Assessment of Heat Mitigation Services Provided by Blue and Green Spaces: An Application of the InVEST Urban Cooling Model with Scenario Analysis in Wuhan, China

Abstract

Natural infrastructure is essential in reducing thermal discomfort caused by the urban heat island (UHI) effect. Optimizing and planning green and blue spaces can help establish nature-based urban heat mitigation strategies that benefit sustainable urban development. Most current studies on urban heat mitigation have focused on the single heat reduction effect of green space or blue space, while there has been a lack of research on the combined cooling effects of blue and green spaces. Moreover, existing heat mitigation models and methods cannot directly guide the optimization of blue–green spatial patterns at the urban scale. This has led to an unclear relationship between heat mitigation effects and blue–green spatial patterns. Based on land use data, meteorological data, and biophysical information as inputs, this paper utilized the InVEST urban cooling model (UCM) and scenario analysis method to simulate urban heat mitigation patterns by setting up different blue–green space configuration scenarios. The relative contribution of blue–green space changes to the variation of heat mitigation benefits was quantitatively estimated using the difference comparison method, and the relationship between heat reduction effects and urban blue–green spatial patterns was elucidated using spatial analysis methods. The results show that the InVEST UCM captured some of the variability in the surface thermal response of Wuhan and can be applied to the modeling of urban heat mitigation patterns. Furthermore, they show that consideration of the cooling effect of water evaporation can improve the simulation accuracy to some extent. In Wuhan, there were regional differences in heat mitigation patterns and the heat mitigation effect was significantly higher in the suburbs than in the city. Additionally, urban parks, lakes, and mountains with surface or block distribution had noticeable cooling benefits. Finally, the scenario simulation results demonstrate that green space was more efficient at mitigating heat, while blue space was more critical for the geographical partitioning of the UHI. These findings can provide a reference for the planning and optimal management of urban blue and green spaces, as well as for the design of heat reduction policies.

1. Introduction

The problem of thermal discomfort caused by high temperatures results from the urban heat island (UHI) effect, which is the difference in temperature between rural and urban areas. This effect is produced by changes in the energy balance driven by the thermal properties of materials used in urban areas and natural infrastructure. Materials such as concrete and asphalt store more heat, while natural infrastructure such as vegetation or water bodies can provide cooling through evaporation or shade. The UHI effect has affected many cities around the world. The rising urban heat has severely affected human health and well-being, leading to increased mortality or morbidity during heat waves. The heatwave of summer 2022, particularly in Europe and Asia, led to thousands of heat-related deaths. The thermal properties of the natural surface have changed as urbanization proceeds, with increasing population concentration bringing a large amount of human-caused heat. Therefore, the UHI effect is continuously expanding and progressing [1]. In addition, as climate change intensifies, the costs for cities to adapt or cope with increasingly frequent and extreme heat waves are rising, along with the potential losses associated with lower comfort and productivity, higher energy demand, and heat-related illness and death. Nevertheless, the benefits of adapting may outweigh the costs, and the consequences of inaction may be even more severe [2]. It is necessary to make informed decisions about natural infrastructure in urban planning and design to accommodate the impacts of more frequent heat waves and higher temperatures in the future, as these decisions often have long-term effects that last for more than 50 years.

In recent years, many cities have experienced heat waves, making urban heat mitigation services a priority. The natural infrastructure reduces the effects of surface UHI by providing shade, altering the thermal properties of urban structures, and increasing evaporative cooling [3]. Urban blue and green spaces, collectively known as urban green areas and water bodies, are essential for maintaining the heat balance and ecosystem health in urban areas. Blue spaces refer to urban water bodies such as rivers, lakes, wetlands, and reservoirs, while green spaces relate to urban green areas such as forests, grasslands, shrublands, and croplands. The high heat capacity and evaporative properties of blue spaces give them distinct ‘thermostat’ and ‘oasis’ effects, which are essential in reducing the surrounding surface temperature [4]. Green spaces can cool the air by reflecting or absorbing solar radiation and lower local air temperatures through evaporative cooling, thereby improving the microclimate and influencing the temperature over broader regions. Therefore, the planning of enhanced blue–green spatial systems can help to provide sustainable, nature-based solutions to urban heat mitigation problems. Within the urban development boundary, the controlled extent and balanced distribution of open spaces, such as structural green areas and water bodies, should be promoted to achieve a high-quality and sustainable urban development.

Present methods for quantifying or modeling heat mitigation services provided by urban blue–green spaces mainly fall into three categories: field measurements, remote sensing inversion and analysis, and model simulations [5,6]. (1) Field measurements are often taken to study the cooling effect of blue–green spaces by obtaining temperature data at specific points [7], with mathematical statistics then utilized to analyze the temporal and spatial temperature variations around water bodies and green spaces and the factors influencing them [8,9]. However, field measurements are limited by the amount of work involved and can be affected by many interfering factors. (2) Remote sensing inversion is usually employed to determine the surface or atmospheric temperatures around water bodies and green areas [10]. A GIS platform is then applied to analyze quantitative relationships between the cooling intensity of urban blue–green spaces and their surrounding environmental characteristics [11,12,13,14]. In recent years, this method has been widely used in the field of mesoscale urban thermal environment research [15,16,17]. (3) Computational fluid dynamics (CFD) models [18,19], empirical models [20,21,22], and three-dimensional microclimate models (e.g., ENVI-met) are used to compare the cooling effects of blue and green spaces under multiple scenarios and to analyze the mechanisms of different influencing factors [23,24]. Currently, commonly available software platforms include ENVI-met [25], Fluent [26], WRF [27], Airpark [28], and InVEST [29], but have issues that must be addressed. For instance, ENVI-met is extensively adopted to simulate microclimate in green areas or water bodies. However, it is limited to micro-scale investigations as its full simulation grid scale is restricted to a 2 × 2 km area [30,31]. Mesoscale models (e.g., WRF) also have some limitations, such as insufficient parameterization of urban microclimate processes and simplified treatment of urban surface features [32]. Further, though the CFD model is based on three-dimensional processes and has high accuracy in simulating temporal and spatial temperature profiles, it is computationally expensive [33]. This study is proposed to be conducted with the support of the urban cooling model (UCM) module in InVEST (Integrated Valuation of Environmental Services and Tradeoffs, version 3.12.0), an open-source model suite designed to support decision makers in incorporating spatial information on the heat mitigation benefits of natural infrastructure into resilience research and practice [29]. The InVEST UCM has two main advantages in the evaluation of heat mitigation services. First, the model simulates the spatial distribution of UHI based on three key parameters: shade, evapotranspiration, and albedo [22,34], which explicitly represent the biophysical mechanisms of UHI formation. Second, the model provides a spatially explicit approach that can be implemented to assess impacts under different urbanization scenarios after calibration and validation for a given city. Therefore, the method helps to design urban-scale solutions and can be used as a decision support system to explore ecosystem tradeoffs. In initial applications of the InVEST UCM, Hamel et al. (2020) [35] demonstrated its ability to represent spatial patterns of nighttime air temperatures during a heat wave in the Île-de-France region in 2003. Bosch et al. (2021) [36] employed a calibrated model to derive initial estimates of air temperature in Lausanne, Switzerland. They discovered that the urban cooling model outperformed preliminary regressions based on satellite data. Zawadzka et al. (2021) [37] utilized the model to evaluate the heat dissipation capacity of urban green spaces and validated the results with daytime surface temperature. They found that the model’s simulated heat mitigation index (HMI) was strongly correlated with changes in land surface temperature (LST) in cities or towns that had a wide range of LST values.

Most recent studies have concentrated on measuring the cooling functions of urban blue–green spaces and analyzing the influencing factors [38,39], usually considering only the individual heat mitigation effect of green areas or water bodies. Less attention has been paid to the combined cooling effects of blue–green spaces [40] and their relationship with spatial patterns [41]. As urbanization progresses, the conflict between urban construction and eco-environmental resources becomes increasingly prominent. Many scholars have focused on maximizing the ecological benefits of heat mitigation services under the constraints of limited urban green space and water resources. Though some models can simulate temperature differences between various underlying surfaces to predict the cooling effects of green areas and water bodies [39], they generally cannot guide the optimization of blue–green spatial patterns at the urban scale [42]. Few studies have attempted to quantify the heat mitigation benefits of blue–green spaces when integrating future land use scenarios and urban climates [43]. Though Fu and Weng (2018) [44] have reported changes in UHI intensity due to future land use change, their study mainly focused on how urban growth affected temperature patterns rather than explicitly examining the impact of the coverage of blue–green space. Predicting future land use is challenging because complex socioeconomic interactions and human-induced factors play an essential role. Scenario simulation can be used as an alternative approach to decompose the unknown future space into manageable components, which can avoid the uncertainty of predicting the exact path of the future [45,46]. This can help to improve the efficiency of urban heat mitigation by guiding the optimization of blue–green spatial patterns.

Wuhan is a core city in central China, where the Yangtze and Han rivers converge. Wuhan has abundant blue and green space resources (comprising 13% and 64% of the total area, respectively). Recently, it has focused on building a national ecological garden city. This study aimed to construct an urban heat mitigation model for Wuhan using the InVEST urban cooling model. It simulated changes in urban heat mitigation patterns by reducing blue and green spaces in the existing urban structure based on scenario analysis methods. This work quantitatively assessed the heat mitigation effects of various blue–green space configurations based on the identification of urban heat mitigation patterns. Its results reveal the link between the heat mitigation effects of urban blue–green spaces and their spatial patterns. The results of this study will contribute to the optimal allocation of urban blue–green spaces and provide urban planners or policymakers with methods and tools to evaluate the cooling effect of blue–green spaces.

The objectives of this study are to calibrate and validate the depiction of heat mitigation pattern in Wuhan using the InVEST UCM and to investigate the heat mitigation efficiency and the contribution of blue–green spaces. The innovations of this study are, firstly, the way it takes into account the cooling effect of water evaporation in the model, not just the cooling of vegetation. Given that it has been acknowledged that blue space can help mitigate UHI effects [3,7,47], the incorporation of the evaporative effects of water bodies should improve the accuracy of heat mitigation modeling. A second innovation is its comparison of the heat mitigation effects of blue and green spaces and exploration of the optimal allocation ratio between them, which will aid in prioritizing blue and green spaces and guide urban planning. Finally, the scenario simulation of blue–green spatial configurations can provide a nature-based solution (NbS) approach, one that helps to further explore the optimization of heat mitigation patterns based on the current blue–green patterns within the urban growth boundary. This should provide policymakers and urban planners with new perspectives and approaches that can help improve urban comfort and promote sustainable urban development.

2. Materials and Methods

2.1. Study Area

Wuhan, which is the capital of Hubei province in central China, is located at latitude 29°58′–31°22′ N and longitude 113°41′–115°05′ E with a total area of 8569 km² (Figure 1). It is situated in the eastern part of the Jianghan plain, with more than 70% of the region being flat. The terrain is generally higher in the southeast and lower in the northwest regions. The Yangtze river and its largest tributary, the Han river, converge in the city to form the three urban districts of Wuhan (Wuchang, Hankou, and Hanyang). The city has a subtropical monsoon climate with hot summers and cold winters. Spring and autumn are shorter, while summer and winter are longer [48,49]. Wuhan has a resident population of 12.33 million according to the 2020 China Census (http://www.stats.gov.cn/sj/pcsj/, accessed on 19 April 2023) and is ranked the 12th most populous city in China. In recent years, Wuhan has experienced rapid urban development with an increase in the frequency of extremely hot weather and a corresponding increase in the pronounced heat island phenomenon [50].

Wuhan boasts vast blue spaces in the form of rivers and lakes, accounting for about 13% of its total area. The Yangtze and Han rivers and their 165 tributaries, including the Jinshui, Tongshun, Fu, and Sheshui rivers, comprise a vast water system unique to any other metropolis. In terms of the number and size of its lakes, Wuhan is one of the top three inland cities in the world. Wuhan has 166 lakes, 33 of which are located in the central urban area. The abundance of water resources in Wuhan has become an ecological foundation for the city’s sustainable development.

Wuhan’s urban green spaces are steadily expanding, accounting for approximately 64% of its total area, and possess a rich diversity of flora. The city’s flora belongs to a transitional zone from evergreen broad-leaved forests of the central subtropics to deciduous broad-leaved forests of the northern subtropics. Southern portions of the Yangtze and Han rivers are dominated by camphor (Cinnamomum camphora), along with other tree species such as heather (Sinocalamus afinis), fir (Cunninghamia lanceolata), and oil tea (Camellia oleifera). In the north of the city, the trees include pine (Pinus massoniana), sequoia (Metasequoia glyptostroboides), phacelia (Platanus orientalis), and poplar (Populus canadensis). According to Li (2005) [51], the vegetation in Wuhan’s central and suburban areas is similar, consisting of a mixture of broad-leaved evergreen and mixed deciduous forests. The distinction is that the broad-leaved evergreen forests predominate in the suburbs.

2.2. Estimation of Land Surface Temperature (LST)

This study used Landsat 8 surface reflectance data to calculate the daytime land surface temperature (LST). The data were collected by operational land imager (OLI) and thermal infrared sensor (TIRS) with high spatial resolutions (30 m and 100 m) suitable for intra-urban studies. Only the three nearest years of summer images from 2019 to 2021 (1 June–30 September) were selected to analyze daytime LST in Wuhan during the study. This period can represent temperature heterogeneity across the city and is also the season when high temperatures have had the most severe health impacts. Low-quality data with clouds and shadows were excluded, and a total of 57 images were finally selected.

LST is a function of brightness temperature and emissivity. Brightness temperature is the temperature value for the 10th band of the Landsat 8 surface reflectance product, and emissivity measures the ability of an object to emit infrared energy [52,53]. The LST can be calculated using the following formula:

$$LST=\frac{BT}{1+(0.0000109\cdot (BT/0.01438)\cdot ln(E))}$$

(1)

where BT is the brightness temperature in Kelvins, and E is the emissivity of the surface in dimensionless units [54]. The calculated LST was then converted to Celsius temperature.

2.3. Principles of the InVEST Urban Cooling Model (UCM)

The InVEST UCM was applied to estimate and analyze the spatial distribution of heat mitigation in Wuhan in 2020 and under different scenarios. The urban cooling model calculates a heat mitigation index (HMI) at the raster scale based on several factors, including shade, evapotranspiration, and albedo, as well as the distance to cooling islands such as parks. The HMI is determined based on the cooling capacity of the neighborhood or its surroundings. If the pixel is unaffected by large green spaces, the HMI equals the cooling capacity (CC) index. However, if the pixel is surrounded by parks or green areas larger than or equal to 2 hectares, these areas will provide additional cooling capacity. The HMI is calculated as a distance-weighted average of green fields and surrounding pixels in this case. The HMI is the main output of the InVEST UCM and ranges from 0 to 1, where 0 means no heat is mitigated in a cell and 1 means the UHI is fully mitigated. Specifically, the HMI is computed as follows:

$$HM{I}_i=\{\begin{array}{c}C{C}_i,ifC{C}_i\ge C{C}_{par{k}_i}orG{A}_i<2ha\\ C{C}_{par{k}_i},otherwise\end{array}\}$$

(2)

$$C{C}_{par{k}_i}={\displaystyle \sum }_{j\in dradiusfromi}{g}_j\cdot C{C}_j\cdot e^{\left(\frac{-d(i,j)}{d_{cool}}\right)}$$

(3)

$$G{A}_i=cel{l}_{area}\cdot {\displaystyle \sum }_{j\in dradiusfromi}{g}_j$$

(4)

where CC_parki represents the CC value provided by each park, which is a distance-weighted average of CC values attributed to green spaces; d(i, j) is the distance between pixel i and pixel j; d_cool is the distance over which the green space produces cooling effects; GA_i is the area of green space calculated based on the search distance (d_cool) around each pixel; cell_area is the area of a cell in hectares (ha); and g_j is equal to 1 if pixel j is green and 0 if it is not.

CC is the local cooling capacity of a single pixel and is calculated differently for daytime and nighttime. Daytime cooling is based on shade, evapotranspiration, and albedo, whereas nighttime cooling is related to the intensity of buildings. In this study, we calculated the CC index for each pixel using the daytime approach from the scheme proposed by Zardo et al. (2017) [22] and Kunapo et al. (2018) [55]. The CC index is calculated as follows:

$$C{C}_i=0.6\cdot shade+0.2\cdot albedo+0.2\cdot ETI$$

(5)

$$ETI=\frac{K_c\cdot ET0}{E{T}_{max}}$$

(6)

The weights assigned to the model’s recommendations for shade, albedo, and evapotranspiration (0.6, 0.2, 0.2, respectively) are based on empirical data. The evapotranspiration index (ETI) represents the normalized value of potential evapotranspiration, which is calculated for each pixel by multiplying the reference evapotranspiration (ET0) and the crop coefficient (K_c) and dividing by the maximum value (ET_max) of the ET0 grid for the area of interest.

2.4. Inputs of the InVEST UCM

Inputs for the InVEST UCM include land use/land cover (raster), reference evapotranspiration (raster), area of interest (vector), and biophysical table. In addition, several parameters need to be specified, including reference air temperature, magnitude of the UHI effect, air blending distance, and maximum cooling distance.

• Land use/land cover (LULC) maps shade, crop coefficient, albedo, and whether the cell is green or not, based on the land cover type of the cell. The model resamples all outputs using the resolution of this layer. The land use data used in this study are from the global land cover product for 2020 at 10 m resolution (WorldCover V2) developed by the European Space Agency (ESA) (https://esa-worldcover.org/, accessed on 19 April 2023) and derived from Sentinel-1 and Sentinel-2 data. The data have been independently validated with an overall global accuracy of approximately 74.4% ± 0.1%. The accuracy is higher for permanent water bodies, tree cover, and cropland and lower for shrubland and herbaceous wetland. According to the WorldCover V2 product, there are eight land cover types in Wuhan: tree cover, shrubland, grassland, cropland, built-up, bare/sparse vegetation, permanent water bodies, and herbaceous wetland. Thus, the blue space includes permanent water bodies and herbaceous wetlands, while the green space includes tree cover, shrubland, grassland, and cropland.
• Reference evapotranspiration (ET0) measures the amount of water evaporating from land into the air over a given period. It is generally expressed as the water depth in millimeters (mm) per unit of time. In this study, the annual average reference evapotranspiration was calculated using the Penman–Monteith equation based on daily meteorological data from the 15 meteorological stations in Wuhan (http://data.cma.cn/, accessed on 19 April 2023), and its spatial distribution was obtained using inverse distance weighted interpolation.
• The biophysical table contains data on shade, crop coefficient (K_c, used to calculate evapotranspiration), and albedo for each land cover. The values for shade, K_c, and albedo are between 0 and 1. In practice, the range of the K_c value can be extended to 0–1.5. The shade above 2 m is assigned a value of 1, and that below 2 m is given a value of 0. K_c and albedo can be determined according to the parameters recommended by Zawadzka et al. (2021) [37] and the InVEST model. The value of green space is assigned as 0 or 1, with 0 denoting that the LULC class is not counted as green area and 1 indicating that the class qualifies as green space (area > 2 ha) with additional cooling. In this study, both green and blue spaces were given a value of 1, primarily to consider the cooling effect of evaporation from blue spaces. Thus, permanent water bodies and herbaceous wetlands were treated as green spaces when considering their evaporative cooling in the HMI calculations. Table 1 displays the specific values of the relevant parameters.

Table 1. Key parameters of the biophysical table assigned to each land cover type submitted to the InVEST UCM.

Land Cover	Shade (0–1)	Kc (0–1.5)	Albedo (0–1)	Greenspace (0 or 1)
Tree cover	1	1.003927	0.140110	1
Shrubland	0	0.967735	0.188667	1
Grassland	0	0.931543	0.192888	1
Cropland	0	0.717214	0.160713	1
Built-up	0	0.327539	0.208190	0
Bare/sparse vegetation	0	0.613452	0.231966	0
Permanent water bodies	0	1.000000	0.056354	0 or 1 *
Herbaceous wetland	0	1.100000	0.141964	0 or 1 *

* The HMI was simulated in two modes in this study: HMI-exclude, where blue spaces (permanent water bodies and herbaceous wetland) were given 0; HMI-include, where blue spaces were assigned 1, taking into account the cooling effect of water evaporation.

Table 1. Key parameters of the biophysical table assigned to each land cover type submitted to the InVEST UCM.

Land Cover	Shade (0–1)	Kc (0–1.5)	Albedo (0–1)	Greenspace (0 or 1)
Tree cover	1	1.003927	0.140110	1
Shrubland	0	0.967735	0.188667	1
Grassland	0	0.931543	0.192888	1
Cropland	0	0.717214	0.160713	1
Built-up	0	0.327539	0.208190	0
Bare/sparse vegetation	0	0.613452	0.231966	0
Permanent water bodies	0	1.000000	0.056354	0 or 1 *
Herbaceous wetland	0	1.100000	0.141964	0 or 1 *

* The HMI was simulated in two modes in this study: HMI-exclude, where blue spaces (permanent water bodies and herbaceous wetland) were given 0; HMI-include, where blue spaces were assigned 1, taking into account the cooling effect of water evaporation.

• Reference air temperature (T_ref) is the baseline air temperature in rural areas for the period of interest. This study was based on data from the Caidian meteorological station (http://www.nmc.cn/publish/forecast/AHB/caidian.html, accessed on 19 April 2023), which is a suburb of Wuhan and has an average air temperature of 17.5 °C from 1981–2010.
• Magnitude of the UHI effect (UHI_max) is the difference between maximum urban temperature and rural reference temperature. In the absence of local research, this value can be obtained from the global surface UHI explorer by Yale University (https://yceo.users.earthengine.app/view/uhimap, accessed on 19 April 2023), and it was accordingly determined to be 2.25 °C in this study.
• Air blending distance (d_air) is the search radius (in meters) applied to consider the air mixing in urban areas. This parameter has been added to the model to solve the wind dynamics in small areas. The model recommends an initial range of values from 500 m to 600 m [56,57], and we chose 550 m for this study.
• Maximum cooling distance (d_cool) is the distance (in meters) at which large urban parks or green areas (>2 ha) will produce cooling effects. These effects decrease with distance from green spaces. The model recommends using 450 m as an estimate due to a lack of local studies.

Table 2. Additional parameters submitted to the InVEST UCM.

Parameter	Description	Value	Reference
Tref	Reference air temperature	17.5 °C	1981–2010 average temperature of Caidian, a suburb of Wuhan
UHImax	Magnitude of the UHI effect	2.25 °C	Global surface UHI explorer
dair	Air blending distance	550 m	Model recommended value range for the initial run: 500 m to 600 m
dcool	Maximum cooling distance	450 m	Model recommended value

shows the values for the above parameters submitted to the InVEST UCM. The weighting factor method was chosen to calculate daytime cooling capacity with weights of 0.6, 0.2, and 0.2 for shade, albedo, and evapotranspiration, respectively, as suggested by the model.

2.5. Scenario Simulation

Scenario simulation is a spatially explicit representation of land cover and can be employed to obtain the main input layer to the InVEST UCM. This study proposed various land use scenarios to investigate the impact of changes in blue and green spaces on the spatial pattern of urban HMI to illustrate the heat mitigation effects of blue and green spaces in Wuhan. This can be achieved in practice through the InVEST proximity-based scenario generator.

The proximity-based scenario generator creates conversion patterns based on user-specified focal habitats and converted habitats. The user decides which habitats can be converted, what they can be converted to, and the type of pattern based on proximity to the edge of the focal habitat. In this way, various land use change scenarios can be generated, including urban expansion from the current blue and green spaces (where blue and green spaces are the focal habitats and urban built-up areas are the converted habitats). The resulting land use maps can be input into the InVEST UCM. Therefore, the proximity-based scenario generator is not intended to predict actual expansion patterns but to develop different land-use change scenarios to examine how the relationship between land use change and ecosystem services vary under different assumptions about land use change.

The land use conversion scheme for this study was carried out through ‘convert nearest to edge’. Specifically, the land use type closest to blue or green space was replaced with built-up areas. Eight land allocation scenarios were simulated based on the land cover status of Wuhan in 2020. These scenarios included the conversion of blue space into built-up areas, the conversion of green space into built-up areas, and the combined conversion of both blue and green space into built-up areas, with a maximum conversion area of 500 km² to 1000 km² and following the principle of converting the areas closest to the edge. These eight land-use conversion scenarios are displayed in Figure 2.

2.6. Quantification of Contributions

The difference comparison method was used to further clarify the response of heat mitigation services to changes in blue and green spaces in Wuhan. The contribution of changes in blue and green spaces to changes in heat mitigation services can be quantified using the following equations.

$$C_b=\Delta b/\left(\Delta b+\Delta g\right)\cdot 100\%$$

(7)

$$C_g=\Delta g/\left(\Delta b+\Delta g\right)\cdot 100\%$$

(8)

where C_b is the contribution of changes in blue space to changes in heat mitigation services; C_g is the contribution of changes in green space to changes in heat mitigation services; and Δb and Δg represent changes in heat mitigation services under the blue space change scenario and the green space change scenario, respectively.

3. Results

3.1. Heat Mitigation Pattern in Wuhan

Ordinary least squares (OLS) regression analysis was conducted to compare the heat mitigation pattern of Wuhan in 2020 simulated by the InVEST UCM with the daytime LST for the summers of 2019–2021. The HMI was obtained by two solutions, including or excluding the cooling capacity of blue space, and the LST was acquired from the thermal infrared band of Landsat 8. The spatial resolution of HMI and LST was inconsistent and needed to be unified at 30 m. The results show that the InVEST UCM captured some changes in Wuhan’s surface thermal response and was applicable to urban heat mitigation simulations. The strength of the correlation between HMI and LST increased when the cooling effect of blue space was considered in the simulations (R²-include = 0.684; R²-exclude = 0.445) (Figure 3). This suggests that considering the heat mitigation effects of blue spaces could help improve the relationship between HMI and LST to a certain extent, especially in Wuhan, which had extensive water bodies. In addition, according to the spatial distribution of HMI (Figure 4), when the cooling effect of water bodies is not considered, Liangzi lake, Axe lake, Lu lake, and Zhangdu lake in the suburbs, as well as the Yangtze river and Han river passing through the entire city, all exhibit modest heat mitigation effects (Figure 4b), which was not quite in line with the reality. Therefore, the cooling capacity of the blue spaces should be considered when evaluating the heat mitigation services in Wuhan. This can increase the strength of the relationship between HMI and LST and reflect the actual situation.

Next, the spatial distribution characteristics of heat mitigation patterns of Wuhan in 2020 were analyzed based on the results of HMI-include (Figure 4a).

(1) There were spatial differences in heat mitigation patterns in Wuhan. Heat mitigation effects were significantly higher in the surrounding suburbs than in urban areas (Figure 4a) due to differences in underlying surface features (e.g., surface albedo and heat capacity) between urban and suburban areas. As the distance from the study area edge increased, the population and building density became lower, which led to stronger heat mitigation effects.

(2) Specifically, areas with urban parks, lakes, and mountains distributed in a faceted or blocky pattern exhibited more pronounced heat mitigating effects compared with built-up regions of the city (Figure 4a). Concentrated and contiguously distributed lakes and green areas were the main cold islands with an intense heat mitigation effect. In particular, Wuhan’s urban lake parks, such as the East Lake Scenic Area, Yanxi Lake Wetland Park, and some mountainous areas, were found to show strong heat-mitigation effects. Both urban lakes, such as East lake, South lake, Tangxun lake, and Houguan lake, and suburban lakes, such as Niushan lake, Liangzi lake, Lu lake, Axe lake, and Zhangdu lake, exhibited strong heat-mitigating effects due to the presence of water bodies and vegetation distributed in a planar or block pattern. Many central urban areas of Wuhan are impermeable, especially areas along the Yangtze and Han rivers, which are highly urbanized and distributed contiguously. These areas showed more prominent UHI and weak heat mitigation effects. Similarly, the built-up areas of district administrative units in the suburbs had weaker heat mitigation effects. Furthermore, the heat mitigation effect proved weak in the dense industrial and economic development zones.

(3) Blue space plays a crucial role in the spatial distribution of heat islands, especially in Wuhan where there is a high proportion of large water bodies in the central city. These can influence the distribution of UHI by breaking up heat islands with a fragmented pattern. This effect can help alleviate the severe UHI conditions in urban districts. In Figure 5, the spatial partitioning of UHI can be observed from rivers and lakes in the central urban area of Wuhan.

3.2. Scenario Simulation of Land Use

Three types of land use conversion scenarios were designed for this study, which includes:

(1)

Conversion scenario of blue space to built-up areas

In this scenario, built-up areas were simulated by converting blue space at a rate of 500 km² and 1000 km² (Figure 6b). The results show that the proportion of blue space in Wuhan progressively decreased from 13.16% (using 2020 as a reference, Figure 6a) to 7.34% and 1.52%, while the proportion of built-up areas gradually increased from 10.99% (in 2020) to 16.82% and 22.64% (Figure 7).

(2)

Conversion scenario of green space to built-up areas

In this scenario, built-up areas were modeled with a conversion ratio of 500 km² and 1000 km² from green space (Figure 6c). The results indicate that the percentage of green space in Wuhan continued to decline from 63.64% (in 2020) to 57.82% and 52.00%, while the share of built-up areas grew gradually from 10.99% (in 2020) to 16.82% and 22.64% (Figure 7).

(3)

Combined conversion scenario of blue and green spaces to built-up areas

In this scenario, built-up areas were simulated by transforming blue and green spaces for the following three patterns: b500g500 (both blue and green space converted to 500 km²), b500g1000/b1000g500 (blue and green space, one converted to 500 km² and the other converted to 1000 km²) and b1000g1000 (both blue and green space converted to 1000 km²) (Figure 6d). The results suggest that the proportion of blue and green spaces declined steadily from 76.80% (in 2020) to 65.16%, 59.34% and 53.52%, while the proportion of built-up areas rose incrementally from 10.99% (in 2020) to 22.64%, 28.46% and 34.28% (Figure 7).

3.3. Scenario Simulation of Heat Mitigation

The above eight land use scenarios were input into the InVEST UCM for HMI simulations. Figure 8 and Figure 9a exhibit spatial distributions and numerical statistics of these scenarios. By spatially overlaying the simulation results with the HMI map in 2020, we obtained the statistics of HMI changes in different intervals under different scenarios (Figure 9b).

According to the statistics in Figure 9a, the low-value zones (0.0–0.2) grew relatively small (less than 1 km²) in all eight scenarios, and high-value zones (0.8–1.0) exhibited varying degrees of decline in different scenarios. Therefore, we would focus on the reduction in high-value areas. Generally, the b500 scenario had the smallest decrease in high-value areas, while the b1000g1000 scenario showed the largest drop. Furthermore, the decline in high-value areas was smaller, and the mean value was higher in the b1000g500 scenario compared with the b500g1000 scenario, indicating a more effective heat mitigation effect of green space.

• As blue space was converted to built-up areas at different magnitudes, urban heat mitigation patterns changed accordingly, with high-value regions (0.8–1.0) experiencing the most significant changes (Figure 8b). Specifically, when blue space was declined by 500 km² and 1000 km² (scenarios of b500 and b1000, respectively), high-value regions (0.8–1.0) were decreased by 1289.89 km² and 1751.30 km², respectively (with respective 15.02% and 20.39% reductions in area ratio) (Figure 9b).
• As green space was converted to built-up areas at different rates, urban heat mitigation patterns changed dramatically, with a marked decline in high-value regions (0.8–1.0) (Figure 8c). In particular, when green space shrunk by 500 km² and 1000 km² (scenarios of g500 and g1000, respectively), high-value regions (0.8–1.0) were decreased by 1059.15 km² and 2591.08 km², respectively (with respective 12.33% and 30.16% reductions in area ratio) (Figure 9b).
• Comparison between scenarios of b500g1000 and b1000g500 (Figure 8d): The reduction in high-value areas was somewhat smaller in the b1000g500 scenario (2542.79 km² for b1000g500 vs. 3365.25 km² for b500g1000), while the increase in low-value areas was a little more (0.07 for b1000g500 vs. 0.03 for b500g1000) (Figure 9b). Additionally, the mean HMI increased slightly in the b1000g500 scenario (0.6130 for b1000g500 vs. 0.6109 for b500g1000). This demonstrates that green spaces were more efficient at mitigating heat.

According to the spatial distribution, in scenarios where blue space was converted to built-up areas, the decreased high-value regions were mainly located around lakes and rivers in the study area and were relatively dispersed. In scenarios where green space was converted to built-up areas, the reduced high-value regions were primarily distributed in the suburbs, with the most located in Huangpi and Xinzhou districts. In the combined conversion scenarios of blue and green spaces to built-up areas, except for the common areas of change (high-value area reduction consistent with b500g500), decreased high-value areas were mostly situated in the suburbs of Huangpi and Xinzhou districts in the b1000g500 scenario, while reduced high-value areas were located largely around large water bodies such as East lake, Tangxun lake, Yangtze river, Zhangdu lake, Niushan lake, Liangzi lake, Lu Lake, and Axe lake in both urban and suburban areas in the b500g1000 scenario (Figure 10).

This study examined the contribution of blue and green spaces to heat mitigation in urban and rural areas. For the city as a whole, the overall contribution of green space was higher at 82.86%, while the contribution of blue space was only 17.14%. We found that green spaces were more efficient in providing heat mitigation services in Wuhan, as reflected in the high-value areas and mean value of HMI. It is worth noting that, in urban areas, the contribution of blue space was higher (59.50%) while the contribution of green space was lower (40.50%) (

Table 3. Contributions of blue and green spaces to heat mitigation services and the proportion of blue and green spaces in urban, rural, and the whole city, respectively.

Region	Blue Space		Green Space
	Contribution (%)	Area (%)	Contribution (%)	Area (%)
Urban	59.50%	20%	40.50%	32%
Rural	15.77%	12%	84.23%	68%
The whole city	17.14%	13%	82.86%	64%

). This is due to the high relative share of green space in urban areas. Specifically, the proportion of blue and green space in urban areas was 20% and 32%, respectively, while in rural areas, it was 12% and 68%, respectively. In addition, our study has shown that blue space can have a dispersing effect on UHI, as shown in Figure 5 and discussed in Section 3.1. Overall, optimizing the spatial distribution of blue and green spaces can improve the efficiency of heat mitigation.

4. Discussion

4.1. Heat Mitigation Effects of Blue and Green Spaces and Their Optimal Configuration

According to the results of heat mitigation simulations for the three types of scenarios in this study, it can be seen that spatial patterns can influence the UHI effect. Increasing urban built-up areas by reducing blue and green spaces in cities will lead to a more severe urban thermal environment and increased health risks. In other words, enhancing green space and preserving blue space in urban spatial design and planning can alleviate the UHI effect [8]. This study found that, when the same area of urban blue space and green space turned into built-up areas, high-value regions and mean HMI values of green space declined more markedly in green space than in blue space. This suggests that green spaces are more effective than blue spaces in mitigating the urban heat island effect. Zhang et al. (2015) [14], Yuan et al. (2017) [58], and Gao et al. (2019) [59] have reached similar conclusions. Therefore, efforts to expand urban green spaces, such as creating green roofs, improving green areas in residential areas, and developing urban parks, should be encouraged. In addition, the cooling effect of water bodies was found to be of practical importance in partitioning and controlling the distribution of UHI. Similar conclusions have been drawn in a study by Gao et al. (2019) [59].

The application of blue or green space during rapid urbanization is often patchy and relatively clustered, which greatly diminishes the cold island effect of these natural infrastructures. However, the expansion of blue or green space is usually achieved marginally or by means of enclaves. Off-site compensation cannot fully offset the warming caused by the loss of formerly blue or green space. As a result, there is a clear imbalance between the loss and expansion of blue or green space in cities [60], leading to an uneven distribution of surface temperature differences. Further research is therefore needed to find ways to optimize the regulation of urban land use, especially the strategic allocation of blue and green spaces. For example, in urban landscape planning, the urban landscape pattern can be optimized by gathering vegetation or water bodies and dispersing impervious surfaces (buildings, roads, car parks, etc.). Urban planners and policymakers should prioritize the protection of natural spaces and the restriction of the conversion of natural waters or vegetation to other land uses (e.g., farmland and built-up areas) to help reduce the UHI effect and create a healthier and more comfortable urban living environment.

4.2. Advantages and Limitations of This Study

This study has the advantage of integrating multi-source data and scenario design for heat mitigation. Using the InVEST UCM and scenario analysis methods, we examined the heat mitigation effects under different urban land use patterns in Wuhan. The InVEST UCM contains information on crucial surface attributes that have been proven to determine air and surface temperatures, such as evaporative cooling of vegetation, shade by tall trees, and albedo. It thoroughly uses commonly available spatial datasets (e.g., land use, evapotranspiration), requires fewer in situ measurements, and avoids complex analysis of remotely sensed thermal data [37,61]. The model also enables analysis of the interactions between other ecosystem services provided by blue and green spaces. Furthermore, our research provides evidence for nature-based heat mitigation solutions. The scale and intensity of UHI effects can be offset and reduced by scenario configurations based on blue and green spaces. These solutions involve the incorporation of vegetation and water bodies into urban design and planning, such as through the construction of green cities and sponge cities, and will help provide a better thermal environment for urban dwellers, lowering their health risks. These efforts will additionally co-benefit other objectives related to blue–green natural infrastructure, such as carbon sequestration and flood control [62,63].

This study also has some limitations. Firstly, uncertainties in the input parameters of the InVEST UCM may affect the simulation results [37,59,63,64]. For instance, the CC index relies on experiential weights derived from limited case studies. Such weighting can moderate the influence of critical factors that contribute to cooling effects but may also include high uncertainty. In addition, the model sets a maximum cooling distance (d_cool) generated by large urban parks and green areas (>2 ha). However, the minimum radius (r) for generating a circle of 2 ha area is approximately 80 m, which means that if the cooling distance computed by the model is less than this value, no space can be classified as large green space. In other words, if the cooling distance from urban parks or water bodies is <80 m (as argued by some authors, such as Broadbent et al., 2018 [65] and Motazedian et al., 2020 [66]), the model calculations will result in weaker HMI values. Two parameters, d_cool and r, control the effect of large green spaces and air mixing and their values are difficult to obtain from the literature. This is because they vary with vegetation characteristics, climate (effect of large green spaces), and wind direction (air mixing). This is the main limitation of the model with respect to the uncertainty in the effect of horizontal convection on temperature homogenization. Though the parameters can be calibrated when conditions allow, additional analyses are required to select the correct parameters based on local information, such as using a sliding average algorithm that considers a wider range of convection and air movement [35]. However, limitations in terms of how the urban cooling model represents the spatial air mixing and cooling effects of green spaces seem to be difficult to overcome when monitoring stations are sparse or the spatial resolution of the data is low, though in this case spatial regressions based on remote sensing features such as LST and NDWI do not necessarily perform better than the InVEST UCM [36]. Secondly, the InVEST UCM only considers the cooling benefits of green space. To avoid such limitations, we evaluated the cooling effect of evaporation from water bodies when using the model in this study. We observed a certain degree of improvement in the correlation between simulated HMI and LST when water bodies were incorporated into the cooling function. Though we have studied the cooling effects of blue space in this research, it is only superficially integrated. Further investigation should delve into the combined cooling benefits of green and blue spaces. Thirdly, this study was conducted on an annual time scale and did not consider differences in diurnal and seasonal variations in the cooling effects of blue and green space. Finally, the intensity of HMI may also be associated with the morphology, distance, and spatial relations between blue and green spaces [67], which we have only briefly discussed in this study.

4.3. Further Research

There are several ways to explore this topic further. One approach is to test the model’s sensitivity to relevant parameters or conduct experimental studies to identify and reduce uncertainty. Moreover, model-estimated temperature patterns could be compared with observed or simulated data to gain insight into the relative effects of shade, albedo, and evapotranspiration. For example, Bosch et al. (2021) [36] performed 100 calibrations of the InVEST UCM and found that the weights obtained were very close to those recommended by the model (0.2, 0.2, and 0.6). However, these weights should be modified in consideration of the specific climatic or weather conditions in a given area, due to the fact that cooling effects may vary. For example, it may be more effective in hot and dry climates, where the cooling effect of vegetation and shade will be particularly pronounced, while it may be less effective in areas of high humidity, where the transpiration of vegetation may be limited. Subsequent versions of the model could also be improved to better represent the specific morphology of many cities by separating the cooling distance setting used to calculate the number of green spaces from the search window size. This would address the instability of the model. In particular, when assessing the cooling effect of large green spaces on surrounding areas, the model is simplified based on a distance-weighted average of the CC values from the large green spaces and the pixel of interest, and involves the setting of a maximum cooling distance from the green space, both of which may have some uncertainty. Though the simplification of the model is a well-thought-out choice, it is essential that these issues are taken note of when using the tools, especially when more complex or accepted models are routinely used in a city. InVEST UCM can be used in combination with other models to ensure the high accuracy of the heat mitigation estimates derived from the model [35]. Future work could focus on comparing heat mitigation effects at different time scales to produce more comprehensive results. For instance, seasonal variations and diurnal time differences in the combined cooling effects of blue–green spaces could be analyzed [5,68], accounting for factors such as morphology, distance, and spatial relations of blue–green areas. In addition, this study was conducted in Wuhan, and the applicability of the findings to other cities should be further examined under different thermal environments.

5. Conclusions

InVEST UCM and scenario simulation methods were utilized to construct a heat mitigation model for the city of Wuhan and to assess the impact of changes in blue–green space on spatial patterns of urban heat mitigation. A quantitative assessment of the contribution of blue–green space to heat mitigation services in Wuhan was also conducted. The results were as follows.

(1) The InVEST UCM can capture some changes in the surface thermal response of Wuhan and is suitable for modeling urban heat mitigation. When the cooling of water evapotranspiration was considered in the simulation, there was a certain improvement in the correlation between HMI and LST, and this led to an increase in simulation accuracy.

(2) There were spatial differences in the heat mitigation patterns in Wuhan. Heat mitigation effects were generally higher in surrounding suburbs than in urban areas. Moreover, urban parks, lakes, and mountains with a surface or block distribution had more substantial heat mitigation effects.

(3) Scenario simulation results reveal that both blue and green spaces were effective in mitigating the urban thermal environment, with green spaces being more efficient in cooling and blue spaces being more critical to the spatial partitioning of the UHI. It is especially evident in the central urban districts of Wuhan, which have large water bodies. Furthermore, green spaces contributed 82.86% of the total heat mitigation in the whole city, while blue spaces contributed 17.14%. However, in the urban areas, the contribution of blue space was greater at 59.50% compared with 40.50% for green space, due to the higher proportion of water bodies in the urban areas of Wuhan than in the rural areas.

(4) Further research is required to fully understand the interaction between the cooling effects of blue and green spaces and the balance of the two in urban planning, taken separately or combined.

The InVEST UCM is suitable for urban-scale studies and generally recommends the use of 30 m or higher resolution land cover and meteorological data, as sufficient spatial resolution and data quality are required to capture temperature differences within the city. Because the model does not explicitly consider the effects of topography on wind speed and temperature, it may be more appropriate for cities with flat or gentle slopes and less appropriate for cities with complex topography, such as mountains or canyons. In addition, the model can be used for a variety of climates and weather patterns, but the cooling effect of urban green spaces will vary depending on the specific climate and weather conditions of a particular area. The findings of this study contribute to the rational planning of blue and green spaces and help to improve the efficiency of urban heat mitigation based on natural solutions for sustainable urban development. Moreover, the study can provide some references for the impacts of heat waves in cities with a high proportion of blue space.