Impact of Changes in Blue and Green Spaces on the Spatiotemporal Evolution of the Urban Heat Island Effect in Ningbo and Its Implications for Sustainable Development

Abstract

Blue and green spaces (BGS) play a crucial role in mitigating the urban heat island (UHI) effect by not only lowering land surface temperature (LST) but also regulating the urban microclimate and enhancing ecosystem services. In this study, Ningbo City is selected as the research area. LST data for the years 2014, 2017, 2020, and 2023 were retrieved using Landsat 8 imagery processed via the Google Earth Engine platform, employing an atmospheric correction approach. Simultaneously, land use types were classified using the random forest algorithm. Based on these datasets, a Geographically and Temporally Weighted Regression model was employed to quantitatively assess the spatial and temporal impacts of BGS changes on the UHI effect. The results reveal that (1) from 2014 to 2023, BGS in Ningbo exhibited a consistent decline, while construction land expanded significantly, leading to a gradual increase in the annual average LST; (2) strong UHI zones were primarily concentrated in urbanized zones and closely aligned with regions of elevated LST; the minimum, maximum, and average LST values in blue and green spaces were significantly lower than those observed in cultivated land and construction land; (3) the variation in the influence coefficient of blue space on LST was greater than that of green space, suggesting stronger spatiotemporal heterogeneity in its regulatory effect on the urban thermal environment. Additionally, the green-to-blue space area ratio increased from 9.7:1 in 2014 to 12.8:1 in 2023, deviating progressively from the optimal ecological balance. To promote sustainable urban development, it is imperative for Ningbo to strengthen the conservation and restoration of BGS, optimize their spatial configuration through evidence-based planning, and ensure the long-term stability of ecological functions.

1. Introduction

The urban heat island (UHI) effect refers to the phenomenon whereby urban areas exhibit significantly higher land surface temperatures (LST) than surrounding rural regions, primarily due to anthropogenic activities and the processes of urbanization [1,2]. This effect arises predominantly from the extensive presence of impervious surfaces such as buildings and roads, which possess high thermal mass, coupled with a relative scarcity of vegetation and water bodies that otherwise provide natural thermal regulation. As a result, urban areas experience elevated daytime temperatures and slower nocturnal heat dissipation, leading to the formation of localized thermal hotspots [3,4]. The UHI effect not only compromises urban thermal comfort but also contributes to increased energy demand, aggravated air pollution, and heightened public health risks—particularly for vulnerable populations such as the elderly and low-income communities. Understanding the UHI effect enables a more nuanced comprehension of urban heat distribution patterns and temperature dynamics during urbanization, thereby providing an empirical foundation for the development of scientifically informed urban planning and environmental management strategies [5]. In the context of global climate change and the rising frequency of extreme weather events, mitigating the UHI effect has become a critical component of enhancing urban resilience and promoting sustainable development.

Substantial advances have been made in UHI research, encompassing its causes, spatial and temporal patterns, driving factors, and mitigation approaches [6]. Current studies have predominantly focused on the roles of building density, land cover transformation, vegetative scarcity, and meteorological conditions in intensifying UHI during urban expansion [7,8,9]. Remote sensing, when integrated with in situ observations, has emerged as a pivotal tool for monitoring UHI dynamics and spatial distribution [10]. Findings indicate that the intensity and configuration of UHI effects vary significantly across cities, influenced by local topography, climatic conditions, socioeconomic development levels, and urban design [11]. To address these challenges, researchers have proposed a range of mitigation strategies, including the expansion of urban green infrastructure, the incorporation of water bodies, rooftop and vertical greening, and the implementation of nature-based solutions [12,13,14,15]. Future research is expected to delve deeper into the interconnections between UHI and broader issues such as climate change, energy consumption, and sustainable urban development, thus facilitating the formulation of more integrated and adaptive policy frameworks.

BGS, comprising water bodies (blue spaces) and vegetated areas (green spaces), includes natural or semi-natural urban features such as rivers, lakes, wetlands, parks, greenways, and open spaces [16,17]. These landscapes deliver essential ecological services—enhancing air quality, promoting biodiversity, regulating hydrological cycles—and critically contribute to the moderation of urban thermal environments [18]. Water bodies mitigate ambient temperatures through processes like evapotranspiration and radiative cooling, while green spaces reduce surface heat via shading and plant transpiration [19,20]. Moreover, BGS facilitate urban heat dispersion and energy flow regulation, collectively alleviating thermal stress and enhancing urban livability [21]. Consequently, these ecological systems are vital for improving the urban environment, addressing climate-related challenges, and advancing the goals of sustainable urban development.

Recent years have seen growing scholarly interest in the role of blue–andgreen infrastructure in mitigating UHI effects. A variety of methodological approaches have been employed to assess how such spaces influence urban temperature reduction and microclimatic improvement. Water bodies, particularly large ones such as rivers and lakes, have been shown to exert pronounced cooling effects, especially through nocturnal radiative heat release [22,23]. Concurrently, green spaces significantly reduce surface temperatures by providing shade and intercepting solar radiation [24]. Urban areas with higher greening levels—particularly those incorporating parks, tree-lined streets, and green roofs—consistently exhibit lower LST [25,26]. Furthermore, the synergistic interplay between water and vegetation in blue and green systems has been observed to amplify cooling outcomes. In some urban contexts, integrated blue and green networks serve as effective thermal buffers, substantially diminishing both the intensity and spatial extent of UHI zones [27]. However, the cooling efficacy of blue and green spaces is modulated by several factors, including the size, shape, spatial configuration, vegetation type, and soil characteristics of these ecological elements [28,29,30]. Overall, strategically designed and spatially optimized blue and green systems can create robust cooling corridors within urban landscapes, offering an effective solution for improving urban thermal comfort and resilience.

As a major port city in Zhejiang Province, Ningbo has experienced rapid urbanization in recent years. Driven by industrialization and urban expansion, the city has witnessed significant increases in population density and building intensity, alongside substantial transformations in land use structure. Cultivated land and natural green spaces have been progressively converted into commercial, residential, and transportation infrastructure, leading to widespread replacement of permeable natural surfaces with impervious built environments such as buildings and roads [31]. These changes have markedly reduced the extent of urban green spaces and water bodies, thereby intensifying the UHI effect. The proliferation of impervious surfaces facilitates daytime heat accumulation and hampers nocturnal heat dissipation, resulting in the formation of persistent thermal hotspots [32]. Moreover, the diminished presence of vegetated and aquatic areas weakens the city’s capacity for natural thermal regulation, exacerbating the occurrence and severity of high-temperature events [33]. The UHI effect in Ningbo not only degrades air quality and the ecological environment but also diminishes residents’ quality of life—particularly during the summer months—by increasing energy demand and public health risks. As such, optimizing the spatial configuration of blue–green infrastructure and enhancing urban planning frameworks have become critical priorities for improving the city’s thermal environment.

In this study, Landsat satellite imagery from four temporal nodes—2014, 2017, 2020, and 2023—was employed to retrieve LST data for Ningbo. Land use classification was performed using the random forest algorithm, and blue–green spatial features were extracted accordingly. Subsequently, a Geographically and Temporally Weighted Regression (GTWR) model was utilized to quantitatively assess the impact of BGS changes on the spatiotemporal evolution of the UHI effect. The study aims to offer a scientific foundation for ecological improvement and urban planning initiatives in Ningbo.

2. Research Area and Data Sources

2.1. Research Area

Geographically, Ningbo is a sub-provincial city under the direct jurisdiction of Zhejiang Province, situated between 120°55′–122°16′ E and 28°51′–30°33′ N, in the southeastern coastal zone of China, forming part of the southern flank of the Yangtze River Delta [34]. Administratively, the city comprises six districts, two counties, and two county-level cities. The central urban area includes the Haishu, Jiangbei, Beilun, Zhenhai, Yinzhou, and Fenghua districts, while the surrounding jurisdictions encompass Xiangshan County, Ninghai County, Yuyao City, and Cixi City (Figure 1). The city’s topography is characterized by a diverse landscape—hilly terrain dominates the west while the east consists primarily of low-lying plains, with a general elevation gradient sloping from southwest to northeast [35]. Ningbo enjoys relatively abundant vegetation cover, especially in forested regions surrounding urban and mountainous zones. These forests provide vital ecological functions, including air purification, local climate regulation through evapotranspiration, and biodiversity enhancement, forming a network of green corridors and ecological landscapes that strengthen the city’s blue–green infrastructure.

In terms of hydrology, Ningbo is endowed with a well-developed water network, including prominent rivers such as the Yong River and Yin River, as well as numerous lakes and tributaries. These water bodies significantly contribute to the ecological integrity of the city’s BGS, particularly by enhancing its cooling potential and mitigating elevated summer LST. The synergistic relationship between forested green spaces and aquatic systems provides Ningbo with a robust ecological regulatory mechanism, which greatly enhances the city’s resilience to climate extremes and its capacity for environmental adaptation.

2.2. Data Sources

This study also leverages the Google Earth Engine (GEE) platform (https://earthengine.google.com/ accessed on 9 January 2025)—a cloud-based geospatial analysis system developed by Google Inc. [36,37]. GEE integrates vast datasets, including satellite imagery, meteorological variables, and geophysical layers, and offers powerful computational capabilities through Google’s global infrastructure. It enables users to perform high-resolution data visualization and advanced geospatial analyses, while also supporting the integration of custom tabular and raster datasets, making it an indispensable tool for large-scale environmental and urban studies.

Since its successful launch in February 2013, Landsat 8 has provided stable data acquisition and high-quality imagery, making it a reliable source for long-term environmental monitoring. To ensure temporal consistency and representativeness of the dataset, this study selected Landsat 8 imagery from four three-year intervals—2014, 2017, 2020, and 2023—to systematically evaluate the spatiotemporal changes of BGS in Ningbo and its impact on the UHI effect. Remote sensing data were obtained from the publicly available LANDSAT/LC08/C01/T1_SR collection on the GEE platform, derived from the Thermal Infrared Sensor (TIRS) onboard Landsat 8. These datasets offer a spatial resolution of 30 m and include spectral information across Bands 1–11, providing comprehensive coverage of Ningbo’s land surface conditions.

To enhance data reliability, all imagery underwent atmospheric correction using the Land Surface Reflectance Code (LaSRC) algorithm, which minimizes atmospheric, solar, and sensor-related interference, thereby improving the accuracy of LST retrievals [38]. Additionally, atmospheric conditions such as cloud cover and aerosol concentration were accounted for to mitigate their potential impact on temperature estimation. Rigorous quality control and correction procedures were applied to ensure the precision and comparability of the LST data across time. These high-resolution, atmospherically corrected datasets provide robust support for evaluating the relationship between BGS spatial changes and UHI intensity, offering valuable insights for urban greening strategies, blue–green infrastructure development, and climate-adaptive urban planning.

3. Methodology

The research methodology (Figure 2) comprises the following key steps: (1) retrieval of LST using the GEE platform; (2) classification of UHI intensity levels; (3) land use classification using Landsat 8 remote sensing imagery to extract BGS; (4) analysis of the spatiotemporal variation of BGS and UHI effects; and (5) investigation of the impact of changes in BGS on LST based on the GTWR model.

3.1. LST Retrieval

Various methods are currently employed for LST retrieval, including the atmospheric correction method, the single-window algorithm, and the split-window algorithm [39]. While the latter two rely on simplified assumptions about atmospheric conditions and thermal emissivity, they often fail to capture the complexity of radiative transfer processes across different wavelengths and atmospheric layers. In contrast, the atmospheric correction method integrates radiative transfer modeling with remotely sensed data, offering improved accuracy under diverse geographic and meteorological conditions [40]. Accordingly, this study adopted the atmospheric correction method for LST inversion, which calculates surface temperature by disentangling the total thermal radiation received by the satellite into three components: (1) upward atmospheric radiance, (2) ground-emitted radiance transmitted through the atmosphere, and (3) downward atmospheric radiance reflected from the surface. The fundamental principle is to subtract atmospheric contributions from the satellite-observed thermal signal to retrieve the true surface-emitted radiance, which is then converted into surface temperature estimates [41]. The retrieval process can be divided into the following three steps.

(1)

Calculation of surface emissivity

The Normalized Difference Vegetation Index (NDVI) is an index derived from remote sensing data that effectively reflects the surface vegetation coverage and growth conditions. It is widely used in monitoring plant health and growth. The calculation formula is as follows:

$$NDVI=(NIR-RED)/(NIR+RED)$$

(1)

In the formula, the NDVI values range from [−1, 1], where NIR represents the reflectance of the near-infrared band of the remote sensing image, and RED refers to the reflectance of the red band of the remote sensing image [42].

According to the mixed pixel decomposition model, land cover types are categorized into three types: water surface, impervious surface, and vegetation surface. ε represents the surface emissivity. The calculation formulas for the surface emissivity of water (ε_water), impervious surfaces (ε_building), and vegetation surfaces (ε_surface) are as follows:

$${\epsilon }_{\mathrm{water}}=0.995$$

(2)

$${\epsilon }_{\mathrm{building}}=0.9589+0.0086FVC-0.0671{FVC}^2$$

(3)

$${\epsilon }_{\mathrm{surface}}=0.9625+0.0614FVC-0.0461{FVC}^2$$

(4)

$$FVC={\displaystyle \frac{NDVI-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}}$$

(5)

In these equations, FVC represents the Fractional Vegetation Cover, NDVI_max refers to the NDVI value of areas with pure vegetation cover, and NDVI_min corresponds to the NDVI value of areas without vegetation cover. To mitigate the influence of noise, the values for NDVI_max and NDVI_min are, respectively, set at 0.05 and 0.70 [42].

(2)

Calculation of radiative brightness

The radiative brightness B(T_S) of a blackbody at a temperature T_S in the thermal infrared wavelength range is calculated using the radiative transfer equation. The corresponding formula is as follows:

$$B(T_S)={\displaystyle \frac{[L_{\lambda }-L_{\mu }-\tau (1-\epsilon \left)L_d\right]}{\tau \epsilon }}$$

(6)

In the formula, B(T_S) represents the radiative brightness of the blackbody at temperature T_S, τ denotes the atmospheric transmittance in the thermal infrared band, ε is the surface emissivity, L_λ refers to the thermal infrared radiation received by the satellite sensor, L_μ represents the upward atmospheric radiative brightness, and L_d indicates the downward atmospheric radiative brightness [43].

(3)

Retrieval of LST

After obtaining the radiative brightness of the blackbody at a temperature T_S in the thermal infrared wavelength band, the surface temperature T_S (°C) of the study area is retrieved using the inverse function of Planck’s law. The corresponding calculation formula is as follows:

$$T_S={\displaystyle \frac{K_2}{\mathrm{l}\mathrm{n}\left(K_1/B\left(T_S\right)+1\right)}}$$

(7)

In the formula, the value of K₁ is 774.89 W/(m²·sr·µm), and the value of K₂ is 1321.08 K [43].

3.2. Classification of UHI Intensity Levels

To quantitatively evaluate UHI intensity, the Heat Field Variation Index (HFVI) was employed [44]. This metric expresses the standardized deviation of the surface temperature of each pixel from the regional mean temperature, thereby capturing the relative intensity of thermal anomalies within the study area [45]. HFVI is particularly effective in characterizing the spatial heterogeneity of urban thermal environments and mitigating the influence of seasonal and weather-related variability, thus enhancing the robustness and comparability of LST-based UHI assessments across different time periods. The specific calculation formula is as follows:

$$HFVI={\displaystyle \frac{T_S-T_{mean}}{T_{mean}}}$$

(8)

In the formula, HFVI represents the heat field variation index, T_S denotes the LST (°C) of a specific pixel, and T_mean refers to the mean LST (°C) of the study area [46]. To analyze the variation in UHI intensity across different years, the mean LST of the study area in 2014 is used as the baseline for comparison. Based on the calculation results, the UHI intensity is classified into four distinct levels (

Table 1. Classification of UHI intensity levels.

Level Value	HFVI Value	UHI Intensity Level
1	HFVI > 0.50	Strong UHI
2	0.3 < HFVI ≤ 0.50	Moderate UHI
3	0.1 < HFVI ≤ 0.3	Mild UHI
4	HFVI ≤ 0.1	None UHI

).

3.3. Land Use Classification and Extraction of BGS

Based on the standards and regulations, including the “Land Use Status Classification” (GB/T 21010-2017) [47], and considering the specific context of Ningbo City, the land use types in the city are categorized into four main types: cultivated land, blue spaces, green spaces, and construction land (

Table 2. Classification and content of land use in Ningbo City.

Land Use Type	Detail
Cultivated land	Land designated for agricultural production, including fields for crop cultivation, orchards, and vegetable farming areas.
Blue spaces	Land associated with water bodies, including rivers, lakes, reservoirs, ponds, wetlands, and their surrounding buffer zones.
Green spaces	Land covered by vegetation, such as forests, grasslands and parks.
Construction land	Land designated for urban development, industrial production, commerce, and residential purposes.

).

For land cover classification, the RF algorithm was used to generate high-accuracy land use maps for Ningbo. The classification workflow comprised five key steps: image selection, training sample collection, model training, classification execution, and accuracy assessment. A validation set comprising 30% randomly sampled data points was used to construct a confusion matrix, which evaluated classification performance by comparing the spectral signatures of validation samples with classification outcomes. The overall classification accuracy and Kappa coefficients for each temporal phase exceeded 0.80, indicating that the classification results met the research standards and achieve the desired level of precision.

3.4. Impact of Spatiotemporal Variations in BGS on LST Based on the GTWR Model

Huang et al. introduced the GTWR model as an extension of the traditional Geographically Weighted Regression (GWR) by incorporating temporal heterogeneity into the spatial framework [48]. The core innovation of the GTWR model lies in its ability to dynamically adjust regression coefficients in response to variations across both space and time. Compared with GWR, which accounts only for spatial non-stationarity, and the global Ordinary Least Squares (OLS) regression, the GTWR model demonstrates superior theoretical robustness and empirical applicability. Specifically, by integrating the temporal dimension, GTWR enables the simultaneous capture of spatial and temporal variability, thereby offering a more nuanced and accurate depiction of complex spatiotemporal processes. Furthermore, when handling large-scale spatiotemporal datasets, GTWR effectively reduces model bias through optimized spatiotemporal weighting functions, significantly improving estimation accuracy. In contrast to the OLS model, which assumes spatial independence and often fails to reflect local variations, the GTWR model excels in detecting nonlinear interactions and local spatial autocorrelation, thereby overcoming key limitations of traditional regression approaches in geographic analyses [49]. In this study, the GTWR model was employed to quantitatively assess the impact of BGS changes on the spatiotemporal evolution of LST in Ningbo. The expression of the GTWR model is as follows:

$$Y_i={\beta }_0\left(u_i,v_i,t_i\right)+\sum _{k=1}^m{\beta }_k\left(u_i,v_i,t_i\right)X_{ik}+{\epsilon }_i$$

(9)

In the formula, Y_i represents the response variable for the i-th sample point, while (u_i, v_i, t_i) denotes the spatiotemporal coordinates of the i-th sample point. X_ik is the k-th explanatory variable corresponding to the i-th sample point, and ε_i refers to the model error term. β₀(u_i, v_i, t_i) is the regression constant for the i-th sample point, and β_k(u_i, v_i, t_i) represents the regression coefficient for the k-th explanatory variable at the i-th sample point [50]. The formula for calculating the regression coefficients during model fitting is as follows:

$$\widehat{\beta }\left(u_i,v_i,t_i\right)={\left[X^{T}W(u_i,v_i,t_i)X\right]}^{-1}{X}^{T}W(u_i,v_i,t_i)Y$$

(10)

In the formula, W(u_i, v_i, t_i) represents the weight matrix of sample point i, which reflects the influence of other sample points on this particular sample point [50]. The spatial distance between sample points is calculated using the Euclidean distance formula, expressed mathematically as

$$d_{ij}^S=\sqrt{{(u_i-u_j)}^2+{(v_i-v_j)}^2}$$

(11)

The time distance between sample points is calculated using the following formula:

$$d_{ij}^T=\sqrt{{(t_i-t_j)}^2}$$

(12)

Due to the differing units of measurement for time and space, their variation may impact the results. Therefore, the formula for calculating spatiotemporal distance is as follows:

$${\left(d_{ij}^{ST}\right)}^2=\lambda {\left(d_{ij}^S\right)}^2+\mu {\left(d_{ij}^T\right)}^2$$

(13)

The weight function is based on an adaptive Gaussian function, with the specific weight calculation formula given as follows:

$$W_{ij}=exp\left(-{\displaystyle \frac{{\left(d_{ij}^{ST}\right)}^2}{h^2}}\right)$$

(14)

In this context, h represents the parameter for spatiotemporal distance decay within the weight function, indicating the degree of local calibration smoothing [51]. The optimal bandwidth is determined using the corrected Akaike Information Criterion (AICc). The primary objective of this study is to analyze the impact of spatiotemporal variations in BGS on the LST in Ningbo City across five different years.

4. Results and Analysis

4.1. Spatiotemporal Variations of BGS

The spatial distribution of BGS in Ningbo was extracted through remote sensing image interpretation for the years 2014, 2017, 2020, and 2023 (Figure 3). Land use classification statistics were used to calculate the area of each land use type (

Table 3. Land use type areas (km²) for Ningbo City from 2014 to 2023.

Land Use Type	2014	2017	2020	2023
Cultivated land	2975.99	2924.65	2970.69	2907.91
Blue spaces	457.96	423.64	379.87	341.52
Green spaces	4457.91	4475.69	4381.49	4376.73
Construction land	1462.69	1530.85	1622.77	1728.56

), along with the total BGS area for each administrative unit within Ningbo (

Table 4. BGS areas (km²) for each city within Ningbo City from 2014 to 2023.

City	2014	2017	2020	2023
Haishu District	327.98	327.08	322.78	320.76
Jiangbei District	47.87	47.15	45.3	44.6
Beilun District	252.39	250.36	236.81	235.73
Zhenhai District	43.63	44.18	40.86	40.42
Yinzhou District	376.91	375.61	367.79	364.9
Fenghua District	903.54	909.58	890.24	879.64
Xiangshan County	694.91	692.48	673.96	672.18
Ninghai County	1214.52	1220.73	1193.99	1197.62
Yuyao City	802.15	791.98	775.67	762.36
Cixi City	251.97	240.18	213.96	200.04

).

In terms of land use composition, Ningbo is predominantly characterized by cultivated land and green space. From 2014 to 2017, the area of green space increased by 17.78 km², indicating a brief period of ecological improvement. However, between 2017 and 2023, rapid urban expansion led to significant growth in construction land, resulting in a sharp decline of green space by 98.96 km². Concurrently, the blue space area decreased by 116.44 km² over the nine-year period, primarily due to encroachment from land development and infrastructure projects associated with urbanization.

Spatially, the reduction in BGS was more pronounced in the central urban area, which lost 66.27 km², compared to 131.35 km² in surrounding districts between 2014 and 2022. Specifically, BGS in Haishu District, Jiangbei District, Beilun District, Yinzhou District, Xiangshan County, Yuyao City, and Cixi City exhibited a continuous decline, decreasing by 7.22 km², 3.27 km², 16.66 km², 12.01 km², 22.73 km², 39.79 km², and 51.93 km², respectively. Zhenhai District and Fenghua District exhibited an initial increase followed by a subsequent decline, with net decreases of 3.21 km² and 23.90 km², respectively. In Ninghai County, BGS fluctuations were more irregular but still showed a net decrease of 16.90 km². Overall, all administrative units within Ningbo experienced varying degrees of BGS loss, with Cixi City showing the most substantial decline and Zhenhai District the least.

4.2. Spatiotemporal Variations of LST

To derive LST, the atmospheric correction method was applied to Landsat imagery for the study period, yielding both the spatiotemporal distribution of LST (Figure 4) and the average LST values (

Table 5. Average LST (°C) for Ningbo City from 2014 to 2023.

City	2014	2017	2020	2023
Ningbo City	20.20	22.22	23.64	24.70
Haishu District	20.13	23.38	24.21	24.35
Jiangbei District	21.78	26.86	27.33	25.39
Beilun District	20.37	23.35	23.57	26.34
Zhenhai District	22.63	25.72	26.80	26.33
Yinzhou District	19.91	22.16	23.72	26.62
Fenghua District	19.53	20.83	22.50	25.46
Xiangshan County	19.18	21.32	22.39	24.74
Ninghai County	18.66	20.34	22.16	23.68
Yuyao City	21.37	22.72	24.89	24.34
Cixi City	23.25	24.88	26.38	25.22

).

Analysis of the spatiotemporal LST distribution reveals a consistent pattern in which northern Ningbo exhibits higher LST compared to the southern regions. High-temperature zones were concentrated in Jiangbei District, Zhenhai District, and Cixi City, while lower temperatures were primarily observed in Fenghua District, Xiangshan County, and Ninghai County. Over the nine-year period, Ningbo experienced a steady increase in average LST, rising from 20.20 °C in 2014 to 24.70 °C in 2023, a total increase of 4.50 °C. The rise in LST across all administrative units reflects the compounded influence of BGS degradation and climatic factors. Between 2014 and 2023, LST increased by at least 1.96 °C in every city, with Yinzhou District recording the most substantial increase of 6.71 °C.

4.3. Spatiotemporal Variations of UHI Intensity

Based on the classification results of the UHI intensity in Ningbo (Figure 5), a significant UHI effect is clearly observed. The strong UHI zone is primarily concentrated in urbanized areas, closely aligning with regions of high LST. With the progression of industrialization and urbanization, the UHI effect in Ningbo has gradually expanded from the old urban core to the entire central city, while the UHI effect in built-up areas of other regions has also become increasingly pronounced.

According to the changes in the area proportions of different UHI intensity levels (

Table 6. The area proportion of UHI intensity levels in Ningbo City.

Heat Island Intensity Level	None UHI	Mild UHI	Moderate UHI	Strong UHI	Total
Area proportion in 2014	72.6%	10.8%	10.3%	6.3%	100.0%
Area proportion in 2017	59.2%	14.4%	11.5%	14.9%	100.0%
Area proportion in 2020	43.8%	29.1%	14.3%	12.8%	100.0%
Area proportion in 2023	26.9%	46.1%	13.2%	13.8%	100.0%

), the proportion of the non UHI zone steadily declined from 72.6% in 2014 to 26.9% in 2023. Conversely, the proportion of the mild UHI zone increased significantly, rising from 18.8% to 46.1%. The combined proportion of the strong and moderate UHI zones remained relatively stable across most years, except for 2014 (16.6%), with values of 26.4%, 27.1%, and 27.0% in subsequent years. This trend closely aligns with the continuous expansion of construction land and the reduction of BGS.

4.4. Spatiotemporal Analysis of BGS and UHI Effect

To assess the relationship between LST and land use types, spatial overlays of LST data and land use classifications were conducted (

Table 7. LST (°C) of different land use types from 2014 to 2023.

Year	LST Value	Cultivated Land	BLUE SPACES	Green Spaces	Construction Land
2014	Minimum	−8.90	−12.27	−11.97	−7.50
	Maximum	39.33	35.96	36.10	42.24
	Average	23.41	17.68	17.70	26.10
2017	Minimum	−8.76	−11.38	−11.21	−8.23
	Maximum	39.56	37.20	38.50	41.52
	Average	26.70	16.57	19.20	28.61
2020	Minimum	−7.34	−9.68	−9.72	−6.37
	Maximum	40.08	39.11	37.05	42.94
	Average	26.53	18.54	21.62	29.32
2023	Minimum	−7.93	−9.17	−9.25	−6.59
	Maximum	40.69	38.06	37.76	41.16
	Average	26.39	21.03	23.20	29.42

). The results reveal significant interannual differences in LST across land use types from 2014 to 2023, with the relative ordering of LST values shifting over time. Consistently, construction land exhibited the highest maximum LST, followed by cultivated land. In terms of average LST, construction land also recorded the highest values. Conversely, blue space and green space maintained significantly lower minimum, maximum, and average LST values compared to construction and cultivated land, underscoring their superior thermal regulatory capacity.

The spatial distribution of land use types within various UHI zones further illustrates these thermal disparities. BGS are predominantly located in none UHI zones, while construction land is largely concentrated in moderate and strong UHI zones. This spatial pattern confirms that construction land contributes most significantly to UHI formation in Ningbo, whereas BGS play a critical role in reducing LST and mitigating urban thermal stress.

4.5. The Impact of Changes in BGS on LST

To quantitatively examine the impact of BGS on LST, the GTWR model was applied using the average LST and BGS area for each city. The model achieved a coefficient of determination (R²) of 0.932, a standard error (Sigma) of 0.623, and a bandwidth of 0.196, indicating a high level of model fit and strong explanatory power [52]. The resulting influence coefficients for blue space (

Table 8. Influence coefficient of blue spaces from 2014 to 2023.

City	2014	2017	2020	2023
Haishu District	0.034431	0.038589	0.040434	0.024818
Jiangbei District	0.025309	0.027769	0.029968	0.010004
Beilun District	0.017779	0.020632	0.022535	−0.000735 *
Zhenhai District	0.012930	0.016424	0.017532	−0.007640
Yinzhou District	0.010102	0.013869	0.014324	−0.011663
Fenghua District	0.008495	0.011987	0.012409	−0.013448
Xiangshan County	0.007688	0.010251	0.011491	−0.013366
Ninghai County	0.007601	0.008466	0.011417	−0.011683
Yuyao City	0.008281	0.006686	0.012063	−0.008758
Cixi City	0.009597	0.005120	0.013271	−0.005287 *

Note: * p-value too large (p ≥ 0.05).

) and green space (

Table 9. Influence coefficient of green spaces from 2014 to 2023.

City	2014	2017	2020	2023
Haishu District	−0.005066	−0.006948	−0.006703	−0.000291 *
Jiangbei District	−0.005042	−0.006772	−0.006258	0.000016 *
Beilun District	−0.005019	−0.006574	−0.005794	0.000197 *
Zhenhai District	−0.004915	−0.006287	−0.005278	0.000281 *
Yinzhou District	−0.004672	−0.005867	−0.004702	0.000274 *
Fenghua District	−0.004274	−0.005288	−0.004081	0.000168 *
Xiangshan County	−0.003746	−0.004558	−0.003461	−0.000046 *
Ninghai County	−0.003156	−0.003720	−0.002912	−0.000364 *
Yuyao City	−0.002590	−0.002840	−0.002503	−0.000771
Cixi City	−0.002102	−0.001984	−0.002279	−0.001221

Note: * p-value too large (p ≥ 0.05).

) across the study period were statistically analyzed to explore spatial and temporal variability. A p-value less than 0.05 indicates that the coefficient is statistically significant at the given spatiotemporal location. Conversely, a p-value equal to or greater than 0.05 suggests a lack of statistical significance, and the effects of the associated variable should be interpreted with caution.

A comparison of the blue and green space influence coefficients from 2014 to 2023 indicates that the coefficients associated with blue space exhibit greater variability than those of green space, suggesting that blue space demonstrates stronger spatiotemporal heterogeneity in its cooling effect on the urban thermal environment. This variability may be attributed to the evaporative cooling function of water bodies, which is highly sensitive to climatic factors such as air temperature, humidity, and wind speed. Furthermore, water body morphology, adjacent land cover characteristics, and human interventions may significantly influence its thermal performance across regions and timeframes. In contrast, the cooling effect of green space is primarily driven by more stable mechanisms, such as vegetation cover and evapotranspiration, resulting in relatively consistent influence coefficients.

Throughout the study period, blue space influence coefficients were predominantly positive, indicating that reductions in water body area are generally associated with increases in LST. However, some negative coefficients emerged, suggesting localized cooling effects following blue space reduction beyond a certain threshold. This may reflect conditions where the decline in evaporation, reduced water volume, or lower heat capacity diminishes the heat storage and release potential of water bodies, leading to temperature reductions under specific climatic and spatial configurations. In contrast, green space impact coefficients were generally negative, reaffirming its cooling function. This effect was particularly pronounced during periods of green space expansion in 2017 and partial conversion to cultivated land in 2020. Moreover, from 2014 to 2023, the green-to-blue space area ratio in Ningbo increased from 9.7:1 to 12.8:1, deviating progressively from the hypothesized optimal ratio. This imbalance in the spatial configuration of green and blue infrastructure is likely to impair the connectivity of urban cooling networks, thereby weakening the overall capacity of BGS to regulate urban thermal environments effectively.

5. Discussion

Urban land use change and variations in LST significantly influence the efficiency of socio-economic systems and the formulation of strategies for sustainable urban development. These changes offer critical insights into pathways for mitigating elevated urban temperatures. Persistent increases in LST can adversely affect residents’ quality of life, elevate societal burdens, and exacerbate psychological stress, particularly under conditions of resource scarcity or extreme climatic events [53]. Addressing these challenges requires urgent and effective mitigation of the UHI effect through the scientific and rational optimization of BGS configurations. Such efforts are essential for promoting urban resilience to high temperatures, enhancing environmental livability, and safeguarding public health.

Empirical results from this study demonstrate that BGS in Ningbo plays a crucial role in regulating and mitigating the deterioration of the urban thermal environment. To further leverage this ecological function, sustainable urban development strategies should prioritize the preservation of adequate water body areas and the optimization of hydrological system structures, thereby strengthening the spatial agglomeration and cooling efficacy of aquatic environments [54,55]. Additionally, the spatial form of blue spaces should be scientifically reconfigured to maximize their regulatory efficiency at the landscape scale [56].

However, the evolving priorities of urban development, such as the revitalization of older neighborhoods and rural areas, pose challenges to large-scale BGS expansion, particularly under ecologically oriented territorial governance frameworks. In this context, urban planning must shift from emphasizing quantitative expansion toward enhancing the ecological quality and allocation efficiency of BGS. Multi-scale spatial optimization strategies should be adopted to amplify the cooling functions of BGS, ensuring more widespread and enduring thermal regulation benefits.

The intensifying UHI phenomenon highlights the inadequacy of isolated micro-scale interventions or limited BGS coverage in curbing urban thermal degradation. Therefore, urban planning must evolve from fragmented micro-regulation to integrated macro-level spatial governance, systematically managing both the spatial configuration and ecological functionality of BGS networks [57]. Guided by ecological principles, sustainable urban development should embrace regional-scale planning to ensure the holistic and coordinated layout of BGS, promoting its orderly expansion across administrative boundaries [58,59].

To realize this vision, it is essential to enhance ecological connectivity between landscape elements, reduce spatial fragmentation, and strategically manage high-density development in urban cores. These efforts will support the construction of a more adaptive and resilient urban thermal regulation system. Moreover, by strengthening the connectivity of the BGS network, urban planners can more effectively mitigate the adverse impacts of UHI and prevent further thermal deterioration driven by fragmented heat source distributions [60,61]. As cities expand, the establishment of ecological buffer zones between urban cores and peripheral areas should be informed by scientific delineation and enforced through rigorous regulatory measures. In BGS-rich regions, priority should be given to ecological protection and resource governance policies, ensuring the efficient utilization and equitable distribution of natural resources and thereby enhancing the comprehensive regulatory capacity of urban ecosystems.

6. Conclusions

In this study, Landsat series satellite remote sensing images from 2014, 2017, 2020, and 2023 were utilized to retrieve LST in Ningbo City using the atmospheric correction method. The spatiotemporal changes of the UHI effect were then analyzed in relation to changes in BGS distribution. The key findings are as follows:

(1)

Between 2014 and 2023, Ningbo experienced a persistent reduction in blue–green space alongside an expansion of construction land. This land use transition corresponded with a marked and sustained increase in the city’s average LST.

(2)

Strong UHI zones were primarily concentrated in urbanized areas, showing a high degree of spatial overlap with regions of elevated surface temperatures. Overall, the average LST of BGS areas remained consistently lower than the regional mean, whereas construction land exhibited significantly higher LST values.

(3)

Based on the GTWR model, the influence coefficient of blue space displayed more pronounced variability than that of green space, reflecting a higher degree of spatiotemporal heterogeneity in its impact on the urban thermal environment. From 2014 to 2023, the green-to-blue space area ratio in Ningbo increased from 9.7:1 to 12.8:1, progressively diverging from an optimal balance. This growing imbalance weakened the spatial cohesion and regulatory effectiveness of BGS in moderating urban surface temperatures.

The findings of this research hold substantial practical relevance for policymakers and urban planners. By elucidating the dynamic regulatory roles of blue and green spaces under urbanization-induced thermal stress, this study contributes to a deeper understanding of how BGS can serve as a nature-based solution for climate adaptation. The results provide theoretical support for the formulation of more scientifically grounded and spatially precise BGS planning strategies, which are essential for enhancing urban climate resilience, improving environmental quality, and promoting residents’ thermal comfort and well-being.