Designing multisource blue–green cooling networks by coupling landscape pattern metrics and circuit theory

Abstract

The surface urban heat island effect is a growing environmental concern amid accelerating urbanisation and climate change, posing significant threats to sustainable development and public health. While urban green spaces are recognised as effective in mitigating the surface urban heat island effect, most studies focused on isolated patches or local interventions, and thus lack a systematic, network-based cooling strategy. This study proposes an integrated framework of “cool-source identification–ventilation corridors–network integration”, combining multi-source remote sensing data (2003–2022), the geographic detector model, morphological spatial pattern analysis, and circuit theory to assess surface urban heat island dynamics and construct an urban green cooling network. Focusing on Nanchang’s central urban area, results indicate that from 2003 to 2022, the proportion of low-temperature zones decreased from 41.69% to 19.67%, while high-temperature zones increased from 8.69% to 19.89%; correspondingly, the surface heat island imbalance index rose from − 0.491 to 0.005. The interaction between the normalised difference vegetation index and land use explained over 60% of the surface temperature variation. A green ecological network comprising 56 primary and 60 secondary ventilation corridors was established, significantly enhancing connectivity and heat dissipation potential. Building on previous studies, this network demonstrates the capacity to reduce daytime temperatures by 1–3 °C in densely built-up areas. The findings highlight that optimising the spatial configuration and connectivity of green spaces can maximise cooling benefits under land constraints. This blue–green ventilation integrated approach provides transferable insights for humid, flat and rapidly urbanising regions.

Introduction

Background of study

The surface urban heat island (SUHI) effect refers to the phenomenon in which urban built-up areas experience higher temperatures than surrounding rural areas due to increased impervious surfaces and anthropogenic heat emissions. Since Luke Howard’s initial observation of London’s SUHI in the 19th century¹, it has remained a central topic in urban climatology. Currently, 55% of the global population resides in urban areas, a figure projected to rise to 68% by 2050²(United Nations Department of Economic and Social Affairs, 2019). Rapid urbanisation has intensified SUHI effects worldwide. According to the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report, the interaction of SUHI with global warming significantly elevates heat-related risks in urban zones, where temperatures in some city centres can exceed those of nearby rural areas by 3–12 °C³.

Globally, SUHI contributes to urban overheating by amplifying ambient temperatures, increasing energy demands for cooling and exacerbating heat stress, particularly in low-income and rapidly urbanising regions. Recent satellite-based analyses indicate that SUHI intensity has risen more rapidly in lower-income countries compared to higher-income ones, with an average annual increase of 0.021 °C globally from 2003 to 2018, driven primarily by rapid impervious surface expansion and limited green infrastructure⁴. In tropical and subtropical zones, higher rates of intensification are observed due to unplanned urban growth and vegetation loss. In humid climates, high baseline humidity further diminishes evaporative cooling from remaining green spaces, intensifying perceived heat stress during extreme events^5,6. These patterns underscore the need for context-specific mitigation in vulnerable urban areas.

A recent satellite-based thermal analysis of more than 3,000 cities across 171 countries revealed that over 70% of medium and large cities now exhibit pronounced SUHI impacts, particularly in rapidly urbanising regions of China. For example, in Nanchang, the proportion of extremely low-temperature zones declined from 4.39% in 2001 to just 0.77% in 2020, while extreme high-temperature zones surged during the same period⁷. SUHI not only increases energy use and carbon emissions but also directly threatens public health. According to the latest WHO (2023) report, extreme heat contributes to approximately 166,000 deaths annually, disproportionately affecting vulnerable urban populations⁸.

A fundamental distinction exists between SUHI and the canopy-layer urban heat island. SUHI measures land surface temperature (LST) from satellite data, providing high spatial coverage suitable for large-scale pattern analysis, but it is limited to clear-sky daytime conditions and may overestimate temperatures in vegetated areas due to emissivity variations. Canopy-layer Urban Heat Island (CUHI) measures near-surface air temperature using ground stations, offering direct relevance to human thermal comfort but with limited spatial representation. This study focuses exclusively on SUHI due to its compatibility with remote sensing-based landscape pattern analysis over extended temporal and spatial scales^9,10.

Urban nature-based solutions (NbS), such as green roofs, urban forests, parks, and blue–green corridors, are increasingly recognised as effective measures for mitigating SUHI through enhanced evapotranspiration, shading and airflow improvement^11,12. In dense urban contexts, opportunities for large-scale greening are limited. Therefore, maximising the cooling potential of existing green spaces through a networked spatial approach has become a research frontier.

Literature review

Urban green infrastructure is widely regarded as one of the most cost-effective and synergistic approaches to mitigating SUHI. Vegetation cools the environment through shading and evapotranspiration, while water bodies provide cooling via latent heat absorption and enhanced convection^13,14. Together, these create urban “cool islands”. A global meta-analysis of 131 quantitative studies found that increasing urban canopy cover can reduce near-surface temperatures by 2.0–3.1 °C, with cooling performance influenced by tree height, leaf area index and background climate¹⁵. Additionally, urban impervious surfaces are approximately 3 °C warmer than green areas¹⁶, and canopy structure plays a crucial role in cooling, especially during midday and at night¹⁷. In tropical contexts, differential vegetation types, such as multi-layered forests, exhibit stronger cooling than simple grasslands, emphasising structural complexity^18,19.

Recent studies demonstrated that while fragmented green spaces provide only localised cooling effects, well-connected green–blue networks can link multiple cooling sources within 200–500 m, thereby amplifying regional cooling²⁰. Cold island networks constructed using circuit theory have shown the potential to reduce daytime temperatures by an additional 1–3 °C in several cities⁷(Liu et al., 2024). Empirical evidence from Shenzhen also confirms the effectiveness of corridor planning based on circuit theory in coastal hilly cities²¹. Urban form further modulates these effects. For instance, higher building density and morphology can influence SUHI intensity through shading and wind patterns^22,23. To contextualise this study, Table 1 summarises key literature on SUHI mitigation, highlighting methods, findings and gaps addressed by the present work.

Table 1 Summary of key literature on SUHI mitigation.

Nonetheless, three key research gaps remain: (i) the absence of system-level evaluations across multiple green patches in plain humid cities, (ii) limited understanding of dynamic heat dissipation and airflow in land-constrained contexts, and (iii) insufficient actionable, multifactor-integrated strategies for compact urban environments. These gaps highlight the need for an integrated framework that couples landscape pattern metrics with circuit theory. The present study addresses these gaps by proposing a transferable approach tailored to subtropical humid regions, with cross-regional validation to enhance applicability.

Research novelty and objectives

This study introduces several unique contributions to SUHI mitigation:

• (i) An integrated framework coupling geographic detectors, morphological spatial pattern analysis (MSPA) and circuit theory for multi-source cooling network design.

• (ii) Application to a humid subtropical plain city lacking natural topographic ventilation.

• (iii) Preliminary application and validation of the framework in a tropical context using literature-based comparison to enhance transferability.

• (iv) Multifactor-integrated countermeasures that explicitly incorporate natural, built- environment and land-surface.

Given Nanchang’s characteristics of rapid urbanisation, hot-humid climate, and diverse green–blue spaces, this study selects its central urban area by integrating multi-source remote sensing, geographic detector models, MSPA and circuit theory. The study aims to:

• (i) Analyse the spatiotemporal evolution of SUHI from 2003 to 2022.

• (ii) Identify dominant natural and anthropogenic drivers of thermal patterns.

• (iii) Extract key cooling-source green spaces and assess their cooling effects.

• (iv) Simulate potential ventilation corridors using resistance surfaces and circuit theory to construct a green cooling network.

• (v) Propose strategies for enhancing green connectivity, introducing stepping-stone micro-greens and protecting ventilation paths.

The ultimate goal is to provide a transferable framework and planning reference for SUHI mitigation in Nanchang and other subtropical humid cities.

Methods

Study area and data

Study area overview

Nanchang City (28°36′N–29°11′N, 115°45′E–116°15′E) is located in north-central Jiangxi Province, eastern China, situated along the lower reaches of the Gan River and at the south-western edge of Poyang Lake. It has a subtropical monsoon climate characterised by hot, rainy summers and mild, dry winters. In recent decades, the city has undergone rapid urban expansion. Between 1990 and 2020, its built-up area increased by 433.29 km², while cultivated land decreased by 291.99 km² (Zhang et al., 2023). As urbanisation reduced natural vegetation and water bodies, surface temperatures have risen citywide, with increasingly frequent extreme summer heat. The number of high-temperature days (daily maximum ≥ 35 °C) rose from 25.3 days in 2000 to 41.7 days in 2020, making Nanchang one of the most heat-affected cities in China (Liu et al., 2024).

The study area was delineated based on the urban spatial structure defined in the Territorial Spatial Master Plan of Nanchang (2021–2035)²⁴, covering the Central Activity Zone, core urban districts and surrounding sub-centres (Nanchang Municipal People’s Government, 2024). The total area is approximately 1,710 km² and includes historical urban cores, newly developed suburban towns, wetlands and lakeside ecological zones. This configuration reflects the nested spatial pattern of “core urban area–suburban new towns–ecological base”. The boundary was digitised through a spatial overlay of the official planning map and remote sensing data (Fig. 1).

For SUHI intensity computation, the reference zone was defined from non-urban portions within the study area itself to ensure local climatic consistency. Non-urban zones were identified using 30 m resolution annual land cover data²⁵. Specifically, pixels classified as “Cropland”, “Forest”, “Grassland” and “Water” were aggregated to represent the rural/natural baseline, while pixels classified as “Impervious Surface” were used to define the urban/built-up areas. While the Local Climate Zone (LCZ) scheme was consulted to conceptually distinguish natural landscapes, a strict LCZ-based classification was not applied for the quantitative calculation in this study.

Fig. 1

Geographic location and extent of the study area in Nanchang City, Jiangxi Province. The map shows the administrative layout and spatial coverage of the study area based on the Nanchang Territorial Spatial Master Plan (2021–2035)²⁴. This figure was created using ArcGIS Desktop v10.8 software (https://www.esri.com, Environmental Systems Research Institute, Inc., Redlands, United States).

Data sources and preprocessing

This study integrated multi-source remote sensing imagery, meteorological observations and urban planning data to construct a comprehensive dataset covering the thermal environment, landscape patterns and socio-economic factors. Table 2 presents information on the Landsat images used for spatiotemporal analysis of the thermal environment, while Table 3 summarises additional data sources and preprocessing outputs.

Table 2 Landsat data information.

Table 3 Multi-source data, preprocessing and factor selection rationale.

LST data were obtained directly from the USGS Landsat Collection 2 Level-2 (L2) surface temperature products, which measure land surface temperature (skin temperature) rather than near-surface air temperature. These datasets have been pre-processed by the USGS using standardised workflows, including radiometric calibration, atmospheric correction and geometric correction, ensuring high consistency and comparability across sensors and acquisition periods²⁶. As a result, no additional atmospheric, geometric or topographic corrections were performed in this study.

To minimise the influence of cloud cover and precipitation, only scenes with less than 10% cloud cover under clear-sky conditions were selected. Image acquisition dates were chosen to represent periods with favourable atmospheric conditions, predominantly during peak summer (July), which is most suitable for capturing SUHI characteristics. Due to data availability constraints and the need to maintain approximately five-year intervals for comparative consistency, imagery for 2008 was acquired in late spring (May). The year 2022 was selected as the most recent study period instead of 2023, owing to the unavailability of complete summer-season imagery at the time of analysis.

This seasonal discrepancy for 2008 may introduce potential bias in temporal comparisons of SUHI intensity, as surface temperatures in late spring are generally lower than those in peak summer. However, this effect is mitigated by employing relative LST classification methods based on mean and standard deviation thresholds, which are widely used to reduce seasonal and atmospheric variability in multi-temporal Landsat analyses^37,38,39. In addition, sensitivity analyses (e.g., excluding the 2008 dataset and comparing trends across the remaining years) confirmed that the overall spatiotemporal trends in SUHI intensity and thermal patterns remain robust⁴⁰. All raster datasets were resampled to a unified 30 m spatial resolution using bilinear interpolation and clipped to the study area boundary to ensure spatial consistency.

In addition to satellite and official spatial data, this study also utilised 30 m resolution annual land cover data compiled by Yang and Huang (2024)²⁵, as well as a 100 m gridded population density dataset from Chen et al. (2024)³⁵. Both datasets were standardised and integrated into the analytical framework.

Research methods

To achieve the research objectives, this study proposes a five-stage integrated analytical framework that couples multi-source data processing, thermal environment analysis, driving factor detection, cooling source identification and network optimisation (Fig. 2). This framework enables a systematic transition from cool-source identification to ventilation corridor simulation and final network integration, providing a transferable approach for SUHI mitigation in humid subtropical cities.

Fig. 2

Flowchart illustrating the five-stage integrated analytical framework, from data acquisition and preprocessing to cooling network construction and optimisation.

Spatiotemporal analysis of urban thermal environment

Based on the distribution of LST, the study area was classified into five thermal zones: extreme low (below mean − 1.5 Standard Deviation (SD)), low (mean − 0.5 to − 1.5 SD), moderate (± 0.5 SD), high (mean + 0.5 to + 1.5 SD) and extreme high (above mean + 1.5 SD). This classification approach, using ± 0.5 SD and ± 1.5 SD thresholds, is widely adopted in SUHI studies to effectively delineate moderate and extreme thermal zones while maintaining statistical robustness across multi-temporal datasets⁴¹. Temporal changes in the proportion of each zone were analysed. Spatial autocorrelation and hotspot analysis were conducted to identify hot and cold clusters, and their areal change and spatial migration were calculated to reveal SUHI pattern dynamics (Table 4).

Table 4 Temperature zone classification Method.

$$\:SHII=\:\frac{{A}_{high}-{A}_{low}}{{A}_{total}}$$

where $\:{A}_{high}$ is the total area of high-temperature zones, $\:{A}_{low}$ is the total area of low-temperature zones and $\:{A}_{total}$ is the total study area. This index complements the spatiotemporal analysis of SUHI patterns.

Identification and quantification of SUHI driving factors

To investigate the drivers behind urban thermal patterns, the Geodetector model was employed to quantify the influence of potential factors on LST spatial variability^42,43,44. This model, based on spatial heterogeneity theory, assesses how well a factor explains temperature variance using a q-statistic ranging from 0 to 1. A higher q-value indicates stronger explanatory power; q = 1 denotes full explanation, while q = 0 implies none.

Candidate factors included natural variables (elevation, slope, aspect), built environment features (building height, density, road and population density), and land surface characteristics (land use type, NDVI). Each factor was stratified using equal-interval or natural breaks classification. The study area was partitioned into subregions based on these strata, and the q-value for each factor was calculated as the weighted sum of intra-region LST variance relative to total variance. This allowed identification of key drivers shaping spatial temperature patterns:

$$\:q=1-\frac{{\sum\:}_{h=1}^{L}{N}_{h}{\sigma\:}_{n}^{2}}{N{\sigma\:}^{2}}$$

where $\:{N}_{h}$ is the number of samples in subregion h, $\:{\sigma\:}_{h}^{2}$ is the variance of LST within subregion h, and $\:{\sigma\:}^{2}$ is the variance of LST across the entire study area. This formulation enables the identification of key drivers shaping spatial temperature patterns.

Additionally, the Geodetector can evaluate the interaction between paired factors to determine whether their combined impact on temperature is enhanced, independent or weakened. In this study, particular attention was paid to the ranking of individual q-values and the interaction effects among dominant factors. For example, the joint influence of vegetation and land use type or building density was assessed to examine whether their combined explanatory power exceeded that of individual factors. This analysis provides insight into the complex causes of urban thermal environments and supports the design of multifactor intervention strategies.

Identification of green space cooling sources via MSPA

MSPA was applied to identify ecological units from the 2022 land use map of Nanchang City that are critical for urban cooling sources. These green space cooling sources refer to areas, such as forests, parks and wetlands, that can significantly improve local temperature conditions. MSPA extracts the spatial structure of green spaces by binarising land use data and applying an eight-neighbour pixel connectivity analysis. This process identifies six morphological types of landscape elements: core, islet, perforation, edge, loop and bridge⁴⁵(Table 5). Each type plays a distinct ecological role and contributes to the connectivity of the landscape.

The core areas represent relatively intact green spaces with minimal external disturbance and high potential ecosystem service value. Islets refer to isolated small patches with low connectivity, limiting ecological flow exchange. Perforations are small blank patches within core areas that restrict ecological flows. Edge zones serve as transitional areas between core and non-core regions, often participating actively in ecological processes. Loops are peripheral corridors surrounding core areas, maintaining internal ecological flow. Bridges connect two or more core areas, facilitating ecological exchange between adjacent patches.

Based on the extracted green space elements, a Cooling Index (CI) was constructed to quantify the cooling capacity and spatial contribution of each type of green space element. The CI is calculated as follows:

$$\:CI=(\overline{{LST}_{a}}-{LST}_{mean})\frac{{S}_{a}}{S}$$

where $\:\overline{{LST}_{a}}$ represents the mean LST of the target spatial unit, $\:{LST}_{mean}$ denotes the average LST across the study area, $\:{S}_{a}$ is the area of the target spatial unit, and S represents the total area of the study area.

A higher CI value indicates a greater cooling contribution of the corresponding green space element, resulting from a combination of lower $\:\overline{{LST}_{a}}$ and larger $\:{S}_{a}$. Conversely, a lower CI value suggests limited cooling capacity, which may be due to a small area or high isolation of the element.

Additionally, a landscape connectivity importance index was introduced to evaluate the role of green space elements in the overall ecological network. The spatial layout and functional connectivity of different green space units were analysed to identify critical nodes, such as bridges and loops, which not only have cooling potential but also act as hubs for ecological flow within the network.

Table 5 Morphological characteristics and ecological significance of green space elements.

Construction of the green space cooling network and application of circuit theory

After identifying key cooling sources, circuit theory-based modelling was applied to simulate potential pathways of cool air diffusion and construct an urban cooling network connecting those sources. Originating from ecological connectivity analysis, circuit theory models the landscape as a conductive surface with resistance properties, quantifying spatial connectivity by simulating the flow of “current”. In the context of urban thermal environments, the flow of cool air is analogous to an electrical current, while urban surfaces act as “resistance surfaces” that hinder its movement.

The resistance surface was generated using principal component analysis (PCA), incorporating key driving factors identified by the Geodetector model. Areas with high building density and impervious surfaces, which retain heat and obstruct air flow, were assigned high resistance values. In contrast, open water bodies and green spaces received low resistance values due to their facilitative effect on cool air movement. Higher elevation and steep slope areas were also assigned low resistance owing to cooler temperatures and better airflow.

Using the resistance surface, major cooling-source patches were designated as current injection points (simulating cold air origins). The Circuitscape software was employed to simulate multi-path cool air diffusion and calculate the cumulative current intensity for each grid cell, which represents the frequency of cool air passage. Areas with high cumulative current values indicate potential ventilation corridors.

Circuit theory also identifies network bottlenecks, critical nodes where air pathways are narrow and surrounded by high resistance, with few alternative routes. These bottlenecks are marked by high current density, defined as the intensity of current passing through a single cell. Removing these nodes would disrupt network connectivity and reduce overall cooling effectiveness.

To clarify the structure and logic of this modelling approach, Fig. 3 illustrates the resistance network and current flow structure: Fig. 3A shows the resistance network model, with low-resistance patches as nodes and resistive links as circuit elements; Fig. 3B depicts current flow and bottleneck identification, highlighting primary air corridors and critical points where connectivity may fail.

Fig. 3

Integrated schematic of resistance network and current flow. (3A) Resistance Network Model, showing nodes representing low-resistance patches and resistive links forming a landscape circuit. (3B) Current Flow and Bottleneck Identification, illustrating current pathways and critical nodes where connectivity could be disrupted if removed. Concept and methodology adapted from McRae et al. (2008)⁴⁶.

Finally, the ventilation corridor network was further optimised based on land use and road layouts. For high-current connectivity locations lacking greenery, small green patches or “stepping-stone” parks were proposed to fill gaps. In areas with existing corridors but obstructed airflow (e.g., dense building clusters), interventions such as corridor openings, façade perforation or vegetative retrofitting were suggested to reduce resistance. The final ecological cooling network, therefore, consists of cooling sources (major green and water bodies) as nodes, ventilation corridors as links, bottlenecks as critical restoration targets and stepping-stone greens as auxiliary connectors, forming a complete urban cooling framework.

Results

Spatiotemporal characteristics of the urban thermal environment

Spatially (Fig. 4), in 2003, high-temperature patches were mainly concentrated in the old urban commercial centres, relatively small in size and scattered. By 2008, the high-temperature zones had expanded considerably and extended in strips along the urban edge, indicating the impact of suburban development on localised warming. In 2013, new high-temperature clusters emerged in the newly developed areas west of the Ganjiang River, with the SUHI pattern evolving from a single-core to a multi-core structure. At this stage, suburban areas also began to experience heat intrusion. In 2018 and 2022, the entire south-western area (including Honggutan and Xinjian Districts) became a high-temperature hotspot, with contiguous distributions of high-temperature zones and almost no cold zones remaining in the central urban area. The heat pattern tended to stabilise but became highly polarised.

Fig. 4

Spatial distribution of temperature zones in Nanchang City from 2003 to 2022. The figure illustrates spatial patterns of different temperature zones for each study year. This figure was created using ArcGIS Desktop v10.8 software (https://www.esri.com, Environmental Systems Research Institute, Inc., Redlands, United States).

This evolution can be broadly divided into two stages. The first period was a rapid expansion phase (2003–2013), during which the area of high-temperature zones increased by approximately 40.9%, while the area of low-temperature zones contracted by about 72.6%. This phase coincided with Nanchang’s “One River Two Banks” urban expansion peak. The second was an intensity enhancement period (2013–2022), during which the proportion of high-temperature zones increased by a further 7.65%, although the rate of real growth slowed to approximately 62.5%. This suggests that SUHI deterioration during this phase was primarily driven by intensified development rather than extensive spatial expansion, possibly related to dense development spurred by transit infrastructure construction during this period.

Overall, the SUHI in Nanchang’s central urban area has intensified significantly, with the surface heat island index (SHII) rising from − 0.491 in 2003 to 0.005 in 2022. This reflects a transition from weak to strong SUHI conditions, and from isolated heat spots to widespread heat dominance, with severe conditions now prevailing. Correspondingly, the area and proportion of different temperature zones and SHII values in Nanchang City from 2003 to 2022 (Table 6).

Table 6 Area and proportion of different temperature zones and SHII values in Nanchang City from 2003 to 2022.

It is noteworthy that the thermal characteristics of different land cover types exhibit significant variation. According to the 2022 land use statistics, nearly 80% of pixels in the High Temperature Zone and Sub-high Temperature Zone consist of impervious surfaces, with built-up land dominating these areas. In contrast, water bodies and green spaces predominantly occur in the Low Temperature Zone and Sub-middle Temperature Zone, forming the cooling landscape. For example, in 2022, high-temperature hotspots were mainly associated with dense urban areas and regions along major transportation corridors, while areas such as Aixi Lake, Yaohu Lake and the Meiling Mountains maintained lower temperatures owing to their abundant green and water resources, creating localised “cool islands”.

This pattern underscores a strong correlation between underlying surface properties and SUHI intensity: continuous impervious surfaces substantially increase LST, whereas large contiguous green spaces and water bodies help to maintain cooling islands and reduce surrounding heat loads. Over the past two decades, the expansion of high-temperature zones in Nanchang has largely coincided with the conversion of Cropland into urban construction land (Fig. 5). Therefore, controlling the expansion of impervious surfaces while protecting green and water areas constitutes a key strategy to mitigate the worsening SUHI effect.

Fig. 5

Land use change in Nanchang City from 2003 to 2022. The figure illustrates land use types and their changes, highlighting the reduction of cooling landscapes and the expansion of impervious surfaces. This figure was created using ArcGIS Desktop v10.8 software (https://www.esri.com, Environmental Systems Research Institute, Inc., Redlands, United States) based on land cover data from Yang and Huang (2024)²⁵.

Driving factors of the urban heat Island effect

Geodetector analysis was employed to quantify the impact of environmental factors on the spatial distribution of LST in Nanchang. The results indicate that the NDVI and land cover are the dominant drivers of SUHI formation. According to the factor detection results, NDVI exhibited the highest explanatory power for LST variation (q = 0.500), followed by land cover (q = 0.463) (Table 7).

Table 7 Factor detection results for LST variation in Nanchang.

Table 7 summarises the q-statistics of individual factors and their significance levels. Among the nine environmental variables analysed, aspect was the only factor without statistical significance (p = 0.501), suggesting minimal influence on urban thermal patterns. The other eight factors demonstrated highly significant associations with LST (p < 0.001), with the explanatory power ranked as follows: NDVI (0.500) > Land Cover (0.463) > Building Height (0.326) > Building Density (0.277) > Population Density (0.262) > Road Density (0.140) > Slope (0.116) > Elevation (0.110).

This finding underscores the joint role of vegetation and surface characteristics in shaping the urban thermal landscape. High vegetation coverage and proximity to green or water areas are associated with lower local temperatures, whereas densely built-up areas correspond to higher temperatures. Notably, building density (or impervious surface ratio) and population density also exhibited substantial explanatory power (q ≈ 0.25–0.35), highlighting the contribution of anthropogenic factors to SUHI effects. For instance, areas with high building density tend to experience elevated temperatures, though vegetation interspersed within can mitigate the warming effect.

Topographic factors, such as elevation and slope, had limited explanatory power in the generally flat terrain of the study area. However, local highlands (e.g., Meiling Mountains, elevation > 200 m) showed localised cooling effects. The role of road density in SUHI formation depended on the surrounding land use, as areas with dense road networks often corresponded to high-intensity urban development.

Importantly, multiple environmental factors exhibited interactions, with NDVI showing a strong compensatory influence when combined with built-environment variables, particularly building density, highlighting the mitigating role of green space within densely built areas. As illustrated in Fig. 6, the off-diagonal elements represent interaction q-values derived from the GeoDetector interaction detector, while the diagonal elements indicate the explanatory power of individual factors. Notably, all significant factor pairs exhibited enhancement effects, with interaction q-values, especially those combining natural and built-environment factors, exceeding the maximum single-factor q-values, thereby confirming a pronounced bivariate enhancement effect. For example, NDVI exhibited a single-factor q-value of 0.500, while its interaction q-values with land cover, building height and building density increased to 0.618, 0.570 and 0.555, respectively, clearly illustrating this enhancement pattern.

Fig. 6

GeoDetector q-statistics interaction matrix for LST variation.

The diagonal elements represent the explanatory power (q-values) of individual factors obtained from the factor detector, whereas the off-diagonal elements indicate the q-values of pairwise factor interactions derived from the interaction detector. Warmer colours denote stronger interaction effects on LST variation.

These results demonstrate that the urban thermal environment is shaped by the coupling of natural and anthropogenic factors, including green–grey space configuration and the intensity of human activities. Increasing green coverage, reducing impervious surfaces and lowering anthropogenic heat emissions (e.g., traffic and air conditioning) are fundamental to mitigating SUHI effects. This quantitative identification provides a basis for focusing on key drivers in the subsequent ecological network optimisation, which will target “greening and grey reduction” and “ventilation and heat dissipation” as core strategies.

Ecological cooling patches and cooling effects

Through MSPA analysis combined with cooling efficiency assessment, several ecological cooling patches were identified in the central urban area of Nanchang. These patches primarily include large urban parks, lake wetlands and suburban green spaces, totalling 123 patches distributed across major urban directions. According to the MSPA results, core patches account for approximately 15% of the urban area, and these patches are predominantly covered by vegetation and water bodies, serving as key sources of cooling. Among these, the Meiling Forest and its surrounding areas represent the largest cooling patch (approximately 81 km²), followed by the Gan River (approximately 65 km²), Yaohu Lake (core and edge, totalling about 15.7 km²) and Aixi Lake Wetland (approximately 3.7 km²). The core patches are spatially continuous and intact, acting as “temperature buffer zones” within the urban thermal environment. Overlay analysis with LST data showed that the average surface temperature within these core patches was 2–4 °C lower than the surrounding urban areas, with cooling radii extending several tens of metres.

Beyond the core patches, MSPA identified perforations, bridges and edge patches. The edge patches (CI = − 0.0153) serve as transition zones between the core patches and high-temperature areas, providing significant temperature buffering effects. Perforations exhibited a weak cooling effect (CI = − 0.0045), with average temperatures between those of the core zone (37.88 °C) and the edge zone (40.66 °C). The per-unit-area CI (–0.0011/km²) was notably lower than that of the core patches (–0.0018/km²), suggesting that these small non-green units may weaken the overall cooling capacity, becoming “hot spots within cool islands” and causing localised thermal disturbances. Bridges (CI = 0.0366), acting as linear elements, formed “heat corridors” due to their narrow width and thermal infiltration from adjacent areas (Table 8).

Table 8 Cooling patch metrics in Nanchang.

The connectivity importance index, quantified as the delta probability of connectivity (dPC), was calculated for each candidate cooling patch, and based on thresholds of 0.5 and 1, patches were categorised into three connectivity levels. Patches with higher connectivity were mainly concentrated in the core zone, particularly Meiling, which had the highest dPC value (59.4508), attributed to its large area, wide distribution and strong connectivity with surrounding patches. The Gan River showed slightly lower connectivity due to interruptions by bridges, but still maintained a relatively high level. Several large urban water bodies, such as Qingshan Lake, Aixi Lake, Yaohu Lake, Xiang Lake, Xiazhuang Lake, Qianhu Lake and Jiulong Lake, demonstrated good connectivity due to their large areas and intact shapes. In addition, smaller patches, such as Xiaqiling, Dashitou and Rural Lake, also exhibited strong connectivity.

Overall, patches with high connectivity primarily include mountainous forests, river water bodies and urban parks with large water bodies. Some small green spaces, such as Bayi Park in the city centre (approximately 10 hectares), although limited in area, are located in high-temperature zones and exhibit significant cooling efficiency, thus also classified as cooling patches. Considering both cooling effects and network connectivity, six primary cooling source nodes were ultimately identified, including Meiling, Gan River, Qingshan Lake, Aixi Lake, Yaohu Lake and Rural Lake. The remaining patches were classified as secondary or tertiary cooling nodes based on their connectivity and thermal contribution indices (Fig. 7). These cooling patches collectively form the foundation of Nanchang’s urban cooling network, supporting the subsequent implementation of strategies focusing on “greening and grey reduction” and “ventilation and heat dissipation”.

Fig. 7

Spatial distribution of cooling patches in Nanchang. This figure was created using ArcGIS Desktop v10.8 software (https://www.esri.com, Environmental Systems Research Institute, Inc., Redlands, United States).

Cooling corridors and ecological network pattern

After identifying the locations of cooling sources, circuit theory modelling was employed to simulate the potential cooling corridors in the central urban area of Nanchang. The cumulative current map (Fig. 8A) revealed multiple high-intensity “current” pathways traversing the city, effectively connecting major cooling source nodes. This indicates the optimal locations for ventilation and cooling corridors, while also identifying key centrality areas, ecological pinch points (Fig. 8B) and barrier points (Fig. 8C).

Subsequently, existing land use types and the road network were integrated with the circuit model results to optimise the cooling corridors. A total of 116 corridors were extracted, including 56 major cooling corridors and 60 general corridors. Corresponding optimisation strategies were also proposed. Primary cooling sources, such as Qingshan lake, Yaohu lake, Aixi lake, Meiling and Gan River, exhibit a solid ecological foundation and serve as the core for urban thermal environment regulation. It is recommended to enhance ecological protection and buffer zone construction in these areas to maximise their regional cooling effects. Secondary sources, including Xiang lake, Jiulong lake and Bayi Park, require strengthened ecological restoration and boundary management, with the establishment of localised “cooling chains” through green corridors. Tertiary sources, mainly consisting of community green spaces and peripheral plots, should focus on the conversion of underutilised land and the integration of embedded greenery to construct micro-scale ecological networks.

Major cooling corridors connect primary cooling sources, and it is suggested that their continuity and crossing capacity be enhanced through measures, such as bridge greening and vertical greening. General cooling corridors should emphasise fine-scale management and public participation to improve their synergistic effects. Collectively, these measures aim to construct a structurally stable and functionally comprehensive urban cooling ecological network (Fig. 8D).

Fig. 8

Cooling corridors and ecological network patterns in Nanchang. 8 A: Cumulative current map of cooling corridors. 8B: Pinch point and centrality distribution.

8 C: Barrier point distribution. 8D: Optimised layout of cooling corridors and the ecological network.

This figure was created using ArcGIS Desktop v10.8 (https://www.esri.com, Environmental Systems Research Institute, Inc., Redlands, United States) and Circuitscape v4.0 (https://circuitscape.org, developed by Brad McRae, Viral Shah, Tanmay Mohapatra, and Ranjan Anantharaman, Circuitscape Project, University of California, Santa Barbara, United States).

Discussion

Cooling benefits and mechanisms of the green space ecological network

This study confirms that linking fragmented green spaces and water bodies through an ecological network can establish a synergistic cooling system at the urban scale. This finding is consistent with previous research, showing that the spatial configuration of urban greenery significantly influences thermal mitigation and that highly connected green networks provide greater cooling services than isolated patches. Through an ecological network, cool air that would otherwise be confined within individual green spaces can be channelled along corridors to extend its influence across larger distances. Particularly at night, connected green corridors help to facilitate “cool island ventilation” between urban and rural zones, drawing cool air from the suburbs into the urban core and suppressing nighttime SUHI intensification.

This mechanism is evident in the modelled Ganjiang riverside corridor, where the Gan River functions as a natural ventilation channel linking large-scale water and green resources from the northern and southern suburbs, significantly reducing thermal load in adjacent urbanised areas. Moreover, the network helps to eliminate “thermal discontinuities”, whereas isolated patches only affect their immediate surroundings. A linked network allows local cooling effects to merge into a region-wide chain of cool islands, enhancing the system-wide impact of individual green elements.

Previous studies have shown that the shape complexity and spatial connectivity of green spaces significantly affect LST. Guo et al. (2020) demonstrated, through a multi-city analysis in China, that increased connectivity and morphological complexity enhance cooling performance, while in fragmented landscapes the marginal cooling benefits of additional green space diminish⁴⁷. Similarly, Li et al. (2024) provided empirical validation for a “multi-level urban ecological cooling network” in over 50 cities, confirming that networked green configurations significantly boost collaborative cooling effects²⁰.

The findings from Nanchang further support this conclusion: the interaction between vegetation coverage and land use type proved to be the strongest explanatory factor for surface temperature variation, reinforcing the importance of spatially networked green layouts. When green spaces are isolated by large impervious areas, their cooling effect remains localised; in contrast, embedding greenery within continuous open spaces enables the diffusion of cool air and the lowering of temperatures across broader urban zones. This aligns with international evidence, such as in Germany, where green infrastructure in Munich provides micro-scale heat mitigation through connected urban elements⁴⁸and in Ethiopia, where parks in Addis Ababa effectively reduce UHI intensity through spatial connectivity⁴⁹.

Thus, optimising the spatial distribution of green infrastructure is more effective than merely increasing green area. The concept of networked planning represents a paradigm shift in SUHI mitigation, from isolated interventions to holistic spatial strategies. This is particularly vital for high-density cities, such as Nanchang, where land resources are limited. By consolidating existing green assets and filling critical gaps, substantial cooling gains can be achieved.

Planning implications and multifunctional synergies

Constructing a green space ecological network not only helps to reduce urban temperatures but also improves overall environmental quality. First, the cooling network aligns with urban ventilation corridor systems and provides co-benefits for air circulation and pollutant dispersion. A well-designed green ventilation corridor can both transport cool air and act as an “airway” to reduce the accumulation of pollutants such as PM2.5, thereby achieving dual environmental benefits. In Nanchang’s planning context, the Ganjiang–Fuhe corridor is designated as a major ventilation pathway; our findings reaffirm its cooling potential and underscore the need to maintain its openness and continuity in planning and regulatory efforts.

Second, the ecological network can be integrated with the development of blue–green infrastructure to generate multiple ecosystem services. For example, the proposed riverside cooling corridor can be combined with flood control and drainage systems to form waterfront parks that also function as retention spaces. Newly added small green patches can incorporate rain gardens and habitat components, contributing simultaneously to stormwater management and biodiversity conservation.

This multifunctionality is evident in international cases, such as in the UK, where blue-green infrastructure mitigates UHI intensity in humid climates⁵⁰, and in Portugal, where nature-based solutions in Porto reduce natural hazard impacts, including heat²⁷.

Moreover, the green network can enhance residents’ quality of life and well-being. A connected greenway system facilitates walking and cycling and provides continuous recreational spaces. Studies have shown that green corridors encourage outdoor activity and improve mental health⁵¹. Therefore, the development of cooling networks should be integrated with green mobility systems and public space enhancements to support the idea of multifunctional green infrastructure.

It is also important to consider socio-economic and aesthetic factors in implementing ecological networks. For instance, introducing “stepping-stone” greens may involve land acquisition or reallocation, requiring a careful cost–benefit balance. In practice, underutilised urban spaces or brownfields can be prioritised for micro-green space retrofitting, which is cost-effective and helps to bridge network gaps. In cities such as Beijing, small-scale interventions like pocket parks have proven effective in building high-efficiency nodes within dense urban areas. In historical districts, shaded boulevards along cultural corridors can be created to preserve urban heritage while simultaneously providing cooling benefits. Ecological network planning should be closely aligned with urban spatial structures and cultural contexts to secure public and stakeholder support, thereby enhancing long-term feasibility and sustainability. To integrate driving factors, strategies should prioritise natural variables (e.g., high-elevation greens for ventilation), built features (e.g., reducing density in corridors to minimise heat retention) and land-surface characteristics (e.g., enhancing NDVI through layering in low-connectivity zones).

Validation of framework for transferability

Rationale for selecting Singapore as a validation context

To evaluate the transferability of the proposed multisource cooling network framework beyond the Nanchang case, Singapore was selected as an external validation context. Singapore differs substantially from Nanchang in both climatic regime and urban morphology. While Nanchang is characterised by a subtropical monsoon climate with extensive inland lakes and horizontal river–lake corridors, Singapore represents a tropical rainforest climate with persistently high temperatures, extremely high urban density and limited land availability. Moreover, Singapore has developed one of the most advanced blue–green infrastructure systems globally, integrating urban parks, vertical greenery and an island-wide park connector network.

These pronounced differences provide a stringent test of whether the core mechanisms underpinning the framework, namely vegetation-driven cooling, spatial connectivity of cooling sources, and interactions between cooling elements and the built environment, remain valid under contrasting environmental and urban conditions. Importantly, this validation focuses on mechanistic consistency and methodological plausibility, rather than re-implementing the full analytical workflow in another city.

Vegetation- and blue–green-based cooling effects

A central premise of the proposed framework is that vegetation and blue–green elements act as primary cooling sources through evapotranspiration and shading, as reflected by the dominant role of NDVI and land-cover composition in the Nanchang analysis. Empirical evidence from Singapore strongly supports this mechanism. Richards et al. (2020) demonstrated that different vegetation types in Singapore exhibit markedly different cooling performances, with multi-layered and tree-dominated vegetation providing significantly stronger air temperature reductions than grass-dominated areas¹⁸. This finding highlights the importance of vegetation structure, rather than vegetation presence alone, in regulating urban thermal environments.

Complementary evidence from remote-sensing studies further confirms the cooling effectiveness of urban greenery in tropical cities. Jia et al. (2024) employed a multiscale approach to evaluate the cooling effects of urban greening in Hong Kong (subtropical) and Singapore (tropical), integrating remote sensing and in-situ observations to assess both LST and near-surface thermal conditions¹⁹. Their results indicate that urban green spaces are associated with measurable LST reductions, with cooling magnitudes ranging approximately from 0.02 to 2.69 °C depending on spatial scale and urban context. The study further highlights the importance of vegetation volume and building height, showing that larger and more spatially aggregated green patches generate stronger cooling effects. These findings are fully consistent with the cooling-source identification logic applied in the Nanchang case, validating the framework’s emphasis on vegetation-based cold sources across different climatic settings.

Network connectivity and amplification of cooling effects

Beyond the presence of cooling sources, the proposed framework emphasises the role of spatial connectivity in amplifying and propagating cooling effects across urban areas, a concept operationalised in Nanchang through MSPA and circuit theory. Evidence from Singapore provides strong support for this network-based perspective. High-resolution observational research by Ching et al. (2025) documented the spatial extent and intensity of park cool island effects in a tropical urban park, demonstrating that cooling signals extend beyond park boundaries and vary systematically with surrounding urban morphology¹⁵.

Although explicit circuit-theory modelling has not been applied in these Singapore-based studies, the observed diffusion of cooling effects aligns closely with the framework’s assumption that connected blue–green networks outperform isolated green patches. At the city scale, Singapore’s park connector network links major parks, waterways and neighbourhood green spaces into a continuous system, further reinforcing the notion that spatial connectivity plays a critical role in enhancing urban cooling performance. This correspondence supports the methodological validity of representing cooling processes as network-based flows constrained by urban resistance surfaces.

Interactions with the built environment and implications for transferability

The framework also explicitly accounts for interactions between cooling sources and built-environment characteristics, recognising that vegetation and water bodies do not operate independently of urban form. Evidence from Singapore corroborates this interaction-based perspective. Recent mobile measurements conducted by Chàfer et al. (2022) in Singapore’s central urban area demonstrated that the synergistic effect of shading and greenery can reduce daytime outdoor air temperatures by up to 4.5 °C⁵². To achieve such benefits, Yang et al. (2015) identified that maximising the height-to-width (H/W) ratio of street canyons, with a threshold of approximately 3, is a critical morphological strategy for enhancing outdoor thermal comfort in tropical climates⁵³. This mechanism complements the Nanchang findings, where surface composition and horizontal green–blue corridors play a more prominent role in regulating LST. In Singapore, limited land availability shifts part of the cooling burden towards vertical morphology and three-dimensional urban form. Nevertheless, the underlying interaction logic remains consistent: urban thermal outcomes emerge from the coupling of vegetation, built form and human activity intensity, rather than from isolated factors.

To facilitate a structured assessment of framework transferability, Table 9 synthesises key climatic, urban and cooling-related characteristics of Nanchang and the Singapore validation context. The comparison emphasises the consistency of underlying cooling mechanisms while clarifying context-specific differences in urban form and blue–green infrastructure configuration.

Table 9 Comparison of key urban and Climatic characteristics between Nanchang and Singapore and implications for framework transferability.

Climate- and Morphology-Specific cooling strategies

Building on the demonstrated mechanistic transferability of the proposed framework (Sect.  4.3), this section extends the discussion from framework validation to strategic generalisation across diverse urban contexts. While the core cooling mechanisms, vegetation-driven cooling, Spatial connectivity and built–natural interactions, remain consistent, their relative effectiveness is strongly conditioned by Climatic regime, urban morphology and City scale. South china’s humid-subtropical plains give Nanchang a very different thermal context from arid Northern cities, mountain-basin cities or ultra-dense megacities. Consequently, the balance between greening and ventilation, and the dominant drivers of surface urban heat islands, vary systematically from place to place. Situating the Nanchang findings within this broader comparative landscape enables the identification of climate- and morphology-specific cooling strategies that are both transferable in principle and context-sensitive in application.

• Humid vs. arid climates: A global study of more than 50,000 cities shows that in warm–humid regions, the cooling benefit of vegetation is partly offset by high atmospheric moisture, while the importance of ventilation potential rises accordingly⁴. For Nanchang, this means that the green–blue network must be designed primarily as air-exchange infrastructure to relieve humid heat stress, whereas drier northern cities can deliver larger perceptual cooling gains through additional greening alone.

• Mountain–basin cities vs. flat-plain cities: Beijing and Stuttgart both exploit nocturnal downslope winds that couple with the urban heat-island circulation to form “mountain–plain breeze” corridors. Backwards-trajectory and computational fluid dynamics (CFD) studies in Beijing now underpin statutory ventilation-corridor zoning^54,55(Xu et al., 2021; Zheng et al., 2022). Flat Nanchang lacks this topographic engine, yet our results show that a continuous river–greenway spine can channel cool air almost as effectively, offering a transferable model for other delta or alluvial-plain cities.

• Compact megacities vs. medium-sized cities: In Guangzhou, façade density, building height and floor-area ratio already outrank vegetation as the primary predictors of SUHI intensity (Shi et al., 2019). Rome’s Tuscolano–Don Bosco district confirms the pattern: when the plan-area density (λ_p) exceeds 0.45 and corridor width falls below 100 m, pedestrian-level wind speed drops by ≈ 60%, despite high tree cover²³. By contrast, Nanchang’s moderate scale and well-preserved rural hinterland allow vegetation and impervious-surface ratios to remain the dominant levers.

• Blue–green–wind synergy: Performance-based corridor design along Seoul’s Cheonggyecheon (PB-UVC) raised evening wind speed by 28% and lowered mean radiant temperature by ≈ 1 °C⁵⁶. In a coastal Mediterranean setting, CFD optimisation in İzmir shows that re-orienting seafront blocks and widening inland street canyons can expand the “acceptable wind comfort” zone by 34% without additional land take⁵⁷.

  • Implications for heat-island mitigation strategies.

• Cities with ample ecological resources (e.g., Nanchang, Stuttgart) should prioritise an integrated blue–green ventilation network, linking rivers, lakes and green corridors, to maximise advective cooling.

• Ultra-dense megacities (e.g., Guangzhou, Hong Kong) need a package of form-based controls (lower λ_vol, larger corridor cross-sections) plus high-albedo and rooftop-green technologies to compensate for green-space scarcity⁵⁸.

• Humid coastal or delta cities should weigh wind guidance and moisture removal more heavily than arid inland cities, where evaporative and shading benefits dominate.

The Nanchang’s case reinforces a broader consensus: cooling networks must be climate- and morphology-specific. Where terrain offers natural air pathways, they should be protected; where it does not, well-aligned, continuous green corridors and watercourses can create functional equivalents, provided that local density and material choices are also addressed.

Model uncertainty and research limitations

Although this study provides valuable insights, several limitations and uncertainties remain and should be addressed in future work.

First, the Temporal scope of the data is limited. The selected remote sensing scenes represent typical summer conditions, but do not account for seasonal changes or interannual climate variability that May affect the thermal environment. Given that Nanchang May exhibit different SUHI characteristics between winter and summer, future studies should incorporate multi-seasonal, year-round data to capture a fuller picture. Additionally, the 2008 May data May introduce bias in Temporal comparisons, potentially underestimating peak summer SUHI intensity by 1–2 °C based on seasonal patterns³⁷.

Second, the evaluation of cooling network effectiveness relies mainly on comparative analysis and empirical inference, lacking direct support from fine-scale atmospheric dynamics modelling. Ideally, CFD or weather research and forecasting (WRF) models could be employed to simulate microclimate conditions under different planning scenarios, allowing for a quantitative assessment of temperature reductions and wind environment improvements

Third, socio-economic drivers of urban thermal conditions, such as anthropogenic heat emissions, are only indirectly reflected via population density in this study. More accurate quantification of various heat sources (e.g., traffic, air conditioning) through urban energy balance models would enhance Understanding of SUHI mechanisms.

Fourth, optimisation of ecological networks should consider cross-scale integration, including both city-wide and micro-level applications. This study primarily focuses on the general urban planning scale, without delving into the cooling effects of localised interventions, such as street or courtyard greening. Future work should combine network strategies with urban design at finer Spatial resolutions to validate their effectiveness.

Fifth, while the literature-based validation in Singapore confirms mechanistic consistency, it is preliminary and does not involve full quantitative re-application of the framework. Future research should test the entire workflow in diverse climates to further assess transferability.

Sixth, the classification of urban and rural reference areas in this study relied primarily on land cover types (e.g. Impervious surface versus cropland, forest and water) rather than the local climate zone (LCZ) framework. Whilst land cover data provides a continuous and high-resolution basis for analysis, it May overlook the complex thermal heterogeneity introduced by three-dimensional urban morphology and surface properties. Recent studies indicate that adopting an LCZ framework allows for a more refined and physically meaningful comparison between urban and non-urban areas by categorising landscapes based on their thermal and structural characteristics⁵⁹. The absence of an LCZ-based classification in this study is a limitation, and future research should consider integrating this approach to improve the robustness of SUHI interpretation and methodological transparency^60,61

Addressing these limitations, including the refinement of climate zone classifications, would strengthen the robustness, transferability and practical applicability of the green space network framework in urban heat island mitigation⁵⁹.

Conclusion

This study takes the central urban area of Nanchang as a representative case to systematically evaluate the heat-mitigation effects of a blue–green ecological network, through the integration of remote sensing analysis, landscape pattern metrics and circuit theory modelling. Comparative insights were also drawn from empirical findings in other urban contexts. The main conclusions and implications are as follows:

• Urban heat island evolution and cold-source structure in Nanchang: Over the past two decades, Nanchang has undergone a distinct transition from a cold-island-dominated to a heat-island-dominated thermal pattern. The proportion of low-temperature zones dropped from 41.7% to 19.7%, while high-temperature areas increased from 8.7% to 19.9%, with small, scattered hotspots evolving into large, contiguous high-temperature patches. This shift aligns with the rapid expansion of impervious surfaces and the densification of built-up areas. Meanwhile, major blue–green patches, such as the Gan River, Meiling Forest and the four main urban lakes, have maintained a cooling effect of 2–4 °C relative to their surroundings, acting as core “cold sources” in the city. However, the increasing isolation of these patches due to urban expansion has weakened their overall cooling impact.

• Dominant role of vegetation–land-use interaction and the added value of network connectivity: Statistical modelling shows that the interaction between vegetation cover (NDVI) and land-use type explains over 60% of LST variation, well above the explanatory power of single factors, such as building height or night-time light intensity. When these scattered cold sources are connected through 56 primary and 60 secondary corridors, daytime temperatures in high-heat zones decrease by 1–3 °C on average, and night-time heat dissipation accelerates significantly. These results suggest that “linking nodes into networks” yields far greater cooling benefits than simply “adding more green patches” of equivalent size.

• Generalisable implications for humid-subtropical plain cities: Unlike mountain–basin cities that benefit from nocturnal downslope winds, humid plain cities often lack natural ventilation corridors, and high ambient humidity diminishes the perceived cooling from evapotranspiration. Therefore, blue–green networks in such cities should prioritise horizontal airflow and convective ventilation along corridors. The Nanchang case demonstrates that, even in flat terrain, effective cold-air flow can be achieved by designing continuous greenways along river and lake frameworks, while controlling thermal resistance along the path. This provides practical reference for cities, such as Bangkok, Ho Chi Minh City and Lagos, which share similar climatic and geomorphological conditions.

• Cross-city comparison and differentiated strategies: The drivers of urban heat islands and the appropriate mitigation strategies vary significantly across city types. These differences are reflected in: (1) the dominant driving mechanisms, whether urban morphology or surface land cover; (2) the extent of ecological assets and spatial constraints; and (3) the influence of terrain and built form on local wind environments. In general:

• i. For megacities dominated by compact, high-density built forms, strategies should focus on controlling volumetric density, increasing surface reflectivity (e.g., cool roofs) and applying vertical greening to improve thermal dissipation.

• ii. In medium-sized cities with intact ecological assets and connectivity potential, priority should be given to constructing integrated blue–green corridors that link existing cold sources, enhance ventilation and deliver ecosystem services, leveraging natural (e.g., elevation), built (e.g., density) and land-surface (e.g., NDVI) factors.

• iii. For cities with terrain-driven wind dynamics, such as mountain–basin cities, the protection of natural ventilation paths (e.g., valley exits, foothill buffer zones) should be institutionalised through spatial planning regulations.

• iv. These differentiated approaches should be guided by a combined framework of dominant drivers–ecological endowment–spatial structure, allowing cities to tailor heat-mitigation strategies to their specific context.

• Theoretical contribution, limitations and future directions: By coupling landscape pattern analysis with circuit theory, this study proposes a transferable framework for evaluating multi-source, multi-path cooling networks and introduces the concept of “thermal resistance bottlenecks” to guide corridor prioritisation. However, the current analysis is limited by its reliance on satellite-derived LST, which is restricted to clear-sky daytime conditions and may overestimate temperatures in vegetated areas due to emissivity variations. The 2008 May data introduces potential seasonal bias in temporal comparisons, potentially underestimating peak SUHI intensity by 1–2 °C. The preliminary Singapore validation, while confirming mechanistic consistency (e.g., multi-layered vegetation cooling up to 2.69 °C), lacks full quantitative re-application. Moreover, future land-use expansion and anthropogenic heat emissions were not modelled. Future research should deploy micro-meteorological sensors in pilot corridors, combining CFD and Weather Research and Forecasting - Urban Canopy Model simulations to assess physiological equivalent temperature improvements and evaluate multi-functional co-benefits, such as public health and recreation. These steps are crucial to supporting the refined implementation and adaptive management of blue–green infrastructure in urban regeneration. The Singapore validation further confirms the framework’s transferability, with consistent mechanisms in tropical high-density settings.

In conclusion, the Nanchang case, supported by the Singapore validation, demonstrates that integrated blue–green corridors spanning urban and peri-urban areas provide a cost-effective pathway to mitigate SUHI by 1–3 °C in humid-subtropical plain cities with abundant ecological resources. In contrast, land-constrained or high-density tropical cities require morphological optimisation, reflective materials and vertical greening to achieve comparable cooling, as validated in Singapore. This framework offers actionable insights for tailoring urban heat mitigation to diverse climatic and morphological contexts worldwide.