Introduction

The term Urban Heat Island (UHI) refers to the phenomenon where temperatures in urban areas are significantly higher than those in nearby non-urban areas (Rizwan et al. 2008). This effect results from the accumulation of heat in urban environments due to human activities such as transportation, industry, and buildings (Solecki et al. 2005; Yang et al. 2016). The increase in Land Surface Temperature (LST) caused by the UHI has various negative impacts, including disrupting species composition and distribution (Niemelä 1999), triggering storms and precipitation events (Bornstein and Lin 2000; Dixon and Mote 2003), worsening air pollution by exacerbating the formation of ground-level ozone and other air pollutants (Grimm et al. 2008; Wilby 2008), and contributing to long-term global warming (United States Environmental Protection Agency 2016; Pickett et al. 2001).

Central to addressing the UHI issue is urban and community forestry, the practice of managing and conserving tree populations in urban settings. This concept is essential in mitigating UHI effects as it focuses on increasing green spaces and tree cover in cities, leading to cooler environments. Community forestry extends this concept to involve local communities in the planning, management, and utilization of forest resources, recognizing the critical role urban residents play in sustaining and benefiting from urban green spaces. This approach promotes a more participatory and inclusive model of urban greening.

In exploring the relationship between urban and community forestry and UHI, it’s crucial to consider factors associated with forestry, such as the types of vegetation, the distribution of green spaces, and the management practices employed (Chen et al. 2024). Studies have shown that urban areas rich in tree cover and green spaces tend to have lower UHI intensities. This is attributed to the natural cooling effects of trees and plants, which provide shade and facilitate evapotranspiration. Conversely, areas with sparse vegetation are more susceptible to UHI. Urban greenery can lower surrounding air temperatures by up to 2–8 °C, significantly alleviating the UHI effect (Georgi and Dimitriou 2010). Urban forests also contribute to improved air quality. Trees and plants absorb pollutants like nitrogen oxides, ozone, and particulate matter, making the air healthier to breathe (Nowak et al. 2018). Beyond environmental factors, urban forests enhance mental health and wellbeing, with access to green spaces linked to reduced stress levels, improved mood, and overall better mental health (Zhang et al. 2020).

Urban heat issues in London are closely tied to urban forestry, playing a crucial role in mitigating the UHI effect and enhancing the city’s environmental quality. Research conducted in London found that a substantial green area spanning 111 ha lowered nighttime local air temperatures by as much as 4 °C compared to a neighboring urban zone. These green spaces, including street trees, parks, gardens, and green roofs, contribute to improving air quality by absorbing pollutants like nitrogen oxides, ozone, and particulate matter, with a valuation of approximately £126.1 million per annum in Greater London (Collins et al. 2019). Furthermore, urban forests in London play a vital role in carbon sequestration and storage, with total carbon storage and carbon sequestration valued at an estimated £146.9 million and £4.79 million per annum, respectively (Collins et al. 2019). However, challenges related to maintaining and expanding urban forests in London persist, including limited space, competing land uses, and the need for equitable distribution of green spaces to address environmental injustices (Vaz Monteiro et al. 2016).

Various methodologies have been extensively applied to understand and mitigate the UHI phenomenon. Geographically Weighted Regression (GWR), a local spatial regression technique, has been widely used to explore the spatial variability of the relationship between urban greenery and the temperature reduction. For example, Yang et al. (2022) have applied GWR and multi-scale geographically weighted regression (MGWR) to analyze the spatial heterogeneities of the LST and influencing factors, such as building density, land-cover diversity, and population density. Wang et al. (2022a, b) employed RF to model the complex interactions between 19 distinct urban characteristics and LST. They identified that cities with high thermal comfort should include low-density high-rise buildings, large urban parks with high levels of greenery and numerous large bodies of water. They also highlighted the importance of incorporating urban planning and green infrastructure to mitigate the UHI.

While these studies have provided valuable insights into the relationship between urban influenced characteristics and UHI, they often rely on single or isolated approaches. In this context, our research introduces a comprehensive geo-design framework that integrates multiple spatial analytical approaches. The framework aims to delve into the nuances of urban and community forestry, examining how different greenery types and forestry practices influence UHI. By leveraging the strengths of various methodologies, our geo-design framework contributes to provide insights and recommendations for urban planners and policymakers to optimize urban green spaces, thereby creating healthier and more livable urban environments.

Methodology

Study area and data source

The study employs a robust quantitative research design, integrating spatial and statistical analyses to comprehensively investigate the complex dynamics of UHI effect within the London Borough. To achieve this, a combination of advanced geo-design techniques is applied, including Bivariate Local Moran’s I, Random Tree Regression, and Modeling Suitability. Figure 1 shows the geo-design framework.

Fig. 1
figure 1

Geo-design framework

Full size image

With an estimated population of about 8.982 million, London, as the capital city of the United Kingdom, is one of the largest cities in the world (Zhang et al. 2022; Zhang & Chen 2024). This bustling metropolis, while known for its rich history and cultural diversity, also faces significant UHI challenges (Kolokotroni and Giridharan 2008; Watkins et al. 2002). As urbanization continues to expand, London has experienced a noticeable increase in temperatures, particularly in densely populated and commercial areas (Kolokotroni et al. 2006). This UHI effect has given rise to a range of issues, including elevated energy consumption for cooling, higher health risks during heatwaves, and strain on urban infrastructure. Understanding the current UHI condition in London and the associated problems is paramount for sustainable urban development and the well-being of its residents.

This study focuses on 11 of the 32 boroughs including the City of London, Islington, Westminster, Lambeth, Southwark, Hackney, Waltham Forest, Tower Hamlets, Lewisham, Newham, and Greenwich (a London Borough is an administrative area or local government district that provides public services and facilities to residents and businesses within its boundaries), as depicted in Fig. 2.

Fig. 2
figure 2

Study area

Full size image

Table 1 presents the resolution, application, and origin of various datasets employed in this study. These datasets include residential, industrial, and commercial densities; urban green space density; and urban forestry data. Data for residential, industrial, and commercial densities, as well as urban green space density, were sourced from OpenStreetMap (https://www.openstreetmap.org/) and presented as 50 m by 50 m grid tiles within a multipolygon structure. Meanwhile, metrics on urban forestry were procured from the London Datastore (https://data.london.gov.uk/) and represented as point shapefile. The count of trees within each 50 m by 50 m grid tile was determined by summing the total number of tree points falling within the respective grid. For temporal consistency, data from the year 2019 was specifically extracted and averaged across the aforementioned datasets. Furthermore, Land Surface Temperature (LST) information was derived from Landsat-8 Thermal Satellite data, providing the highest resolution of 30 m publicly available.

Table 1 Data source
Full size table

Procedure

Measuring spatial autocorrelation using bivariate local Moran’s I

Bivariate Local Moran’s I is utilized to explore the spatial patterns of the UHI effect across the borough (Chen et al. 2022). It allows for the identification of local clusters of high or low UHI intensity. By examining the spatial autocorrelation of land surface temperatures and urban factors, this analysis unveils not only where the UHI effect is most pronounced but also how it varies across different areas within the borough. This spatial perspective is crucial for understanding localized variations in temperature and identifying areas of particular concern. The formula for conducting Bivariate Local Moran’s I is shown as Eq. 1:

$${I}_{i}={\sum }_{j}^{n}=1,j\ne i{({X}_{j}-X)}^{2}/(n-1)({X}_{i}-X){\sum }_{j}^{n}=1,j\ne i{W}_{ij}({X}_{j}-X)$$
(1)

The Bivariate Local Moran’s Ii is a measure where Ii represents the index of a specific location, showing the relationship between that location and its surrounding area. This method provides an understanding of how areas with high and low values of certain attributes group together. High-high clusters are neighborhoods where high values of a certain attribute are found not only within the neighborhood itself but also in the surrounding areas. Conversely, a Low-Low cluster occurs when an area with low values is encircled by neighboring areas also having low values (Manojlović et al. 2021; Zhang et al. 2008). In our research, we employed the bivariate local Moran’s I to determine if the values of one variable (X) in a particular area are linked to the values of Land Surface Temperature (LST, or Y) in the adjacent areas.

In this research, the Bivariate Local Moran’s I was employed to determine the spatial association between various urban factors (such as density of tree canopy) and LST. Variables that consistently showed low values or lacked significant spatial association with LST were excluded from further analysis. This ensures that the subsequent models and analyses are based on variables that have a meaningful spatial relationship with LST.

Random forest regression

Random Forest (RF) is a versatile machine learning algorithm capable of performing both classification and regression tasks (Basha and Rajput 2019; Hitesh et al. 2019). It works by creating numerous decision trees during the training process and provides the average prediction of these individual trees for regression tasks (Basha and Rajput 2019). The strength of RF lies in its ability to provide an estimate of ‘feature importance’, which can be pivotal for feature selection and understanding the underlying structure of the data (Guan et al. 2012; Qi 2012). To ensure the robustness of the RF model, in this study, a common practice of splitting the data into training and testing sets was employed. Specifically, 70% of the data was used for training the model, allowing it to learn and understand the underlying patterns. The remaining 30% was reserved for testing, serving as a validation set to evaluate the model’s predictive performance on unseen data. If a certain feature is deemed as highly important by the RF model, then altering the value of that feature in a sample can substantially impact the prediction results of that sample. Conversely, if a feature is determined to be less important, making adjustments to that feature may have minimal effects on the prediction outcomes for the sample. This makes RF an invaluable tool for identifying and prioritizing critical features in a dataset. In this research context, the ‘Feature Importance’ metric plays a pivotal role in determining the weighted ratio of independent variables related to land use and air quality to mitigate the dependent variables, such as LST.

Modeling suitability

The output from the RF (Random Forest) analysis, combined with a key weighting system, forms the basis for the suitability analysis. The tool ‘Weighted Overlay’, as a pivotal component of Modeling Suitability, simplify the overarching overlay analysis process by merging multiple steps into a single, efficient tool. Its functionality extends to the reclassification of values within input raster, translating them onto a common evaluation scale that signifies suitability, preference, risk, or a comparable metric.

The use of ‘Weighted Overlay’ tool is to incorporate the weighting of importance for each cell value in the input raster. These combined weights produce an output raster reflecting overall suitability scores. In the context of suitability modeling, the output raster generally assigns higher values to areas considered more suitable for the intended purpose. Conversely, in applications like generating cost surfaces, the tool assigns higher values to areas with higher travel costs through the landscape. To harness the full utility of the ‘Weighted Overlay’ tool, a firm grasp of the scale values applied to input raster is crucial, ensuring that the values in the output raster are interpreted accurately.

All scale values in the input raster are uniformly set at ‘1’. This decision aligns with the random forest regression results Sect. 3.2 which underscored the significant impact of variables on LST. Furthermore, the results of the ‘Feature Importance’ are integrated as ‘weights’ within this ‘Weighted Overlay’ process. These weights play a pivotal role in generating the suitability areas aimed at mitigating the UHI effect in the inner city of London, UK.

Results

Spatial autocorrelation of urban forestry to LST

The spatial autocorrelations between various urban factors and LST were explored using the Bivariate Local Moran’s I, as shown in Table 2.

Table 2 Bivariate local Moran’s I value of independent variables to LST
Full size table

Urban & Community Forestry The results from Table 2 indicate that density of tree coverage is negatively correlated with LST. The value of − 0.084 indicates a weak negative local spatial correlation with LST. This suggests that in areas where tree coverage is dense, there tends to be slightly cooler land surface temperatures. The negative value reflects an inverse relationship, which is typical for vegetation providing cooling effects in urban settings. This means urban forestry not only mitigates the immediate temperature rise associated with urbanization but also highlights the critical significance of robust urban forestry practices in effectively combating the formation and exacerbation of urban heat islands, thereby fostering more sustainable and livable urban spaces.

Communities’ Land Use The results indicate that densities of residential, commercial, and industrial areas all have weak positive correlations with LST. This suggests that areas with higher human settlements, increased commercial activities, and greater industrial operations tend to experience elevated temperatures, contributing to localized warming. These findings align with the UHI effect, where urbanized areas typically exhibit higher temperatures than their rural counterparts due to factors such as heat-absorbing materials, reduced vegetation, and anthropogenic heat sources.

Overall, these findings highlight the complex interplay between different communities, greenery, and LST. While certain urban factors exacerbate the heat, elements like tree coverage offer respite. The spatial relationships, however, can vary across different regions, and a visual representation, such as Fig. 3, can provide a more detailed understanding of these local associations. Apart from the global perspective by illustrating the overall correlations between independent variables and LST, it is also essential to recognize that the relationships between these variables exhibit substantial spatial heterogeneity, meaning that they fluctuate significantly across different geographical locations. Figure 3 provides a visual representation of this spatial heterogeneity, highlighting the specific and often unique local relationships that exist between the independent variables and LST across our study area.

Fig. 3
figure 3

Spatial heterogeneity of impact of urban factors to the LST

Full size image

The Fig. 3 presents a Moran cluster map illustrating the spatial relationship between residential, commercial, and industrial densities, green spaces, and Land Surface Temperature (LST) in London, with a focus on the UHI effect.

For tree coverage High-High clusters suggest that areas with extensive tree coverage can still experience high LSTs, perhaps due to insufficient tree density or the influence of other urban factors. The presence of High–Low outliers indicates areas where increased tree coverage successfully reduces LST, highlighting the potential for trees to mitigate UHI effects. However, Low–High clusters show that areas with sparse tree coverage tend to have higher LSTs, underscoring the need for more greenery to help cool these urban regions. For green spaces: High-High clusters are often near Low–High outliers, indicating that green spaces can affect LST in adjacent areas. The City of London shows a Low-Low cluster due to limited green spaces, leading to higher LST, while areas like Battersea Park are High-High clusters. Increasing green spaces could be a key strategy in reducing LST and UHI effects. For residential areas, there’s a mix of High-High (dark red) and Low–High (light blue) clusters in boroughs like Southwark, suggesting that dense residential areas might influence LST nearby. This indicates a need for targeted urban planning to mitigate UHI in these areas. However, the presence of spatial outliers around clusters indicates no definitive spatial autocorrelation pattern. For commercial areas: Fewer clusters and more outliers (High-High and Low–High) are observed, especially in Southwark and the City of London, suggesting varied impacts of commercial density on LST. The spatial patterns indicate both positive and negative correlations in different areas. For industrial areas: There’s a noticeable spatial division between North and South London, with North showing low impact and South showing a higher impact of industrial density on LST. This suggests a region-specific influence of industrial areas on UHI.

Feature importance of LULC, urban forestry, and air quality to LST

The random forest regression model, as shown in Table 3, provides a quantitative assessment of the relative significance of various urban forestry and community land use variables in a predictive model.

Table 3 Model 2 feature importance of independent variables
Full size table

The Density of Industrial Areas is identified as the most influential factor, with a feature importance of 0.29, suggesting that industrial land use, characterized by its built environment and exposed surfaces, is a key driver in the model in predicting LST. The model attributes high importance to both the Density of Tree Coverage and the Density of Residential Areas, each with a feature importance value of 0.22, indicating their substantial impact within the model’s analysis. The Density of Green Space and the Density of Commercial Areas are given lower importance values of 0.14 and 0.12, respectively, indicating a moderate influence on the model’s outcomes.

Suitability analysis for Anti-UHI

The term “Anti-UHI” refers to strategies, measures, or phenomena that counteract or mitigate the UHI effect (Fierravanti et al. 2017; Hendel et al. 2016; Zhang and He 2008). Anti-UHI measures aim to reduce the temperature differential between urban and rural areas, thereby making urban environments cooler and more comfortable, especially during hot periods (Mushtaha et al. 2021). The relative importance of each factor, as derived from the Random Forest analysis, was used as weights in the suitability model.

The outcome of the ‘Weighted Overlay’ procedure exhibits a map delineating areas based on their suitability levels. As depicted in Fig. 4, the map employs a cluster counting technique to identify areas with varying levels of predicted LST. In this context, “cluster” refers to groups of spatially contiguous pixels that share similar LST values. Darker purple regions and higher cluster counting numbers signify elevated temperatures, suggesting a greater need for cooling interventions. Conversely, lighter shaded areas with lower cluster denote regions with lower LST, making them more suitable for habitation and less affected by UHI. The map provides a quantitative scale for the cluster counts, aiding in the interpretation of the data and identifying priority areas for Anti-UHI practices.

Fig. 4
figure 4

Suitability analysis for UHI

Full size image

The results from the map indicate UHI effect in highly urbanized zones, especially in the northern parts of the borough. Areas like the City of London, central Hammersmith and Fulham, central Westminster and central Hackney exhibit higher temperatures, indicating diminished Anti-UHI effects. In contrast, the less urbanized or more rural sections, predominantly in the southern parts of the borough, display a moderate suitability, suggesting a more balanced temperature profile. These findings are instrumental in pinpointing regions within the London Borough that offer optimal living conditions by effectively countering the UHI effect. Residents in areas with diminished UHI effects, such as the southern sections, may benefit from a more comfortable and sustainable urban living experience due to the mitigated heat stress.

In this suitability analysis aimed at countering the UHI effect, the importance of ‘Density of Tree Coverage’ becomes evident as a critical factor in reducing elevated temperatures within urban areas. Areas characterized by a denser tree canopy demonstrate a higher suitability for mitigating the UHI effect, ultimately leading to a more comfortable and livable urban environment. The presence of trees plays a pivotal role in enhancing urban resilience and promoting the development of green infrastructure. This information can guide urban planning and policy decisions aimed at enhancing urban resilience, green infrastructure development, and sustainable land use practices to create healthier and more livable environments throughout the borough.

Discussion

Discussion on findings

The findings of this research align with existing literature, reinforcing the established understanding of the relationship between urban and community forestry features and LST. The density, diversity, and health of urban forests are critical in maximizing this cooling effect. Healthy, mature trees with large canopies are particularly effective in reducing LST, highlighting the need for well-planned and maintained urban forestry (Donfack et al. 2021). The beneficial impact of urban forests in LST reduction is attributed to their shade provision, decreased solar radiation absorption, and the facilitation of evaporative cooling through transpiration processes (Oke 1989). Additionally, urban forests serve as critical ecological networks within cities, supporting biodiversity and providing essential ecosystem services beyond temperature regulation (Hu et al. 2021). Such findings underscore the importance of urban forestry practices in the context of managing urban heat and enhancing the overall livability of cities.

The findings of this research also align with existing literature that underscores the pronounced UHI effect in commercial areas (Connors et al. 2013; Mohan et al. 2013). Urban and different communities, especially commercial zones, characterized by dense infrastructure and limited vegetation, are particularly susceptible to higher temperatures. This vulnerability is linked to the prevalence of heat-absorbing materials like asphalt and concrete, coupled with limited green cover (Li et al. 2020). This underscores the potential of integrating urban forestry into commercial zones as a strategy to mitigate UHI effects. Strategic placement of trees and green spaces in these areas could provide substantial cooling benefits. It also underscores the well-established principle that urbanized zones, with their vast expanses of paved surfaces and heat-absorbing roofs, are more prone to heat retention compared to areas with natural landscapes (Andoni and Wonorahardjo 2018; Rotem-Mindali et al. 2015).

Moreover, this research brings to light the considerable role of forestry elements in mediating the UHI phenomenon in the London Borough. It becomes apparent that the amalgamation of dense commercial zones, high residential density, and a shortage of green spaces, particularly urban forests, significantly intensifies the UHI effect. This observation has profound implications for forestry management, public health, and environmental sustainability. The heightened temperatures due to the UHI effect can lead to health challenges, particularly during heatwaves, and elevate the vulnerability of urban populations (Zeren Cetin et al. 2023). Moreover, the increased demand for cooling in warmer urban environments can strain energy resources and contribute to higher energy consumption and associated environmental impacts (Wang et al. 2022a, b). In this context, urban forestry emerges as a key component of sustainable urban development, offering a nature-based solution to the challenges posed by the UHI effect. Recognizing the interplay of these urban factors underscores the importance of adopting strategies that prioritize green infrastructure, and sustainable land use practices to combat the UHI effect, promote public health, and enhance environmental sustainability within urban areas (Shen 2022).

Strategies for Anti-UHI development and significant to the urban forestry

The findings underscore the significant role of urban forestry in mitigating the UHI effect and enhancing urban environments. Among all the variables included in this study, urban forest, measured by tree coverage, emerges as the sole variable negatively correlated with urban heat. Urban planners and policymakers should prioritize the development and maintenance of urban forests by accurately assessing tree coverage and implementing a long-term strategic plan. While this study provides an example of using GIS to identify urban forests, it is important to explore other data-driven methods such as remote sensing, spatial image segmentation, and maintenance records for improved identification and maintenance (Zhang et al. 2023, 2022). Another critical strategy is to uphold long-term city planning and sustainability goals for urban forests, including setting targets for increasing tree canopy coverage over time and regularly monitoring progress. Long-term planning ensures that the benefits of urban forestry are sustained and contribute to reduced urban heat and heathy urban environments.

In addition, the findings of this research provide evidence and advocating for Anti-UHI strategies. Emphasizing the development of green spaces, especially in UHI hotspots like the City of London and central Hackney, can provide immediate relief. Increasing urban forestry density can act as a natural coolant, offsetting the heat generated by commercial and residential areas. Improving air quality, especially in industrial zones, can further mitigate the UHI effect. A re-evaluation of urban zoning, ensuring a balanced mix of different land uses, can distribute heat more evenly, preventing the formation of UHI hotspots.

To mitigate the UHI effect in the London Borough, comprehensive strategies and considerations are imperative. Anti-UHI development should focus on increasing urban green spaces, implementing cool roofing and paving, controlling industrial emissions, promoting mixed land use planning, engaging communities, enhancing public transportation, and integrating climate-responsive urban planning. Notably, urban forestry emerges as a key player in this endeavor. Urban forestry practices should prioritize tree coverage, diversify vegetation, expand tree canopies strategically, and involve the community in planting and maintenance initiatives. Monitoring and maintenance, community involvement, and long-term planning are essential for the sustained success of urban forestry initiatives. The presented research underscores the importance of interdisciplinary collaboration, policy advocacy, and public awareness in implementing these measures. The suitability analysis using the Random Forest Model provides a visual representation of areas with varying UHI effects, guiding urban planning decisions for healthier and more livable environments. While the study identifies limitations and suggests avenues for future research, its findings align with existing literature, emphasizing the critical role of urban features, particularly green spaces and tree coverage, in mitigating the UHI effect and fostering sustainable urban development.

Understanding the UHI phenomenon is a complex endeavor, necessitating a multifaceted approach. This research, by delving deep into the various variables and their interplay, provides a blueprint for urban planners, policymakers, and public health officials. As cities continue to grow, addressing the UHI challenge becomes paramount, and this study offers the findings for urban planning, public health, and environmental sustainability.

Limitations and future directions

While our framework marks a significant advancement in UHI mitigation strategies, it is not without limitations. The reliance on available data sets and the potential for variability in remote sensing accuracy. The urban forestry dataset we conducted cover the street trees and trees in public urban green space. However for some trees in private gardens and green roofs is insufficient cover. While these are a small fraction compared to the public urban forest, their contribution to the city’s UHI is still not negligible. Additionally, the practical implementation of our strategies depends on policy support, funding availability, and community buy-in, factors that can vary widely across different urban settings.

Future research should focus on expanding the framework’s applicability to a broader range of urban contexts, including cities with differing climatic conditions and urban morphologies. Exploring the integration of emerging technologies, such as machine learning for predictive modeling of UHI dynamics, could further enhance the precision and effectiveness of urban forestry strategies (Doick et al. 2014; Sun and Chen 2017). Moreover, longitudinal studies are needed to assess the long-term impacts of implemented green infrastructure on urban temperature patterns, public health, and social well-being (Parker 2010; Sun et al. 2022).