We addressed these topics in relation to three interlinked questions:

1. How are objective aspects of wellbeing (distributed according to socioeconomic and sociodemographic characteristics) related to subjective assessments of wellbeing (life satisfaction)?

To investigate these questions in real world settings two comparable but contrasting secondary cities of the Global South were selected (based on criteria including population and growth rates, mix of formal and informal growth, range of environmental concerns, relatively under researched) as representative examples in which to explore these concepts (see Figure 1 for details).

Nakuru, located within the Great Rift Valley, 160 km northwest of Nairobi, is the fourth-largest city in Kenya (after Nairobi, Mombasa and Kisumu) and the county capital. Nakuru lies at an altitude of 1,850 m and has a Mediterranean climate (Köppen-Geiger climate classification is Csb) remaining temperate throughout the year with no annual dry season. According to the County Integrated Development Plan 2018-2022, Nakuru town had an estimated population of 405,000 in 2018 which is expected to reach 458,000 by 2022 (a 13% increase). It has a mixture of built environments, including informal and unplanned settlements and both green and blue spaces. Rapid growth in Nakuru is putting development pressure on the public realm including greenspace.

Udon Thani in northeast Thailand is a small city of 130,000 residents facing rapid development due to its strategic location near the Laotian border. The city has a tropical savanna climate (Köppen-Geiger classification Aw) with warm dry winters followed by a 6-month monsoon season. Through the Udon Charter for 2029, a multi-stakeholder vision for the city, the city is committed to achieving six policy points, driven by the objective of becoming a green city focused on MICE (Meetings, Incentives, Conventions and Exhibitions). It seeks to have a walkable urban core, invest in green transport and green infrastructure including parks and public realm spaces.

Wellbeing can be considered a key component for a person’s quality of life and encapsulates both objective and subjective elements. The objective dimensions define wellbeing in terms of quality-of-life indicators including access to basic needs resources (e.g. food, housing, income) and social attributes (education, health, political voice, social networks). The subjective dimension emphasizes people’s own life evaluations including satisfaction (a cognitive evaluation) and happiness (relative emotional state) (Western and Tomaszewski, 2016). Subjective wellbeing encompasses hedonic functions such as pleasure attainment and pain avoidance, and eudemonic linked to a meaningful existence related to personal functioning (within individuals own mental and physical constraints) (Nordbakke and Schwanen, 2013).

This paper reports on the findings from two surveys detailed below: a bespoke neighbourhood survey investigating dimensions of socio-economic, environmental and wellbeing conditions (see 2.2.4.1 and 2.2.4.2 below); and a validated scale questionnaire exploring individual mood effects in different urban settings (see 2.2.4.3). All the survey tools received individual ethical approval via the relevant University of York, UK committee and participants gave informed consent. To facilitate accurate completion surveys were translated into local languages appropriate for each city.

To assess the quantity of greenspace satellite imagery processed to depict vegetation (normalized difference vegetation index (NDVI)) was accessed. Landsat imagery was processed by Climate Engine (climateengine.org) to determine the mean NDVI values for the 12 months prior to the survey period to assess the most recent variations in greenness that could affect wellbeing. These images were clipped to official neighbourhood boundaries for both cities and the distributions of 29 m pixel values determined for input to statistical tests.

One-way ANOVA and Chi2 tests were utilized in IBM SPSS Version 26 to assess the differences between variables based upon age, gender and location. Tukey and Cramer V post-hoc tests determined the significance of any emerging associations or differences. Linear regression analysis was used to assess the explanatory strength of relationships between variables. Kruskal-Wallis H test was used to assess the differences in the distribution of NDVI pixel values by neighbourhood.

The following sections results present findings relevant to our initial research question of ‘how are objective aspects of wellbeing (distributed according to socioeconomic and sociodemographic characteristics) related to subjective assessments of wellbeing (life satisfaction)?’ All the significant statistical analysis presented in this results sections are in table 2 below.

Our subjective wellbeing metrics varied within our case study cities by neighbourhood and with gender.

Looking at the influence of route on the participants in the transect walk, those who began their walk in the park and ended in the market did not see a significant change in hedonic tone (happiness). However, participants who began their transect walk in the market and ended in the park saw a significant decrease (pre-mean = 23.74; post-mean = 22.32) in hedonic tone (indicating increased sadness). These decreases effected both men and women significantly. These results contradict the findings from Nakuru where there were hedonic benefits attributed to greenspace for men.

In Kenya, our data indicates how informal and poorly implemented infrastructure has resulted in unequal access to the provision of basic services. Further exploration of these results highlight that poor water access and quality, and solid waste pollution contribute to measurable differences in objective wellbeing impacts between planned and less affluent districts. However, across all neighborhood’s people perceived the limited water access and crime incidence were undermining their wellbeing demonstrating that some challenges are ubiquitous.

In comparison, in Thailand, overall infrastructure and socio-economic conditions were largely un-problematic. However, despite the effects on objective wellbeing being marginal there were a greater number of differences between neighborhoods related to variations in air quality, noise pollution and traffic congestion. This indicates that as relative affluence increases, the significance of marginal inequalities between neighborhoods can become more pronounced. Our Udon Thani data also mirrors findings from other middle income countries (Colombia) (Scopelliti et al., 2016)that mid-affluent communitie’s wellbeing can benefit most from infrastructure availability and environmental improvements (refer to Figures 5, 6).

In Nakuru, subjective wellbeing predominantly lay in the ‘good’ range (scores of 26–28) (Ng Fat et al., 2017) whilst perceived stress fell within the moderate range (scores of 14–26) (State of New Hampshire Employee Assistance Program, 1983). In Udon, wellbeing was “good” to “excellent” (28+) but with between neighbourhoods differences becoming significant. Stress ranged from “moderate” to “low” but did not vary by neighbourhood indicating additional lifestyle factors beyond local environmental conditions were becoming influential on individual mental health. The absence of basic infrastructure (access to water; sanitation) causes significant stress alongside the obvious direct human-health effects (Ritchie and Roser, 2019). The granularity of our findings (at the neighbourhood scale) indicates that unequal access to basic services linked to affluence within LMIC cities significantly affects inequalities in resident’s subjective wellbeing. This contradicts the findings of Kelley and Evans (Kelley and Evans, 2017) who concluded from national survey data that inequality in income distribution boosted wellbeing in low-income countries. Overall, our findings show that city form and quality (both physical infrastructure provision and social interactions) in LMIC cities can have measurable impacts on the subjective wellbeing and stress of residents, potentially undermining their long-term mental health.

Significantly, our Kenya case results demonstrates that residents who make regular use of greenspace (greater than 2-h per week) show benefits to their subjective wellbeing independent of their neighborhoods conditions. This indicates the psychologically restorative benefits of greenspace can offset stress even for those living in informal settlements. In Thailand we did not find associations between wellbeing improvements and greenspace use. This could indicate that other factors influencing ability to spend time in greenspace (e.g. age, employment status) that also affect stress or wellbeing, are masking any benefits of time spent in natural surroundings. The Thai satellite data revealed, whilst there were variations in greenness between the city-centre and peri-urban fringe, most neighbourhoods had significant levels of vegetation. We hypothesize an alternative explanation for these findings that as urban vegetation is more equitably distributed across a city, routinely exposing people to nature, spending time specifically in greenspace has less discernable mental health benefits. The young cohort of Udon Thani transect walk participants indicated a subjective preference for a sociable retail space over a city park again indicating that perhaps greenspace may be less appreciated when it is widely available.

Our mixed-method approach highlights the complexity of these inter-relationship; however, they do identify a prioritization for urban planners when considering delivery of life satisfaction improvements. Our cross-city comparison highlights that delivering basic needs infrastructure or services universally must always be the primary city development priority. However, once these services are widely available urban form (distribution of public realm or greenspace) and management (socio-environmental conditions) require greater attention. Wellbeing effects associated with variations in these factors begin to take on a gendered and age dimension independent of neighbourhood affluence (employment and housing status). This conclusion is supported by other studies undertaken in higher income locations (Modai-Snir and van Ham 2018; Patel et al., 2017).

Our findings demonstrate that accessible public realm greenspace and neighbourhood greenery can offset some of the negative impacts on wellbeing of urban living even in challenging environments (socio-economic conditions) including informal settlements counteracting some income related health inequalities (Scopelliti et al., 2016). This supports findings on wellbeing impacts for low-income residents from park use in Indian and Colombian cities (Scopelliti et al., 2016; Ahirrao and Khan, 2021). When combined with the recognized physical health benefits (Siqueira Reis et al., 2013; Canterbury District Health Board, 2016; Adhikari et al., 2019), improved neighbourhood economic prosperity (Ahirrao and Khan, 2021) and co-benefits for active travel (Fluhrer et al., 2021) delivering these features more equitably across cities should be a key consideration for planners.

Our results also highlight that neighbourhood greening needs to be culturally appropriate and relevant for local communities including the urban poor (Ramaswami et al., 2016). This supports call for studies investigating distinct cultural and environmental conditions to make urban greenspace recommendations locally relevant in the Global South (Scopelliti et al., 2016) as the lived experience of residents from African, Asian or Latin American cities can vary distinctly due to factors including interactions of environment and infrastructure (Nagendra et al., 2018). For example, our Thai findings reveal local preferences for incorporating green infrastructure into retail and built public realm spaces to the maximize the distribution of wellbeing and ecosystem service benefits in this urban setting.

Our results highlight that distributing greenery throughout cities enables a wider cross-section of residents to enjoy benefits to their underlying wellbeing without needing to spend dedicated time in specific parkland destinations (Cocks et al., 2016; Markevych et al., 2017). This implies cities should incorporate greenspaces through street trees, greened roadside verges, or small-scale pocket parks rather than prioritizing larger but scarcer public parks supporting the findings of (Siqueira Reis et al., 2013). This could begin to counteract the emerging crisis in the rise of non-communicable diseases linked to inactivity and stress identified across South Asia (Adhikari et al., 2019). We add support to the social and spatial justice arguments for widening the distribution of urban greening (Camargo et al., 2017; Rigolon et al., 2018; Zuniga-Teran and Gerlak, 2019; Ahirrao and Khan, 2021) by adding in quantitative evidence on the wellbeing and livability benefits such improvements could bring.

Such distributed greenspace would also ensure equity in other ecosystem service benefits such as urban cooling; shading; biodiversity increases; and surface water flood mitigation (Panagopoulos et al., 2015; Canterbury District Health Board, 2016). Unfortunately, urban greenspace is declining across Global South cities especially rapidly growing secondary cities (Nero, 2017; Adhikari et al., 2019; Fluhrer et al., 2021). As highlighted by (Bai et al., 2018) city planners need greater access to neighbourhood scale data to truly understand the distributional impacts of urban form on resident’s health and city function. For example, internally displaced people residing in Nakuru county have been shown to have poor mental health, quality of life and life satisfaction (Getanda et al., 2015) contrasting with our Nakuru city participants who reported good overall life satisfaction and moderate stress. This demonstrates how high-resolution data is required to identify issues for specific places or population groups understanding local preferences to ensure city developments are appropriate and not merely transferred from different contexts (Nagendra, 2018; Cocks and Shackleton, 2021).

Cross-cutting development issues by their complex nature benefit from an integrated, multi-sector, consultative approach to problem solving if identified solutions are to be resilient (Mitra et al., 2017) in the context of diverse and dynamic city environments. New configurations of actors and collaborations are needed that include vulnerable groups and those typically excluded from city planning (Cinderby et al., 2021; Shackleton et al., 2021). This ambition to make improvements locally relevant (Patel et al., 2017) requires city authorities to plan using participatory co-design approaches that harnesses the collective creativity of people working together in a development process (Lam et al., 2017). Such approaches enable the development of improved shared understandings of complex problems allowing diverse stakeholder to collaborate and agree on locally relevant solutions (McArthur and Robin, 2019). This consensus building aids decision makers identification and delivery of more effective actions (Adelina et al., 2020). These approaches are particularly pertinent when addressing greenspace justice as public institutions typically fund these assets meaning all citizens should enjoy their benefits related to delivery of SDG 11.7 s (Daniel, 2014) ambition to “provide universal access to safe, inclusive and accessible, green and public spaces.”

The cross-sectional survey data used in this analysis represents a snapshot of conditions at a particular moment. Cities and communities are dynamic - investigating wellbeing’s relationship to changing urban environments would therefore benefit from a long-term longitudinal approach, similar to cohort studies from health sciences. Including a wider range of quantitative data with which to compare subjective wellbeing results and environmental perceptions would also provide a more robust picture of the relationship between people and cities. This could include measuring environmental factors known to affect wellbeing such as air and noise pollution, temperature, and humidity, but also quantitative recording of locally relevant socio-economic conditions such as crime or fluctuations in employment levels. Also improving our understanding in a more nuanced way of the interactions of people and places beyond home neighborhoods would explore temporal and seasonal dimensions. Incorporating more qualitative data from participants would add significant richness and additional context to the findings. Results from a complementary survey undertaken by the paper authors in both case study cities on the cultural ecosystem services that different urban spaces provide addresses this shortfall to a certain extent (Cinderby et al., 2021). We also recognize that focusing on greenness in our analysis lacks inclusivity of other spaces that may be valued within different cultural contexts. We would advocate for a wider definition of beneficial urban infrastructure to include natural (brown-, green-, blue-, and barren spaces) alongside built PRS (indoor and outdoor spaces), and their combinations when looking at the interactions of urban form and wellbeing in the Global South. Finally, this study was only undertaken in two cities; collecting similar data from a wider range of locations would significantly improve the robustness and transferability of our findings allowing a generic set of recommendations for a healthy, liveable city to be identified.