How to plan urban green space in cold regions of China to achieve the best cooling efficiency

With the acceleration of urbanization, the urban heat island (UHI) effect has intensified. Urban green space can retard the UHI effect. However, most existing studies have only focused on hot regions, while little attention has been given to cold regions that also have summer heat protection requirements. Furthermore, existing researc has not classified urban green spaces according to the presence or absence of water, which can lead to inaccurate results. This paper takes four cities in cold regions of China as examples and studies the cooling effects of two different types of urban green space. The results indicate that in cold regions of China, green spaces containing water bodies have a stronger cooling effect than those without water. For green spaces without water, the cooling intensity is related to the background temperature and green space areas, while for green spaces containing water bodies, the area of the internal water body is the key influencing factor. Specifically, there is a threshold value of efficiency (TVoE) for the green space areas without water in cold region cities of China, which is approximately 0.52 ha, while there is no TVoE for the green space areas containing water bodies. Additionally, there is a TVoE for the water/land ratio of the green spaces containing water bodies of approximately 0.5. The methods and results of this study can provide a reference for future research and for urban planners and managers designing urban green spaces.

correlated with land surface temperature (LST) [26] . Liang et al. found that the area of green space is negatively correlated with the LST within a certain threshold: that is, within the certain threshold, the larger the green space area, the lower the LST [27] . Mikami and Sekita found that, if the green space area exceeds 20 ha, its cooling intensity will not increase with a further increase in area [28] . Moreover, Jaganmohan et al. concluded that increasing the spatial complexity of smaller green space has a negative effect on the cooling intensity but that increasing the spatial complexity of green space larger than 5.6 ha has a positive effect [29] .
Yu et al. [26] [30] proposed the concept of the threshold value of efficiency (TVoE) to obtain the optimal scale of urban green space and optimize urban green space design. Le et al. [31] found that the TVoE of green space in a tropical city (Hanoi) is 1 ha. Meanwhile, Yang et al. [32] found that the TVoE of green space in a high-latitude city (Copenhagen) is 0.69 ha. Fan et al. [33] investigated seven low-latitude Asian cities, and showed that the TVoE of the cities ranged from 0.6 to 0.95 ha. Yu et al. [30] found that the TVoE in cities with a Temperate Monsoon climate and a Mediterranean climate is generally around 0.5 ha. Studies have shown that TVoE is highly correlated with urban background climate conditions [31][32][33] . Therefore, it is necessary to conduct research based on specific climate regions with different background climatic conditions. The cold region is one of the five climatic regions in China. It refers to the area where the average temperature of the coldest month is (-10)-0 ℃ and the average number of days when the daily temperature is ≤5 ℃ is between 90 and 145 days. In summer, even in cities in cold regions, there are times when the temperature in the city is high. Therefore, it is necessary to pursue research on cold regions. However, existing research has mainly focused on cities in hot regions or individual cities in cold regions or has merely taken individual parks as examples. Meanwhile, the number of studies on specific climatic regions is relatively less, and there is a lack of comprehensive research and comparative analysis on cold regions. It also has the defects of a small number of samples and a single type of green space. Additionally, existing research shows that water bodies also have strong cooling effects [34][35][36] . Studies also found that the area and shape index of the water body are positively and negatively correlated with its cooling effect, respectively [37][38][39] . Broadly speaking, urban green spaces include urban blue-green spaces and urban green spaces. However, most existing studies have not separated urban blue-green spaces from urban green spaces. Due to the overlooked influence of water bodies, the research results will be inaccurate. Moreover, there are many uncertainties in existing research on green spaces containing water bodies, especially in terms of the threshold size and the optimal proportion of blue-green spaces [40] . Therefore, this article selected four cities in cold regions of China as case studies; sample green spaces were selected in each selected city according to whether they included water bodies. By analyzing and calculating the TVoE of the cooling effect of two different kinds of green space in each city, this study aimed to find the optimal green space areas in cold regions and to explore the relationship between the UCI effect of urban green space and landscape indicators and background temperature. We expect this work to provide a valuable reference for urban planners and managers seeking to mitigate the impact of UHIs.

Study area
This study selected four cities in cold regions of China for analysis, namely Beijing, Tianjin, Xi'an, and Zhengzhou. The selected standards are as follows: (1) Although the selected city is in a cold area, there are requirements for heat protection in summer; (2) The selected city has a permanent population of more than 1 million; (3) The selected city has a high urbanization rate and a serious urban heat island effect; and (4) The selected city can represent the general geographic characteristics of China's cold regions in terms of temperature, population density, and urban structure. The populations and geographic details of these cities are shown in Table 1 Table 1 Overview of selected cities

Source: CHINA CITY STATISTICAL YEARBOOK (2019)
Note: The study area is the main urban area of each city.

Land surface temperature (LST) retrieval
Previous studies showed that reliable and accurate LST can be obtained by using the Radiative Transfer Equation (RTE) algorithm [16] [22] . Therefore, the RTE algorithm proposed by Jiménez-Muñoz et al. [23] was chosen to calculate LST in this study.
The principle of the RTE involves estimating the impact of the atmosphere on the surface thermal radiation and then subtracting this part of the atmospheric impact from the total amount of thermal radiation observed by the satellite sensor to obtain the surface thermal radiation intensity. After that, LST can be obtained by converting this thermal radiation intensity. The RTE is calculated as follows: [ ] where λ L is a radiance pixel value received by a satellite sensor， ε is the surface emissivity, is the ground radiance, and τ is the atmospheric transmittance. The atmospheric transmittance, the atmospheric downward radiance where 1 K = 774.89 W/(m2*μm*sr) and 2 K = 1321.08 K in Landsat images. Landsat 8 images collected in summer 2019 were used as input data for the LST inversion ( Table 2). Figure 1 shows the results of the LST.  Table 2 The information of Landsat 8 images

Land cover classification
In this study, the ENVI software (Exelis Visual Information Solutions, USA) was used to  Table 2). The Landsat 8 satellite was launched by the National Aeronautics and Space Administration (NASA) on 11 February 2013. The satellite carries two main payloads-the Operational Land Imager (OLI) and the Thermal Infrared Sensor (TIRS)-which are currently used in various fields such as land planning, regional planning, land use, forest monitoring, and agricultural yield estimation.
Visual interpretation methods were used to select training samples for five types of land cover, namely green space, cultivated land, construction land, water, and unused land. Green space refers to land covered by vegetation, which is mainly a mixture of grass, shrubs, and trees; cultivated land refers to land on which crops are grown; construction land refers to buildings and impervious land; water refers to rivers, lakes, and ponds; and unused land refers to land other than agricultural and construction land. The accuracy of the land cover classification results of the four cities were 83.6%, 86.9%, 85.2% and 88.8% respectively.

Sample green space extraction
In previous studies on the cooling effect of urban green space, there were problems such as a small number of samples, only considering a single type of green space, and ignoring interference from other cooling factors. Due to the limited accuracy of land cover classification results, this study uses high-resolution Google Earth satellite images to select sample green spaces for each city based on the results of land cover classification. The selection principles are as follows: (1) The area of the green space is different; (2) The urban green space is classified and selected according to the presence or absence of internal water bodies; (3) There must be a certain distance between the sample green spaces to prevent mutual interference due to proximity; and (4) It was attempted to avoid large external areas of water within 100 m of the green space, which could cause inaccurate data due to the influence of the external water [33] .
According to this principle, based on the results of the land cover classification, and supplemented by high-resolution Google Earth images, the boundary of the sample green space vector was drawn in the ArcGIS software (Esri, Redlands, CA, USA). After screening, in each city, 30 green spaces without internal water bodies and 20 green spaces with internal water bodies were selected. The specific locations of the green spaces selected in each city are shown in Figure 2.

UCI effect analysis
In this study, as in the work by Yu, the UCI effect was defined as the difference in LST between a green space and its surrounding urban areas. Two indicators were selected to characterize the UCI effect of green space: the UCI intensity and the UCI extent. UCI intensity refers to the LST difference between the green space and the turning point of the first LST drop outside the patch. The UCI extent is the distance from the edge of the green space to the first turning point of cooling.
In order to calculate the UCI effect of each green space, we first calculated the average LST inside each selected green patch and then created different numbers of 30m buffer zones for different types of green patch. Finally, the UCI intensity and extent of each green space were obtained, and correlation analyses of the obtained results were carried out with the Microsoft Excel software.

Calculation of TVoE
Yu et al. [26] proposed the use of a TVoE to estimate the optimal green patch area.
According to the "law of diminishing marginal utility," before TVoE is reached, as the area of green patches increased, the UCI effect of green patches also increased significantly. However, after the green patch area reaches or exceeds the TVoE, continuing to increase the patch area will produce a relatively insignificant increase in UCI intensity. Therefore, we believe that it is cost-effective to expand the patch area until the TVoE is reached. This paper used the Excel software to conduct logarithmic regression analysis of the area of green patches and UCI intensity. The TVoE occurs where the slope of the obtained logarithmic function is 1.

Selection of landscape indicators
Previous studies have shown that the UCI effect of green space is correlated with various landscape indices, and the main influencing factors are the size, shape, complexity, and climate background of the green space [41][42][43][44] . The present study started from the patch scale and landscape scale, selected the following indices, performed linear regression analysis, and where E indicates the perimeter of the patch and A is the area of the patch. The larger the LSI value, the more complex the shape of the patch. LSI≥1, with no upper limit.

Basic information
The basic information of each city is shown in    Table 3 Basic information of selected cities   Figure 4 shows the results of the linear regression analysis of the patch area and UCI intensity and UCI extent for different types of green space in each city. As shown, the area of green spaces without water in each city is positively correlated with UCI intensity to varying degrees-that is, as the area of green space increases, the UCI intensity also increases (R 2 =0.24， 0.20，0.32，and 0.28 for Beijing, Tianjin, Xi'an, and Zhengzhou, respectively). There is no correlation between the area of green spaces with water and the UCI intensity in any city.

The influence of Patch Area on UCI intensity and UCI extent
There is no correlation between the area of green space in each city and the UCI extent.

The relationship between LSI and UCI intensity and UCI extent
In this study, the LSI was selected to research the relationship between the shape of green space and UCI intensity and UCI extent. The average LSIs of the two different types of green space in the four cities are shown in Table 4. The results for the linear regression of the LSI for each type of green space and UCI intensity and UCI extent are shown in Figure 5. It can be seen from Table 3 that the average LSIs of the two different types of green space in the four cities are not high, indicating that the shape complexity of the selected green space patches is low and that the shape is basically close to a square.  Table 4 Average LSIs of the two different types of green space in the four cities. It can be seen from Figure 5 that there are no correlations between the LSI of the two different types of urban green space, UCI intensity, and UCI extent in any city. Figure 6 shows the results of the logarithmic regression analysis between the UCI intensity and the area of the two different types of green space in the four cities. It can be seen that the TVoEs of green spaces without water in Beijing, Tianjin, Xi'an, and Zhengzhou are 0.53 ha, 0.57 ha, 0.55 ha, and 0.44ha, respectively. Therefore, for cities in cold regions of China, the optimal patch area of green spaces without water is between 0.44 ha and 0.57 ha.

Analysis of TVoE
The logarithmic regression result (R 2 ) for the UCI intensity of the green spaces with water in the four cities is close to 0, so it is considered that there is no TVoE.
Water has a high heat capacity and low thermal conductivity. The evaporation of water is the main cooling mechanism for water bodies. These characteristics of water lead to a significant reduction in sensible heat transfer capacity, which leads to a change in the heat transfer mode, the so-called "constant temperature effect". The constant temperature effect helps to form a more stable climate, including lowering the maximum temperature and increasing the minimum temperature [45] . Therefore, we believe that for green spaces containing water bodies, due to the influence of internal water bodies, there may not be a universal optimal green space patch area; however, we can further explore the relationship between the internal water-body area of green space and UCI intensity. Zhengzhou, respectively); that is, the larger the proportion of water in the green space, the stronger the cooling effect of the green space.

TVoE of the water/land ratio
In order to quantify the cooling effect of urban green spaces containing water bodies, this paper proposes a new quantitative water/land ratio. The water/land ratio refers to the ratio of the water area to the green area (including the green area and the area for a small number of buildings and road paving) in a single urban green patch. The calculation results of the TVoE of the water/land ratio of the green spaces containing water bodies are shown in Figure   8, which are 0.42, 0.52, 0.50, and 0.55 for Beijing, Tianjin, Xi'an, and Zhengzhou, respectively.

Relationship between landscape indicators and UCI intensity and UCI extent
The results of this study showed that, in cold regions in China, for green spaces without water, the higher the BGT, the lower the UCI intensity and the lesser the cooling effect. For green spaces containing water bodies, different situations were observed in the four cities due to the cooling effect of the water inside the spaces. In Beijing and Xi'an, the correlation between the BGT of the green spaces containing water bodies and their UCI intensity is extremely low, while Tianjin and Zhengzhou have strong negative correlations. Therefore, we believe that there is no correlation between the BGT and the UCI intensity. Studies have shown that water bodies can reduce the maximum temperature and increase the minimum temperature due to their "constant temperature effect". In the present study, it was found that the proportion of water bodies of green spaces in the four cities is weakly negatively correlated with the spaces' internal temperature and is positive correlated with the UCI intensity, which shows that the water bodies inside of the green spaces have a certain impact on the BGT and the cooling effect of the patch. Therefore, researchers should analyze green spaces with water separately in the future. Except for the green spaces without water in Beijing and the green spaces with water in Tianjin and Xi'an, the BGT of all of the green spaces in other cities is negatively correlated with the UCI extent. Therefore, the results indicate that there is no correlation between the BGT and the UCI extent for two different types of green space in cold regions of China. Most of the UCI extent of the green spaces without water in Beijing is 150 m; we think this may be related to the relatively large width of the urban roads in Beijing. Paved roads have a lower specific heat capacity, absorb heat quickly, and dissipate heat slowly, which affect the cooling effects of green spaces. The lack of correlation between the BGT of green spaces with water and the UCI extent in Tianjin and Xi'an may be due to the influence of the water bodies inside the spaces.
The area of green spaces without water in each city is positively correlated with UCI intensity to varying degrees, which is consistent with previous research results [46][47][48] . There is no correlation between the area of green spaces containing water bodies and UCI intensity; however, the proportion of water body in the water-bearing green space is positively correlated with UCI intensity in each city. This indicates that, for the green spaces containing water bodies, the proportion of the area of the internal water body to the total area of the green space is an important factor affecting the cooling effect. Studies have shown that, by increasing the size of the water body, the cooling intensity of the water body increases and its cooling efficiency decreases [49][50][51] . Except for Beijing, the area of green spaces without water in each city is positively correlated with the UCI extent to varying degrees. The correlation between the area of green spaces containing water bodies in Beijing and Tianjin and the UCI extent is extremely low, while that in Xi'an and Zhengzhou is positive. This may be related to the higher BGT in Xi'an and Zhengzhou. Studies have shown that the higher the urban background temperature, the stronger the cooling effect of urban water bodies [52][53] . Therefore, in Xi'an and Zhengzhou, where the urban background temperature is relatively high, green spaces containing water bodies have a stronger cooling effect.
There is no correlation between the LSI of the selected green spaces in the four cities and UCI intensity or UCI extent, which is different from the conclusion of Yu that the LSI of green space is positively correlated with UCI intensity [26] . This may be because the four cities selected in this study all use a square road network, so the selected sample green spaces have relatively low shape complexity and are basically close to a square.

TVoE
In this study, we found that there is a TVoE for green spaces without water in cities in cold regions, which ranges from 0.44 ha to 0.57 ha. Therefore, it can be considered that the TVoE of green spaces without water in cold regions of China is about 0.52 ha. Additionally, this study found that the TVoE is related to the average BGT of the green spaces but not to the average NDVI of the green spaces, which is different from the conclusion of Yu [33] that the TVoE of green spaces is highly correlated with the NDVI and the urban background temperature. For For green spaces containing water bodies, none of the four cities has a TVoE for its area.
Furthermore, considering that water is an important factor affecting the cooling effect of the entire green space, we continued to study the TVoE of the water/land ratio in the green spaces.
We found that the water/land ratio of the green spaces containing water bodies in the four cities is positively correlated with the UCI intensity, with TVoEs of 0.42, 0.52, 0.50, and 0.55 for Beijing, Tianjin, Xi'an, and Zhengzhou, respectively.

Guidance on urban planning and management
Urban green space can play an important role in mitigating the UHI effect due to its cooling effect. This study confirms that green spaces containing water bodies have a stronger cooling effect than those without water bodies. In cold regions of China, as water resources are not abundant, there are few green spaces containing bodies of water. Therefore, water should be designed within green spaces. Previous studies have confirmed that the size of a green space is positively correlated with its cooling intensity; however, there is a TVoE-that is, when the area of green space exceeds this threshold, the cooling efficiency of green space will become lower. This study found that, in cold regions of China, there is only a TVoE for green spaces without water, and the threshold ranges from 0.44 ha to 0.57 ha. However, the average area of the selected sample green space is 3.33 ha, which is obviously bigger than the TVoE. This means that, for green spaces without water in cold regions of China, designers should try to ensure that the area is about 0.52 ha to optimize the cooling efficiency of the green space. For green spaces containing water bodies, although there is no TVoE for their area, when the ratio of the internal water area to the area of the green space is 0.5, the cooling efficiency is the highest. Therefore, more emphasis should be placed on the area ratio of the various design elements inside the green space.

Research limitations and future research directions
First, this study analyzed the cooling effects of two different types of urban green space in cold regions of China. In future research, more cities and more sample green spaces can be selected, and more high-precision satellite images can be used to improve the accuracy of the research results. Secondly, every city is unique, an open and complex giant system with nonlinearity and high uncertainty. Although cities in the same climate region have roughly the same background conditions, there are still internal differences and many factors that affect the cooling effect of urban green spaces-one of which is the presence of a water body.
Additionally, the external environment of the city is also an important factor. Therefore, in future research, more factors that influence the cooling effect of urban green spaces need to be included for a more comprehensive analysis. Generally speaking, an urban water body is called an urban blue space. For urban spaces with both a water body and green space, besides external urban factors, the location and shape of the water body inside the green space will also affect the cooling effect of the entire green space. Determining how to quantify the cooling effect more accurately and put forward guiding opinions that can further guide the planning and design of urban green spaces are also key considerations for future research.

Conclusion
This paper mainly studies the cooling effects of urban green spaces in cold regions of China. Considering that a water body inside an urban green space may affect the cooling effect of the green space, we divided urban green spaces into two types for the present research. The following findings were made: (1) Green spaces containing water bodies have a stronger cooling effect than green spaces without water; (2) For green spaces without water, the BGT is negatively correlated with the UCI intensity while its area is positively correlated with the UCI intensity to varying degrees; due to the differences in the internal climate and construction conditions of each city, the BGT and the area of this type of green space are not related to the

Ethics approval and consent to participate
Not applicable.

Consent for publication
Written informed consent for publication was obtained from all participants.

Availability of data and materials
The data that support the findings of this study available from the corresponding author upon reasonable request.