Measuring Urban Green Space Exposure Based on Street View Images and Machine Learning
[ad_1]
1. Introduction
Urban green spaces (GSs), as the sole natural elements within urban structures, play a pivotal role in residents’ daily lives. They provide a range of services, including mitigating urban heat island effects [1], promoting urban hydrological cycles [2], enhancing biodiversity [3], and improving air quality [4]. Presently, rapid urbanization processes lead to chaotic urban expansion, population surges, and increased construction [5,6]. With escalating urban challenges and human habitats being at risk, urban GSs arise as nature-based, sustainable solutions [7,8]. This situation compels urban managers, planners, and researchers to reassess urban GSs, emphasizing their benefits to humanity [9].
GS exposure is typically defined as the extent of an individual’s (or group’s) interactions with natural settings [10]. This directly affects the health impacts of GSs on urban residents, making GS exposure a burgeoning interdisciplinary research area. There are various advantages of GS exposure [11,12]. On the one hand, GSs provide recreational areas, promoting outdoor exercise and physical health [13,14]. On the other hand, they can directly enhance residents’ mood, alleviate stress, restore focus, and, consequently, diminish the onset of disorders like depression [15,16,17]. The global trend of increasingly high population concentrations in urban areas is intensifying [18]. The exposure to GSs in the streets or at small scales heavily influences the daily lives of residents and is crucial in improving environmental quality. Such small-scale renewal and enhancement are also major aspects of future urban planning and construction [19,20]. Therefore, it is necessary to measure GS exposure at the street level to improve residents’ wellbeing.
Early research in environmental science and ecology focused on the significance of the natural environment for human society, emphasizing the protection of biodiversity, ecosystem services, and natural resources [21]. With the development of public health science, attention began to shift towards the impact of GSs on human health, such as improving the water quality, regulating microclimates, and purifying air [22,23]. Furthermore, psychological research has found that GSs have a positive impact on individuals’ mental health and emotional wellbeing, with the potential benefits of this impact even surpassing other aspects of greenery. Studies have shown that more GSs in people’s living environments are associated with enhanced social security, which helps improve their mental health [24]. With the acceleration of urbanization, people’s expectations for the quality of urban life have gradually increased, thereby focusing more attention on urban GSs as a part of residents’ daily lives [25,26]. In addition, the impact of urban GSs on residents has attracted further interest from scholars [15]. In this process, the concept of “GS exposure” was gradually proposed to describe the extent of people’s contact with GSs in their daily lives [10]. This concept quickly gained recognition in the academic community and was explored as a novel perspective on the relationship between humans and nature. Gong et al. argue that the positive effects of GS exposure are significant for vulnerable groups, such as children and the elderly [27]. Zhang et al. categorize the impact of GSs on residents’ health into three pathways: reducing harmful exposures, psychological restoration, and encouraging health-related behaviors [28]. Numerous studies have also combined GS exposure with epidemiological investigations, revealing the potential of GS exposure assessments in public health research [29,30].
The measurement of GS exposure is a foundational aspect of this area of research. The method of calculating GS exposure based on processing remote sensing data using GIS platforms has been widely used in previous research [31]. The Normalized Difference Vegetation Index (NDVI) can be used to measure the spatial distribution of urban greenery; therefore, it is often used as an important indicator of GS exposure [32]. Statistics Canada assessed GS exposure based on the NDVI within a 500 m buffer zone around residents’ dwellings [33]. Zhang et al. demonstrated the ways in which GS exposure can be achieved based on remote sensing imagery [34]. However, critics are concerned that the results of such measurements may be biased [35]. First, relying on geographical associations, such as the distance between communities and GSs, is an indirect method of measurement. This can result in challenges in aligning the multidimensional features of GS exposure, subsequently affecting the accuracy of the results [36]. Second, due to the resolution constraints of remote sensing imagery, it is difficult to precisely identify the exposure to GSs on a smaller scale [37]. Third, measurements of GS exposure should ideally focus directly on human subjects themselves. However, most existing studies have not paid attention to human perceptions or selection preferences. This oversight can lead to measurements that do not fully and accurately reflect the true conditions of GS exposure [38]. Additionally, the complexity of computation methods and the pronounced regional characteristics of geographical spatial data restrict the usability and generalizability of research outcomes. This limitation impedes the translation of study results into practical policies. Thus, there is a pressing need to establish a unified framework for measuring GS exposure [8].
How can we more accurately assess urban GS exposure from the perspective of residents? Is there a comprehensive and user-friendly method for measuring urban GS exposure, thereby providing better support for urban managers and planners? To address these questions, this study proposes a human-centric, multidimensional framework for assessing urban GS exposure. This study’s primary contributions are as follows: (1) Leveraging machine learning techniques and street view image data, this study establishes a wide-ranging, high-precision measurement framework for urban GS exposure, encompassing availability, accessibility, and attractiveness. (2) We evaluate the distribution of GS exposure in downtown Shanghai and explore its connection to urban spatial organization patterns. (3) From both detailed and broader perspectives, we examine the built environment’s influence on GS exposure. This study strives to characterize residents’ daily GS exposure, thereby better guiding urban planning and management strategies amidst urban development challenges.
1.1. Assessment of GS Exposure
Assessment methodologies for GS exposure can broadly be classified into two main categories: actual exposure and potential exposure [31]. Actual exposure assessments typically rely on self-reported data about individuals’ utilization of GSs, often collected via surveys or face-to-face interviews and detailing the frequency or duration of visits [39,40]. Recently, some studies have harnessed mobile signaling data or GPS data to obtain real-time exposure data, monitoring urban residents’ actual interactions with GSs [41,42]. The actual exposure can provide relatively accurate individual-level data and, through ongoing monitoring, yield temporal data series, facilitating insights into the “dose–response” relationship between GS exposure and resident health [43,44]. However, this method is constrained by challenges in data acquisition and the associated manual costs, making large-scale studies difficult. Potential exposure assessments are primarily concerned with bidimensional spatial exposure and three-dimensional visual exposure metrics [34]. Focusing on space as the subject of study, it underscores quantitative measurements of the characteristics of the GSs themselves. This allows for consistent data acquisition on spatial exposure [10], making it apt for large-scale measurements and research. By using street view images, researchers can precisely measure and analyze the distribution, types, and features of GSs, subsequently assessing the potential GS exposure levels of different communities and resident groups. Potential exposure assessments have the scope to delve deeply into health effects when combined with population data or activity trajectories [45]. Such a comprehensive approach aids in a holistic understanding of the influence of GSs on urban residents’ health and quality of life, providing a scientific foundation for urban planning and public health policy decisions [37].
Previous research assessing GSs often focused on the quantity and size of these spaces, only subsequently addressing issues such as the equity of their spatial distribution [46,47]. In fact, considering only the quantity and size of GSs, while neglecting the location of these spaces and their connection with residents, can lead to significant biases in the measurement of GS exposure. Incorporating the accessibility of GSs into the measurement of GS exposure is key to solving this problem [48]. Some researchers also suggest that the quality of a GS should be included in the factors affecting GS exposure [49], thus providing a more diversified perspective for its measurement. At the same time, most studies focus only on urban forests or parks, with little mention of street greening and residential GSs. This further widens the discrepancy between research findings and the actual effects of GS exposure.
1.2. Potential of Street View Images in GS Exposure Studies
In recent years, street view images have emerged as a pivotal resource for geospatial data collection and urban analysis [50,51]. Unlike remote sensing data, street view images allow researchers to observe and analyze urban spaces from a more human-scale perspective, virtually mimicking the effects of on-site inspections [52]. Street view images now cover the vast majority of cities globally and are gradually expanding to natural areas and rural regions, further increasing their data volume and application potential [53]. Google and Baidu are frequently used platforms for studies using such images. Google, being the world’s largest street view provider, pioneered and continues to deliver this service with impressive coverage and resolution [54]. In contrast, Baidu Street View offers superior coverage of Chinese cities, with their data accuracy matching that of Google [55]. Machine learning facilitates the extraction of meaningful results from massive datasets [56]. The integration of machine learning with street view images offers novel methodologies and perspectives for urban studies. It has been widely adopted in fields such as urban environmental measurement [57], property value assessments [58], and urban color analysis [59].
In urban research, measuring the proportion of GSs by using street view images can effectively replace the NDVI [43]. Street view images bring unique advantages to the assessment of GS exposure. Firstly, direct or indirect sensory stimulations play a crucial role here, with visual stimuli being particularly significant, especially on a psychological level [60,61]. This visual-exposure-based measurement approach is widely recognized in existing research. Secondly, street view images are stereo high-definition images that are collected based on a human perspective, allowing us to explore urban GS exposure from the scale of streets or smaller, and their accuracy is guaranteed. Thirdly, in previous research, due to the difficulty of conducting large-scale surveys, GSs could only be evaluated in terms of their quantity and location, making it difficult to assess GSs from the perspective of residents’ actual feelings about them [62]. Combining street view images with machine learning for perceptual measurements can effectively compensate for this shortcoming, thereby expanding the possible dimensions of GS exposure that can be measured [63]. It must be emphasized that based on open-source street view big data, this study’s proposed GS exposure assessment framework has great potential for expansion and application in research across different cities, which is crucial for urban planning and management [64].
4. Discussion
4.1. Multidimensional Framework for Measuring Urban GS Exposure
Although the significance of urban GSs is universally recognized, neglecting issues such as their accessibility to residents, their ease of use, and residents’ daily interactions with these spaces can lead to disparities between GS construction and utilization. This results in actual imbalances in their distribution. Hence, devising a comprehensive, scientific, and practical framework for measuring GS exposure is of paramount importance. The combination of street view images and machine learning can help us revisit GSs from the perspective of streets or at even smaller scales, integrating previously hard-to-obtain perceptual data into the research on GSs [63,94]. This approach allows for a more comprehensive and accurate discussion of the efficacy of GSs, moving beyond simply measuring their quantity. GS exposure is a human-centric evaluation indicator, and this framework is conducted from a human-centric perspective, specifically reflected in three aspects: (1) The research data are derived from street view images. These are façade images that are shot from a human viewpoint, providing a method for observing the urban environment from a human-centric perspective, laying the foundation for this study. (2) The assessment framework that is constructed in this study incorporates subjective human perception. Residents’ perception of the attractiveness of GSs is an important indicator of this framework, helping us to examine GSs from the residents’ perspective. (3) This study focuses on the street scale, assessing the GS exposure that is most relevant to residents’ daily lives, revealing the impact of daily GS exposure, and assisting in enhancing the social welfare effect of urban GSs.
More importantly, our proposed GS exposure measurement framework has scalability and application potential. Acquiring high-resolution remote sensing images often requires approval from governments or relevant agencies. In contrast, high-precision street view images are entirely open-source and much easier to acquire than high-resolution remote sensing images, and they can also support the measurement of GS exposure. With the support of this open-source street view big data, the GS exposure measurement that we developed in this study can be conveniently transferred for use in different urban spaces and widely applied in future urban planning and policy making.
4.2. Spatial Variation in GS Exposure in Downtown Shanghai
This study delved into the GS exposure in downtown Shanghai, revealing a strong correlation between GS exposure and urban spaces. The study found spatial differences in the availability, accessibility, and attractiveness of GSs, which were closely tied to factors such as the urban construction density, transportation conditions, and landscape quality. Our measurement results indicated a “low–high–low” spatial distribution pattern for GS exposure, which was directly related to the city’s developmental pattern and construction density. Central areas exhibited lower GS exposure, possibly due to their high construction density and land use restrictions. As they are typically hubs for commerce, culture, and politics with high land prices, the allocation of GSs in these regions is often constrained. This phenomenon is observed in many major cities, reflecting the contradiction between commercial and residential space demands and GS provisions [95,96]. In contrast to the city center, while peripheral urban regions might offer more room for GS development, the GS exposure remained low in peripheral urban regions due to inadequate infrastructure and geographical constraints. This suggests that, during the initial phases of urban development, the planning and construction of GSs are often overlooked, potentially impacting a city’s sustainable growth and its residents’ quality of life. Transition areas between the city center and the outskirts, with better road construction and transportation conditions, exhibit higher GS accessibility [97]. The elevated GS exposure levels in these regions can be attributed to their unique position during urban development phases. Being focal areas of urban expansion with superior infrastructure and public services, GS planning and construction are often focused on more here [98].
Considering the population distribution of Shanghai, the differences in GS exposure directly impact the equity of residents’ wellbeing. The population distribution in Shanghai is characterized by a dense center and sparsely populated outskirts [99], which directly interact with the “low–high–low” distribution pattern of GSs. This means that a large population in the central area has limited exposure to GSs. Moreover, the central area comprises many old neighborhoods, which are predominantly inhabited by elderly people [100] who have limited mobility. The negative effects of a low GS exposure are more pronounced in these living areas. With the trend of an ageing population, this issue warrants significant attention. Our study also found that in the central urban areas of Shanghai, the exposure to GSs was not clearly associated with residential areas. This suggests that the construction of GSs at the street scale has been somewhat neglected in past residential area developments. This also adversely affects the wellbeing of urban residents.
4.3. Micro and Macro Influences of Built Environment on GS Exposure
This study, at both the micro- and macroscales, extracted features of the urban built environment and conducted a cross-sectional analysis of GS exposure in relation to these features, aiming to uncover how urban construction influences GS exposure. Utilizing structural equation modeling, we regressed urban built environment factors at both scales. At the microlevel, we identified that the building and sky view indexes had significant negative correlations with GS exposure, while the sidewalk view index was positively correlated. An increased building view index typically signified building clusters encroaching upon GSs. Areas with a high sky view index often had lesser-developed infrastructures, leading to limited greenery. Well-constructed sidewalks increased the accessibility of GSs and the likelihood of residents visiting them, consequently enhancing GS exposure, which is consistent with a prior study [101]. By optimizing urban spatial layouts and improving the combination of sidewalks and GSs, GS exposure can be effectively enhanced [102].
At the macroscale, the building density was positively correlated with GS exposure. Although many studies have demonstrated that a high building density negatively affects the quantity and density of GSs [96], this is not necessarily the case for GS exposure. GS exposure needs to be considered comprehensively, taking into account the availability, accessibility, and attractiveness of GSs. In underdeveloped urban areas with a very low building density, there may be abundant greenery, but the area’s accessibility is poor, which discourages residents from entering, and the quality of the GS is not guaranteed. Paul and Bardhan argue that for areas lacking blue and GSs, increasing building sites is an effective improvement strategy [103], which supports the results of this study. This outcome also illustrates that, for cities, both excessive development and a lack of construction are detrimental to enhancing GS exposure.
A diverse range of POIs fostered GS exposure, while an excessive number of POIs could lead to spatial congestion and organizational chaos, emphasizing the importance of quality over quantity in urban planning. This underscored the importance of maintaining diverse POIs, avoiding overconcentration of similar types [104,105], and thereby sustaining adequate GS exposure. Geographical detector analysis further revealed the interactive effects among built environment factors, indicating their operation within a complex, multifactorial context that affects GS exposure. These findings provide a more comprehensive understanding of the relationship between the urban built environment and GS exposure, offering strong theoretical support for future urban planning and resource management.
In conclusion, through empirical analysis of downtown Shanghai, this study highlights the significant impact of the urban built environment on GS exposure, offering scientific backing for urban planners and managers and aiming to provide a comprehensive and in-depth theoretical foundation and practical guidance for sustainable urban development.
4.4. Policy Suggestions
China has fully transitioned from incremental development to stock development, and this is particularly true for Shanghai, a megacity in China. Optimizing the spatial structure and improving the quality of landscapes directly affect Shanghai’s urban vitality and residents’ wellbeing and are central to the city’s future urban planning and GS construction. The “Shanghai Municipal Land Space Planning (2020–2040)” framework, released by the Shanghai Municipal Government, clearly proposes to optimize the layout of urban GSs and enhance urban ecological functions, providing policy support for this study.
The findings of this study have important implications for urban planning. Firstly, urban planners need to re-examine the configuration of GSs in central urban areas and explore how to improve GS exposure under conditions of a high construction density. We suggest exploring innovative greening methods, such as rooftop greening and vertical greening, to make full use of limited spaces. Additionally, through urban renewal projects, transforming abandoned industrial lands into public GSs can bring more GSs to urban centers. Secondly, for the outskirts and newly developed areas, GS planning and construction should be thought through in the early stages of urban planning to ensure a sufficient proportion of public green lands and ecological lands, promoting sustainable urban development. Lastly, given the characteristics of higher GS exposure in transitional areas, it is important to strengthen the connectivity and accessibility of GSs in these areas by setting up more pedestrian paths and bike lanes, among other infrastructures, to increase residents’ utilization of and satisfaction with GSs. This will help improve the quality of life for urban residents and the overall environmental quality of the city.
In summary, there is a significant correlation between the results of GS exposure measurements and urban development patterns, providing valuable references for urban planning and management. By optimizing the configuration of GSs and enhancing their accessibility and attractiveness, we can promote the green development of cities, creating more livable and sustainable urban environments.
4.5. Limitations and Future Directions
We constructed a GS exposure assessment framework based on machine learning and street view imagery. Subsequently, we extracted urban built environment factors at both the micro- and macroscales, revealing the impact of the urban built environment on GS exposure. Our study results deepen our understanding of GS measurement and facilitate the practical application of our findings in urban construction. However, our study has its limitations, which will require further exploration and improvement in future research.
Firstly, we established our GS exposure assessment framework by selecting three dimensions: availability, accessibility, and attractiveness. Nevertheless, there are other aspects of the pathways through which GSs function. For instance, Gascon et al. suggested that the benefits of GSs to human health are achieved through “mediators”, such as improved air quality or noise reduction [106]. These comprehensive factors should be incorporated into future assessment frameworks to achieve a holistic understanding of the role of GSs.
Secondly, our study utilized a cross-sectional analysis to examine the influence of urban built environment factors on GS exposure, lacking a consideration for time series. Xing et al. pointed out that a comprehensive understanding of the spatial and temporal dynamics of GSs is crucial for planners and decision makers [107]. Future research should encompass broader spatial and temporal dimensions, incorporating socioeconomic factors and policy environments, to fully understand the relationship between urban development and GS exposure.
Lastly, it is essential to recognize that the goal of GS exposure is to enhance human wellbeing. Future studies should integrate population data or human mobility patterns to promote precise measurements at the resident level. Furthermore, incorporating public health data will shed light on the intricate mechanisms through which GS exposure impacts residents’ health.
5. Conclusions
In the context of rapid urbanization, evaluating GS exposure is crucial for enhancing urban quality and residents’ wellbeing. This study addressed the limitations of previous evaluation methods by employing machine learning and street view images to construct a novel, human-centric framework for assessing GS exposure. It comprehensively examined the efficacy of GSs regarding the aspects of availability, accessibility, and attractiveness. Supported by open-source street view big data, this framework boasts considerable scalability and operability, allowing for its application across different cities and holding the potential to impact urban planning, city management, and public health.
Using downtown Shanghai as a case study, we demonstrate the practicality of this evaluation framework. This study found that the spatial distribution of GS exposure in downtown Shanghai exhibits a clearly layered structure, increasing and then decreasing from the center to the outskirts. This reflects significant spatial heterogeneity in GS exposure, which becomes evident with urban development. Thanks to the establishment of this assessment framework, we further reveal the complex relationship between GS exposure and the urban built environment on both the micro- and macroscales. On a microscale, the building and sky view indexes are significantly negatively correlated with GS exposure, while the sidewalk view index is positively correlated, indicating that both the excessive clustering and dispersion of buildings are detrimental to GS construction, and that well-developed pedestrian pathways can enhance residents’ exposure to GSs. On a macroscale, the building density is positively correlated with GS exposure, showing that areas with higher urban development levels have greater GS exposure. Additionally, a diverse range of POIs can promote GS exposure, while an excessive number may have the opposite effect.
Overall, this study provides an efficient and scientific tool with which urban researchers and managers can assess and improve urban GS exposure. By thoroughly analyzing the case of downtown Shanghai, this study not only reveals the spatial distribution patterns and influencing factors of GS exposure but also offers valuable experiences and methods for other cities. This contributes to promoting sustainable urban development, enhancing the wellbeing of city residents and providing scientific guidance for urban planning and policymaking.
[ad_2]
