Variable—Economic factorsSeoul City Commercial District Analysis Service
https://golmok.seoul.go.kr/main.do
(accessed on 30 October 2024)
The number of retail shopsThe total number of retail shops in a developed commercial district
The annual rate of new retail shop openingsThe proportion of newly opened retail shops to the total number of retail shops
RevenueThe total amount of income generated by a business
IncomeThe maximum income of residents in a developed commercial district
Foot trafficThe number of pedestrians passing through a developed commercial district
Working populationThe total population that works in a developed commercial district
Survival rateThe proportion of retail shops that remain operational for 5 years compared to the total number of retail shops
POIThe number of points of interest (POI) in a developed commercial districtNational Geographic Information Institute
https://map.ngii.go.kr/mi/oprGuide/portalIntro.do (accessed on 30 October 2024)
GRDP per capitaGross Regional Domestic Product (GRDP) per capita within the municipal boundaries of the commercial districtStatistics Korea
https://kostat.go.kr/anse/ (accessed on 30 October 2024)
Predictor Variable—Built Environment factorsGoogle Street View images analyzed using semantic segmentation
Street Tree ViewThe proportion of segmented objects categorized into the following categories: street trees, sidewalks, signs, sky, roads, and buildings.
Sidewalk
Sign
Sky
Road
Structure
Table 2. Regression results for the developed commercial districts.
Table 2. Regression results for the developed commercial districts.
PredictorThe Average Monthly Rent of a Developed Commercial District
Full Model ResultsReduced Model Results
Coefficientsp-ValueCoefficientsp-Value
Intercept307,5000.000 ***315,3000.000 ***
shop_cnt_y160.000 ***13.980.000 ***
shop_open_y70630.065 .64690.064 .
revenue0.000032660.491
income_max−0.0012290.746
poi−890.282
pop_travel−0.016730.281
pop_work1.6850.796
surv_rate5−1470.600
grdp_cap1480.000 ***1540.000 ***
Street Tree View−34630.029 *−33900.028 *
sidewalk12270.621
sky−38690.003 **−39810.001 **
road−17260.158−19390.050 .
structure−15810.160−20670.045 *
Sample246
F-statistic10.61 20.99
Multiple R squared0.3945 0.3847
Adjusted R squared0.3573 0.3664
Significance: ‘***’ 0.001/‘**’ 0.01/‘*’ 0.05/‘.’ 0.1.
Table 3. Regression results for the whole commercial districts.
Table 3. Regression results for the whole commercial districts.
PredictorThe Average Monthly Rent of the Whole Commercial Districts in Seoul
Full Model ResultsReduced Model Results
Coefficientsp-ValueCoefficientsp-Value
Intercept123,3260.001 ***142,3170.000 ***
shop_cnt_y230.000 ***230.000 ***
shop_open_y14560.002 **14600.001 ***
revenue0.00002310.006 **0.00002160.010 **
income_max49300.044 *49660.039 *
poi−440.701−510.653
pop_travel−0.001090.877
pop_work−4.0575290.502
surv_rate5−120.912
grdp_cap2070.000 ***2080.000 ***
Street Tree View4920.186 0.028 *
sidewalk7280.652
sky−17530.009 **−22610.000 ***
road−11400.001 ***−11680.030 *
structure−12460.160−14610.000 ***
Sample1082
F-statistic28.92 44.82
Multiple R squared0.2751 0.2734
Adjusted R squared0.2656 0.2673
Significance: ‘***’ 0.001/‘**’ 0.01/‘*’ 0.05.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Park, M.; Wang, J.; Yim, B.; Park, D.; Lee, J. Street Trees’ Obstruction of Retail Signage and Retail Rent: An Exploratory Scene Parsing Street View Analysis of Seoul’s Commercial Districts. Sustainability 2025, 17, 6934. https://doi.org/10.3390/su17156934

AMA Style

Park M, Wang J, Yim B, Park D, Lee J. Street Trees’ Obstruction of Retail Signage and Retail Rent: An Exploratory Scene Parsing Street View Analysis of Seoul’s Commercial Districts. Sustainability. 2025; 17(15):6934. https://doi.org/10.3390/su17156934

Chicago/Turabian Style

Park, Minkyu, Junyoung Wang, Beomgu Yim, Doyoung Park, and Jaekyung Lee. 2025. "Street Trees’ Obstruction of Retail Signage and Retail Rent: An Exploratory Scene Parsing Street View Analysis of Seoul’s Commercial Districts" Sustainability 17, no. 15: 6934. https://doi.org/10.3390/su17156934

APA Style

Park, M., Wang, J., Yim, B., Park, D., & Lee, J. (2025). Street Trees’ Obstruction of Retail Signage and Retail Rent: An Exploratory Scene Parsing Street View Analysis of Seoul’s Commercial Districts. Sustainability, 17(15), 6934. https://doi.org/10.3390/su17156934

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop