People–Place Relationships in Regenerative Urban Assemblages: Streetscape Composition and Subjective Well-Being of Older Adults

Abstract

Cities are undergoing rapid transformations due to global trends such as population aging, climate change, and increasing social diversity. In order to address these challenges, urban planning must adopt regenerative approaches that enhance subjective well-being by fostering meaningful relationships between people and their surroundings. Streetscapes, which serve as accessible urban landscapes, are important, especially for older adults, who depend on their local environment due to mobility constraints. This study examines the composition of streetscapes and the subjective well-being of older adults in a Japanese municipality. Using streetscape imagery and semantic segmentation, we quantified landscape elements—including vegetation, sky, roads, and buildings—within various walking distances from participants’ residences. Subjective well-being was measured using an 11-point Likert scale and analyzed by ordinal logistic regression. The results revealed that specific streetscape elements significantly impacted subjective well-being differently across spatial thresholds, showing that micro-scale urban landscapes are substantially important in promoting well-being among older adults. This study provides evidence-based insights for adaptive, inclusive, and regenerative urban planning strategies that promote the well-being of diverse demographic groups.

1. Introduction

Cities are rapidly transforming in response to global issues such as climate change [1,2,3], social diversity [4,5], and demographic shifts [6,7,8]. These challenges highlight the limitations of conventional urban planning, as cities have often been viewed as static entities rather than dynamic systems that are adaptive and resilient [9,10]. In this context, regenerative urbanism has emerged as a conceptual framework that mitigates environmental and social risks, actively adapts to threats, and enhances urban ecosystems [11]. By incorporating nature-based solutions, circular economy, and participatory governance, regenerative approaches seek to establish positive loops throughout the multi-scale urban environment. Such approaches offer a strategic pathway to address the immediate risks associated with the Anthropocene while enhancing urban resilience and livability [12].

In order to promote the design and planning of cities that fully realize their potential for urban renewal, it is essential to recognize that cities are more than just physical infrastructure but are composed of a complex assemblage of diverse and interconnected components, including built form, social relationships, ecological processes, and governance structures [13,14]. Assemblage is the reinterpretation of the concept of “assemblage (agencement)” discussed in the writings of Deleuze and Guattari that has been developed by Manuel DeLanda, offering a dynamic perspective on urban transformation, from macro-scale infrastructure to micro-scale everyday space, highlighting how components of different scales interact and exhibit emergent properties [15]. This perspective is consistent with the multi-scalability that regenerative urbanism emphasizes, which ensures that interventions are not limited to large-scale projects but are across various dimensions. As the most familiar and accessible urban landscape, the streetscape serves as an important site where regenerative urban strategies can directly impact the quality of life. Through their impact on social interactions, environmental experiences, and individual mobility, streetscapes hold great potential for fostering inclusive, resilient, and regenerative urban environments [16,17].

Aging is a pressing issue among the many challenges facing contemporary urban environments, especially in a hyper-aged society such as Japan [18]. Older adults are disproportionately affected by the design and accessibility of their immediate environment, as they often experience mobility constraints [19,20]. In this context, micro-scale urban landscapes, including streetscapes, have been noted to directly influence perceptions of safety, walkability, and social inclusion, which are essential for maintaining quality of life [21,22,23,24]. Nevertheless, previous research has focused on limited types of spatial elements and their impact on behavior and cognition, leaving room for further investigation of experiential and place-based aspects of well-being.

When examining the relationship between streetscapes and individual well-being, subjective well-being emerges as a significant concept. This term encompasses an individual’s self-evaluation of happiness or satisfaction, attracting enormous attention from diverse academic and practical fields [25]. Many studies have shown a strong connection between subjective well-being and physical health, suggesting that a positive evaluation of well-being is linked to better health outcomes [26,27]. In light of its extensive implications, subjective well-being is a crucial indicator for assessing older adults’ health and quality of life.

Studies have been conducted on the effects of urban environments and landscapes on subjective health and well-being, with natural elements as the primary focus [28]. However, the results regarding natural elements vary. For example, one study showed that exposure to the natural outdoor environment within 300 m of their residence improved subjective health [29]. Another study found that exposure to vegetation and waterscapes within 100 and 150 m has been shown to decrease depression [30]. On the other hand, one study has concluded that the amount of greenery within 400 m, 800 m, and 1200 m of one’s home has no significant effect on subjective health [31]. These studies use the Normalized Difference Vegetation Index (NDVI) to indicate the amount of greenery [32]. In addition to natural landscapes, studies on the urban environment have also shown the impact of gentrification [33], public safety [34], and walkability [35] on subjective health perceptions.

Based on these arguments, this research aims to investigate the relationship between streetscapes and subjective well-being, which seems essential for older adults, who depend on the local environment for their daily activities. Research that has focused on various landscape elements and analyzed them at the micro-scale of the living area, with the street as the unit, which is a spatial scale directly related to the living experience of older adults, remains scarce. This study examined the relationship between the configuration of streetscapes and the subjective well-being of older adults in a municipality in Japan. Specifically, we quantified how streetscape at various spatial thresholds affects the sense of well-being by semantically subdividing the street-level images and evaluating the configuration of important landscape elements. The introduction and comparison of multiple spatial thresholds are based on the defining characteristic of assemblages, in which individual components express their characteristics as an aggregate at any scale [13], and is an attempt to incorporate the idea of “city as assemblage” as a dimension of urban planning and spatial design. The findings contribute to developing evidence-based urban design strategies that enhance livability through regenerative and inclusive planning methodologies.

The structure of the paper is as follows: Section 2 introduces the methodology, including data collection, image analysis, and statistical modeling. Section 3 presents the results, emphasizing significant relationships between streetscape elements and subjective well-being. Section 4 discusses the implications of these findings for regenerative urban design and planning. Finally, Section 5 concludes with recommendations for future research and policy applications.

2. Materials and Methods

2.1. Study Area

This study was conducted in Kunitachi, Japan. The country provides a suitable context for this study, given that individuals aged 65 and older constitute 26.6% of the population, according to the 2015 national census [36]. Kunitachi, situated approximately 30 km west of central Tokyo, was chosen for its blend of urban and suburban environments that characterize Japanese metropolitan suburbs. From a statistical perspective, Kunitachi is a typical situation in the Japanese metropolitan area. It is ranked as nearly average among municipalities in Tokyo in terms of the municipal aging rate (23.93% as of January 2023, ranked 21st/49th in Tokyo), population density (9346 persons per square kilometer as of January 2023, ranked 31st/49th in Tokyo), per capita income per taxpayer (4,807,000 yen as of 2022, ranked 16th/49th in Tokyo), and average residential land prices (356,500 yen per square meter as of July 2023, ranked 21st/49th in Tokyo) [37,38].

The Fujimidai neighborhood is located in the central part of Kunitachi. The area includes a variety of urban structures, from densely populated, walkable areas and boulevards designed for automobile traffic to agricultural lands. This mixture establishes an optimal setting for investigating the relationship between streetscape composition and subjective well-being. The area has an approximate population of 18,000 and is approximately 2.5 km from east to west and 1.0 km from north to south.

2.2. Data Collection

2.2.1. Street-Level Imagery

Street-level imagery was obtained from Mapillary [39], an online street imagery platform. Images taken in bad weather or at night were excluded from the dataset to ensure the quality of the analysis (see Figure 1 for examples).

2.2.2. Subjective Well-Being and Other Individual Data

Data on subjective well-being and other personal information were collected from a municipal questionnaire survey, the “Care Prevention and Daily Living Area Needs Survey”, in 2016. This survey, mandated for municipalities following the Long-Term Care Insurance Law, includes questions on perceived physical and mental health, social interactions, and financial circumstances. It is aimed at residents aged 75 years and older who have not been certified by the local government as requiring long-term care.

Subjective well-being was assessed by asking participants, “How happy do you feel now?” Responses were recorded on an 11-point Likert scale, ranging from 0 (not at all happy) to 10 (extremely happy). The survey also collected personal information, including age, sex (binary variable), and family structure (categorized as living alone, living with a spouse aged 65 or over, living with a spouse aged under 65, living with children, or other). Housing type was classified into seven categories: owned single-family detached house, owned unit in a housing complex, public rental housing, private rental detached housing, private rental unit in a housing complex, company housing, and others. Additionally, subjective financial status was assessed on a five-point scale (stable, somewhat stable, neither stable nor unstable, somewhat unstable, unstable). The survey also inquired about the frequency of going out, willingness to participate in community activities, and willingness to organize community activities. One thousand one hundred forty-six responses were collected from people living in the Fujimidai neighborhood. Those responses were geocoded to enable spatial analysis. The residential addresses were included in the basic data on individual residents, which was provided under a joint research agreement concluded between Kunitachi City and the authors. To avoid the leakage of personal information through the use of the geocoder, an anonymized ID was assigned to the basic data and questionnaire response data for use in this study, and a file containing only this ID and text indicating their addresses was used to convert addresses to latitude and longitude using the geocoder publicly available online, after which the latitude and longitude were reflected in the questionnaire response data using the ID. We obtained permission to use address information in our analysis, provided that we carried out the above operations and that we would not disclose individual results.

Table 1. Sociodemographic and personal characteristics of the survey respondents.

Category	Item	Number of Respondents	Proportion (%)
Age	75–79	312	34.1
	80–84	411	44.9
	85–89	148	16.2
	90–94	37	4.0
	95 or over	4	0.4
Sex	Male	386	42.2
	Female	529	57.8
Family structure	Living alone	298	32.6
	Living with a spouse aged 65 or over	399	43.6
	Living with a spouse aged under 65	9	1.0
	Living with children	111	12.1
	Other	98	10.7
Housing type	Owned single-family detached house	229	25.0
	Owned a unit in a housing complex	171	18.7
	Rental public housing	406	44.4
	Rental private detached house	3	0.3
	Rental private unit in a housing complex	92	10.1
	Company-provided housing	7	0.8
	Other	7	0.8
Subjective financial status	Stable	7	0.8
	Somewhat stable	43	4.7
	Neither stable nor unstable	546	59.7
	Somewhat unstable	253	27.7
	Unstable	66	7.2
Frequency of going out	Five or more times a week	385	42.1
	Two to four times a week	414	45.2
	Once a week	84	9.2
	Rarely	32	3.5
Willingness to participate in community activities	Would love to participate	70	7.7
	May participate	487	53.2
	Do not want to participate	358	39.1
Willingness to organize community activities	Would love to organize	24	2.6
	May organize	271	29.6
	Do not want to organize	620	67.8

presents the sociodemographic and personal characteristics of the survey respondents. The sample consisted of 915 older adults residing in the Fujimidai neighborhood, with a mean age of 80.66 years (SD = 4.09). Among the respondents, 42.2% were male, and 57.8% were female. Regarding family structure, 32.6% lived alone, 43.6% resided with their spouse aged 65 or over, 1.0% lived with their spouse under 65, 10.7% resided with their children, and 10.7% lived with other family members or acquaintances. Housing type varied, with 25.0% owning a single-family detached house, 18.7% living in an owned unit in a housing complex, 44.4% living in public rental housing, and the other respondents residing in private rental, company-provided housing, and others. Subjective financial status, measured on a five-point scale, showed that 5.5% of participants perceived their financial situation as stable and 34.9% as unstable, whereas 59.7% considered it neither stable nor unstable. Regarding daily activity and community engagement, the frequency of going out varied across individuals, with 42.1% reporting daily outings (five or more times a week), 45.2% reporting occasional outings (two to four times a week), 9.2% going out once a week, and 3.5% rarely going out. 60.9% of respondents expressed a willingness to participate in community activities eagerly or somewhat eagerly, while 32.2% showed interest in organizing such initiatives.

2.2.3. Additional Data

Other datasets, including land use maps and road network data, were integrated for spatial analysis. Land use and building information were extracted from the Tokyo Metropolitan Government’s Geographic Information System data, and “street units” were established, comprising segments of streets delimited from intersection to intersection [40]. Street network data were obtained from OpenStreetMap.

2.3. Image Processing and Feature Extraction

Streetscape elements were analyzed using semantic segmentation, a deep learning technique that classifies image pixels into predefined categories. This study employed DeepLabv3+ [41] trained on a Mapillary Vistas Dataset [42] with urban landscape elements. Key elements extracted included the following:

• Object: pole, utility pole, traffic sign front, traffic sign back, traffic sign frame, streetlight, billboard, traffic light, trash can, junction box, car, bus, truck, motorcycle, bicycle;
• Construction: building, wall, fence, rail track, road, sidewalk, curb, guard rail, other barrier;
• Human: person;
• Nature: sky, vegetation, terrain, mountain, water;
• Marking: crosswalk zebra, general;

2.4. Spatial Analysis

Network-based distance measures were used instead of conventional circular buffers to better reflect the everyday walking experiences of older adults. Street network distances were calculated using GIS-based routing algorithms, providing a more accurate representation of walkable access to urban features. The distances between the center points of all street unit pairs within and around the Fujimidai neighborhood were computed using the QNEAT3 network analysis tool, a QGIS plug-in. Spatial thresholds of 50 m, 100 m, 150 m, 200 m, 250 m, and 300 m were applied to calculate the average percentage composition of each landscape element within all these scales from each street unit.

2.5. Modeling

We employed hierarchical ordinal logistic regression to examine the relationship between streetscape composition and subjective well-being. This method was chosen due to the ordinal nature of the dependent variable, which represents subjective well-being on a graded scale. The hierarchical approach was adopted to assess the incremental explanatory power of landscape composition variables beyond demographic and personal factors. By structuring the analysis in stages, we aimed to isolate the contribution of streetscape characteristics while controlling for individual-level covariates.

The variables were hierarchically incorporated into the model as follows:

• Baseline Model: The first model included only demographic and personal variables, such as age, gender, housing type, subjective socioeconomic status, and self-reported everyday behaviors. This model served as a reference to assess the effect of individual characteristics on subjective well-being.
• Streetscape Composition Model: The second model introduced streetscape composition variables, calculated as the proportion of different landscape elements within predefined network distance thresholds. These elements, extracted through semantic segmentation of street-level images, included vegetation, sky, road, and building, as listed in the previous section.

As mentioned above, the independent variables for landscape composition were aggregated based on network distances rather than circular buffers, ensuring a more accurate representation of physical accessibility in older adults’ daily experiences.

To assess model performance, we examined key goodness-of-fit measures, including the corrected Akaike Information Criterion (AICc) and Nagelkerke R-squared. We used those statistical indicators to compare those models and determine whether including streetscape variables significantly improved explanatory power. By employing this hierarchical modeling approach, we aimed to disentangle the relative contributions of individual elements and streetscape composition to subjective well-being. This approach provides a robust framework for evidence-based urban design interventions.

3. Results

This section presents the results of the analysis based on survey responses and spatial analysis, examining the relationships between subjective well-being among older adults and both demographic/personal factors and the proportions of streetscape elements at multiple spatial thresholds from their residences. The findings indicate statistically significant relationships between certain streetscape elements and subjective well-being, highlighting the importance of micro-scale urban environments. The following subsections offer a detailed account of the descriptive statistics, the relationships between variables and subjective well-being, and the development of hierarchical ordinal logistic regression models.

3.1. Subjective Well-Being

The average subjective well-being score was 7.10 (SD = 2.18), assessed on an 11-point Likert scale. The distribution of subjective well-being (Figure 2) shows three peaks; therefore, we decided to label them as low (marked 0–6), mid (7–8), and high (9–10).

3.2. Streetscape Configurations

To identify potential impacts of correlations among landscape elements on the analysis, we first calculated the proportion and distribution of each landscape element in the streetscape configuration. Semantic segmentation was performed on 2374 street units in and around the Fujimidai neighborhood, specifying the latitude and longitude of their center points and selecting the street-level image of the nearest point. As a result, the percentage of each landscape element introduced in Section 2.3 was calculated in the image. Figure 3 shows the distribution in terms of the number of street units of the compositional proportions of the four typical elements: road, sky, building, and vegetation (these elements combined, on average, account for about 69% of the total area of the image). Roads accounted for an average of 13.5% of the image (SD = 0.059), and when divided by two percentage points, 10–12% of the image was the largest number of street units (316). Similarly, sky averaged 7.8% (SD = 0.047), buildings averaged 28.6% (SD = 0.116), and vegetation averaged 20.0% (SD = 0.119).

3.3. Bivariate Analysis

Bivariate analysis of sociodemographic and personal variables and subjective well-being is conducted as a prerequisite for creating a model to identify factors that influence subjective well-being. Since subjective well-being is a three-level ordinal scale of “low”, “medium”, and “high”, Pearson’s chi-square test was used to assess their significance. In all cases, the significance level was set at p < 0.05.

Table 2. Difference between sociodemographic and personal characteristics.

Category	Item	Subjective Well-Being (%)			p-Value
		Low	Mid	High
Sex	Male	41.7	38.3	20.0	0.0004
	Female	33.3	35.5	31.2
Family structure	Living alone	37.6	39.6	22.8	0.6732
	Living with a spouse aged 65 or over	37.3	34.3	28.3
	Living with a spouse aged under 65	22.2	33.3	44.4
	Living with children	36.0	35.1	28.8
	Other	34.7	39.8	25.5
Housing type	Owned single-family detached house	27.5	37.6	34.9	0.0005
	Owned a unit in a housing complex	30.4	39.8	29.8
	Rental public housing	44.3	35.7	20.0
	Rental private detached house	33.3	0.0	66.7
	Rental private unit in a housing complex	37.0	33.7	29.4
	Company-provided housing	57.1	28.6	14.3
	Other	42.9	57.1	0.0
Subjective financial status	Stable	14.3	42.9	42.9	<0.0001
	Somewhat stable	9.3	41.9	48.8
	Neither stable nor unstable	30.4	38.1	31.5
	Somewhat unstable	49.4	36.4	14.2
	Unstable	62.1	22.7	15.2
Frequency of going out	Five or more times a week	31.2	36.4	32.5	<0.0001
	Two to four times a week	36.5	38.9	24.6
	Once a week	52.4	33.3	14.3
	Rarely	68.8	21.9	9.4
Willingness to participate in community activities	Would love to participate	22.9	38.6	38.6	0.0006
	May participate	33.3	39.8	26.9
	Do not want to participate	44.4	32.1	23.5
Willingness to organize community activities	Would love to organize	20.8	36.7	26.5	0.0056
	May organize	29.9	40.2	29.9
	Do not want to organize	40.5	35.3	24.2

p-values < 0.05 were considered statistically significant.

shows the percentage of subjective well-being levels and p-value for each ordinal and nominal scale. Women had significantly higher levels of subjective well-being than men, and significant differences were found for housing type, subjective financial status, frequency of going out, willingness to participate in community activities, and willingness to organize community activities. On the other hand, no significant differences were found regarding family structure.

3.4. Ordinal Logistic Regression

Hierarchical ordinal logistic regression analysis was then conducted. As a first step, a baseline model was created that did not include streetscape configurations but only sociodemographic and personal variables. Based on the results of the previous section, age, sex, housing type, subjective financial status, frequency of going out, willingness to participate in community activities, and willingness to organize community activities were entered as variables. The statistical analysis software JMP Pro (Version 18.0.1) was used for the following analyses.

A baseline model was created by selecting variables using the maximum likelihood method. The model had a negative log-likelihood (NLL) of 921.51, an AICc of 1886.06, and a Nagelkerke R-square (NR2) of 0.1676. The Area Under the Curve (AUC) was fair, with 0.6998 for “medium” and 0.7082 for “high” subjective well-being.

Table 3. The statistical profile of a baseline model.

Characteristics	Degree of Freedom	Wald Chi-Square	p-Value
Age	1	3.8975	0.0484
Sex	1	17.7344	<0.0001
Housing type	6	13.0190	0.0427
Subjective financial status	3	26.7374	<0.0001
Frequency of going out	4	41.7977	<0.0001
Willingness to participate in community activities	2	2.4772	0.2898
Willingness to organize community activities	2	3.1912	0.2028

p-values < 0.05 were considered statistically significant.

shows the results of the Wald test for variable effects. These results indicate that the baseline model should include age, sex, housing type, frequency of going out, and subjective financial status.

The second step is to create a streetscape composition model with the percentage of streetscape elements within a specific road network distance from the respondent’s residence as a variable. As mentioned above, the distance thresholds are 50 m, 100 m, 150 m, 200 m, 250 m, and 300 m. In selecting the distances, we are based on the methodology in a previous study [29], where the analysis was conducted in the range of 50 m to 250 m in a radius from the place of residence. For each respondent, we take the center point of the street unit that the geocoded residence faces as the starting point, calculate the mean value of the percentage composition of streetscape elements in all street units closer than a specific distance threshold from it, and use these as the values of the landscape variables linked to the individual. In each distance threshold, the correlations among the streetscape elements were checked. It turned out that, for example, in the 50 m model, the VIF exceeds 15 for the four elements that occupy a large area in the image (road, sky, building, and vegetation). Therefore, a principal component analysis of the four component proportions was performed, and a new VIF was calculated by adding the first and second principal components (PC1, PC2) instead of the four components, resulting in a maximum VIF of 8.275 for the first principal component, which is sufficiently small. The same procedure was followed for the models from 100 m to 300 m.

For each distance threshold from 50 m to 300 m, parameters were again estimated by the maximum likelihood method to create a streetscape composition model.

Table 4. The statistical details and the effects of variables in streetscape composition models.

Variable	Model_50 m		Model_100 m		Model_150 m		Model_200 m		Model_250 m		Model_300 m
	Log Value	Wald Chi-Square	Log Value	Wald Chi-Square	Log Value	Wald Chi-Square	Log Value	Wald Chi-Square	Log Value	Wald Chi-Square	Log Value	Wald Chi-Square
Age	0.483	1.013	1.091	3.196	1.190	3.530	1.104	3.193	1.160	3.417	0.950	2.618
Sex	3.674	13.722 **	4.247	16.052 **	4.776	18.255 **	4.663	17.848 **	4.893	18.771 **	4.906	18.789 **
Housing type	0.878	7.618	0.850	9.772	0.512	7.389	0.903	10.309	0.568	7.908	1.237	12.552
Frequency of going out	2.867	15.390 **	7.237	35.141 **	6.437	31.452 **	6.503	31.720 **	6.007	29.328 **	6.106	29.909 **
Subjective financial status	5.687	31.081 **	8.311	42.930 **	9.276	46.818 **	9.096	46.122 **	8.417	43.020 **	7.836	40.375 **
PC1	0.201	0.236	0.123	0.103	0.749	1.913	0.755	1.905	0.044	0.015	1.863	6.075 *
PC2	1.136	3.316	0.197	0.225	1.001	2.728	0.106	0.074	0.313	0.500	0.009	0.001
Pole	0.356	0.598	0.540	1.124	0.301	0.465	0.618	1.370	0.034	0.009	1.178	3.438
Utility pole	2.355	7.950 **	0.052	0.020	0.050	0.019	0.469	0.924	0.199	0.230	0.130	0.113
Traffic sign (front)	0.299	0.451	0.067	0.033	0.287	0.430	0.928	2.452	0.092	0.058	0.022	0.004
Traffic sign (back)	0.166	0.175	0.576	1.228	0.731	1.720	0.510	1.022	0.555	1.178	0.162	0.163
Traffic sign (frame)	1.234	3.706	1.982	6.741 **	1.258	3.709	1.626	5.232 *	0.070	0.036	0.491	0.981
Streetlight	1.299	3.562	2.078	6.921 **	2.019	6.841 **	1.642	5.206 *	0.647	1.464	0.058	0.026
Billboard	0.079	0.042	1.462	4.320 *	1.822	5.851 *	1.544	4.801 *	0.984	2.691	0.667	1.593
Traffic light	0.316	0.504	0.231	0.305	0.145	0.135	0.844	2.164	0.654	1.526	0.207	0.257
Trash can	0.104	0.074	0.109	0.082	0.287	0.416	0.003	0.000	0.151	0.143	0.335	0.541
Junction box	2.596	8.826 **	1.070	3.057	0.575	1.282	0.828	2.122	0.347	0.583	0.141	0.127
Car	0.089	0.056	0.060	0.026	0.264	0.356	0.468	0.882	0.318	0.489	0.149	0.142
Bus	1.274	3.763	1.068	3.029	1.437	4.310 *	2.351	8.039 **	0.323	0.511	0.845	2.211
Truck	0.714	1.956	0.128	0.102	0.275	0.401	0.190	0.219	0.189	0.217	0.494	1.013
Motorcycle	1.309	3.824	0.885	2.380	0.451	0.866	0.779	1.888	0.046	0.015	1.404	2.990
Bicycle	1.179	3.484	0.080	0.048	0.268	0.392	0.220	0.272	0.133	0.117	0.072	0.037
Wall	1.264	3.545	0.431	0.796	0.362	0.632	1.892	6.284 *	0.377	0.678	0.052	0.021
Fence	0.334	0.545	0.401	0.722	0.138	0.121	0.311	0.476	0.336	0.556	1.738	5.513 *
Rail track	1.291	3.833	1.338	4.196 *	0.713	1.789	0.176	0.188	0.331	0.568	0.126	0.101
Sidewalk	1.811	5.678 *	0.265	0.367	0.150	0.140	0.938	2.394	1.307	3.788	0.214	0.261
Curb	1.452	4.375 *	0.098	0.063	0.296	0.453	0.244	0.317	0.147	0.138	0.774	1.915
Guard rail	0.262	0.364	0.323	0.527	0.136	0.124	2.291	7.938 **	0.142	0.136	0.221	0.275
Other barriers	0.729	1.674	1.030	2.771	2.176	7.312 **	1.544	4.990 *	0.492	0.999	0.904	2.496
Person	0.053	0.021	0.011	0.001	0.182	0.200	0.297	0.447	0.910	2.403	0.214	0.268
Terrain	1.694	5.126 *	0.756	1.838	0.555	1.187	0.096	0.063	0.322	0.507	0.178	0.189
Mountain	0.307	0.446	0.071	0.037	0.098	0.064	0.024	0.005	0.055	0.023	0.526	1.080
Water	1.394	3.457	0.059	0.024	0.032	0.008	0.807	1.997	0.485	0.944	0.233	0.305
Crosswalk	0.195	0.218	0.605	1.348	1.765	5.868 *	3.021	10.893 **	0.312	0.499	1.569	5.003 *
Marking (general)	1.804	5.745 *	0.436	0.804	1.071	2.967	1.165	3.287	0.943	2.510	0.233	0.296
Statistical details
NLL	648.81		874.19		905.24		902.53		914.97		911.47
AICc	1398.92		1847.79		1909.69		1904.26		1929.14		1922.15
NR2	0.2227		0.1975		0.2011		0.2066		0.1812		0.1884
AUC (mid)	0.7442		0.7219		0.7211		0.7288		0.7099		0.7149
AUC (high)	0.7161		0.7197		0.7254		0.7230		0.7116		0.7159

** p < 0.01, * p < 0.05.

presents the statistical details of the model for each distance threshold and the log values and Wald test results for each variable. NLL, AICc, NR2, and AUC were calculated for each model. The 50 m model showed a significant change in NLL compared to the baseline model, improving the model’s explanatory power. In comparison, while NLL and AICc remained similar to the baseline model for all models except the 250 m model, they increased AUC by over 0.01. No change in any statistical indicators was found on the 250 m model, and no significant streetscape elements were detected that significantly impacted it. The results indicate that, in addition to sociodemographic and personal variables, different streetscape elements affect the subjective well-being of older adults within different spatial boundaries.

Table 5. The odds ratio of statistically significant streetscape elements.

	Model_50 m	Model_100 m	Model_150 m	Model_200 m	Model_250 m	Model_300 m
PC1						0.003
PC2
Pole
Utility pole	9.728
Traffic sign (front)
Traffic sign (back)
Traffic sign (frame)		0.075		0.007
Streetlight		8.370	12.613	9.202
Billboard		0.087	0.098	0.088
Traffic light
Trash can
Junction box	0.079
Car
Bus			4.080	10.591
Truck
Motorcycle
Bicycle
Wall				41.619
Fence						0.072
Rail track		4.929
Sidewalk	5.375
Curb	0.121
Guard rail				282.808
Other barriers			0.094	0.113
Person
Terrain	6.598
Mountain
Water
Crosswalk			0.212	0.056		24.751
Marking (general)	5.185

shows the range of odds ratios for the significant streetscape elements in each model, expressing the odds ratio of subjective well-being between the largest and smallest value of the relevant streetscape element tied to the individual who responded to the survey.

4. Discussion

The hierarchical ordinal logistic regression analysis conducted in this study reveals that certain streetscape elements significantly affect subjective well-being at different spatial scales around the residence of older adults in relatively good health. Notably, these effects are more eminent at smaller spatial scales. It may be due to the diminishing differences between the places of residence as the distance increases. For example, terrain, which showed a statistically significant impact in the 50 m model, represents small-scale greenery and soil, and the distribution per street unit is shown in Figure 4a. In contrast, vegetation represents relatively large greenery, and the distribution per street unit is shown in Figure 4b. While the standard deviation of terrain is greater, street units with large proportions are scattered inside the city blocks, making them accessible to many individuals even in shorter distance thresholds. On the other hand, since large-scale vegetation is concentrated and distributed on some blocks and along arterial roads, many individuals can access them only when the spatial range extends to several hundred meters. Comparing the mean and standard deviation of the composition of terrain and vegetation for each individual from 50 m to 300 m, we see a gradual decrease in individual differences (Figure 5).

The revealed effects of streetscape elements on subjective well-being vary across spatial scales. In the 50 m model, the smallest spatial scales addressed in this study, the impact of streetscape elements, distinguished from sociodemographic and personal variables, was most eminent, with elements like sidewalks and terrain having a more positive impact. This finding suggests that enhancing a well-designed walking environment is important in the very vicinity of the residence of older adults. At 100 m to 200 m from the residence, streetlights, buses, walls, and guard rails contributed positively, while billboards and crosswalks had a negative effect. These results suggest that primary roads nearby with moderate traffic and pedestrian infrastructure play a role in well-being, while boulevards having commercial billboards and crosswalks may be too large to be present near their homes. Significant streetscape elements were not detected in the 250 m model. Finally, the positive impact of a lower PC1 value in the 300 m model was found.

To better interpret PC1 in the 300 m model, we will now investigate the principal component analysis of the composition ratio of the four main elements of the streetscape: road, sky, building, and vegetation. Figure 6 displays the component vectors of the four elements with PC1 on the horizontal axis and PC2 on the vertical axis for each model from 50 m to 300 m. Across all distance thresholds, PC1 has positive contributions from roads, sky, and buildings and negative contributions from vegetation. In PC2, road and sky are positive, buildings are negative, with vegetation remaining neutral. The range odds ratio of 0.003 for PC1 in the 300 m model suggests that a higher proportion of vegetation at this scale enhances subjective well-being. This finding aligns with previous research findings that access to large green spaces at the neighborhood scale is desirable for improving subjective health and well-being and reducing the risk of certain mental illnesses among older adults.

The results of this study contribute to neighborhood-scale planning and design through the lenses of regenerative urbanism and assemblage thinking. The insignificance of PC1 and PC2 below 250 m, coupled with the statistical significance of smaller-scale elements such as terrain and ground-level facilities, suggests a strong connection to the principles of regeneration and resilience. Unlike large-scale infrastructure such as buildings and roads, small-sized elements such as sidewalks, terrain, and guard rails can be modified or experimentally implemented in the short term, opening up the possibility of intervention through tactical efforts in the micro-scale urban environment. Additionally, the difference in significant streetscape elements at different spatial scales suggests that individuals may perceive a certain “semantic cohesion” at a sub-neighborhood scale. This empirical evidence supports the assemblage thinking, in which various components of a city dynamically assemble and disassemble across scales to express their emergent properties.

In discussing the findings obtained in this study, it is also necessary to consider the effect of the urban planning regulations. In the study area, the Kunitachi City Community Development Ordinance stipulates procedures for making development projects meet certain criteria, such as the submission of prior consultation documents, holding meetings to provide explanations to neighborhood residents, and confirmation of conformance with approval criteria. The ordinance sets standards, such as minimum lot size, green space, building height restrictions, road widths, and sidewalk installations [43]. The possibility that these standards influence the high percentage of sidewalks and vegetation, which can be extracted from street-level images, and subsequently contribute to improving the subjective well-being of older adults provides a rationale for the legitimacy of regulations and guidelines related to building and spatial design that have been introduced in various regions.

5. Conclusions

This study examined the relationship between streetscape elements and subjective well-being at different spatial scales and revealed that significant environmental factors vary depending on distance. Specifically, the desirable direction of streetscape design to enhance the subjective well-being of older adults includes the development of a safe and pleasant walking environment with terrains and sidewalks in the very vicinity of their homes and having a living primary road of moderate size located a short distance away from their homes. The study also revealed that large-scale greenery is beneficial if located within a neighborhood scale. Implications of these findings for urban design include the fact that multiple spatial scales should be considered simultaneously rather than relying on a single scale. This can be seen as an indication that assemblage thinking applies to urban design. In addition, given the fact that the streetscape elements for which standards are set in the Kunitachi City Community Development Ordinance, such as the provision of green space and sidewalks, were shown to contribute to the subjective well-being of older adults, it is expected that the findings on streetscape and subjective well-being presented in this study can be used in the formulation of such ordinances and guidelines.

Future research can contribute to a more comprehensive urban design and planning approach by exploring three directions while recognizing the limitations of this study. The first is to incorporate short-term temporal variability. Although open-source street-level imagery was used in this study, the images at each location were only taken at a single moment, and the study could not account for transient factors, such as time of day and weather conditions. This limitation in data availability could be solved by the emergence of new methodologies for capturing and sharing streetscapes. For example, mobility-mounted devices could facilitate more dynamic analyses. Data sources other than movies and images, such as audio, may also be combined to provide a more detailed understanding of the situation. Second, a long-term temporal variation could be added, as subjective well-being and the streetscape are expected to evolve. A longitudinal study tracking physical and perceptual changes in the streetscape could contribute to regenerative and assemblage-based urbanism theories. Finally, although this study uses a typical urban neighborhood in a Japanese metropolitan suburb as its case study, the relationship between streetscape and subjective well-being may differ in different urban contexts. Conducting comparative analyses across a broader range of people and cultures is essential to developing theories that account for cultural differences between countries and regions. This will ultimately lead to practices depending on those contexts.