Cooling effect of the pocket park in the built-up block of a city: a case study in Xi’an, China

Pocket parks, the green infrastructures with small sizes and flexible layouts, are suitable for thermal environment improvement in the urban built-up block with limited green space. To quantify the relationship between pocket parks and the thermal environment in western China, two parks in the built-up block of Xi’an were selected. By field measurement, the cooling effect could be extended 100 m from the park boundary, connecting two parks. Furthermore, the road and greening within the block demonstrate significant influence on the cooling diffusion by regression analysis. Based on ENVI-met simulation, the ratio of the tree and the grass, the layout of the tree and the grass, and the layout of the paving were analyzed at different proportions of greening and paving of the park. Finally, a combination of the daytime physiological equivalent temperature (PET) and nighttime air temperature (AT) was proposed to choose the optimum layouts: the trees concentrated in the center and the pavement with more roads. Results can provide insights for designing pocket parks based on the thermal environment improvement.


Introduction
Due to the rapid urban expansion, the urbanization process has led to significant changes in the urban climate. One of the most obvious features is the urban heat island (UHI), which is defined as the temperature difference between urban areas and surrounding rural areas (Oke 1982;Kalnay and Cai 2003). It affects the outdoor thermal environment, which is closely related to residents' well-being in urban areas (Wang and Akbari 2017;Skelhorn et al. 2017;Li et al. 2020a, b). Several measures were proposed to mitigate UHI, such as changing building surfaces and pavement materials (Evola et al 2017;Taleghani et al 2016), and reconstruction of urban space (Peng et al. 2015;Shaeri et al. 2018). Among them, as a more universal and effective measure, the arrangement of green areas was studied frequently (Gómez et al. 2011;Gunawardena et al. 2017;Du et al. 2019Chen et al. 2020b).
The urban park, a public green area, shows a potential of reducing environmental stress produced by UHI and improving the thermal environment (Feyisa et al. 2014;Yan et al. 2018). At the urban scale, there is a strong correlation between the temperature decrease and the appearance of urban parks (Wong and Yu 2005). for the local scale, the temperature difference from the urban built-up area is more than 3 °C in summer for a single park (Potchter et al. 2006;Toparlar et al. 2018). However, different parks provide cooling effects in different levels (Vieira et al. 2018;Grilo et al. 2020), depending on their characteristics (Georgakis and Santamouris 2017;Lu et al. 2017). Some studies indicated that the cooling effect increases with the size of the park (Chang et al. 2007;Erell et al. 2012) and decreases with the shape index of the park (Feyisa et al. 2014;Chibuike et al. 2018). Furthermore, it is also related to the interior proportions or layout of elements (Amani-Beni et al. 2018;Li et al. 2020a, b;Qiu and Jia 2020). Some studies reported that the temperature of the park is reduced by 1-7 °C by increasing the tree proportion and decreasing the impervious surface proportion ( Oliveira et al. 2011;Ren et al. 2013;Chang and Li 2014;Wang et al. 2018).
In addition, the park's cooling effect also influences its surrounding area (Aram et al. 2019). The cooling degree is attenuated with increasing distance from the park boundary (Doick et al. 2014). For a park with a medium or large size, the range of cooling effect is from tens of meters to more than 1 km (Feyisa et al. 2014;Yan et al. 2018;Aram et al. 2019;Toparlar et al. 2018;Arellano Ramos et al. 2020). Simultaneously, the diffusion intensity of the cooling effect varies in urban spaces with different functions (Grilo et al. 2020). For example, around the parks in Barcelona, the temperature decreases by 2.21 °C in the industrial area and 1.05 °C in the residential area (Arellano Ramos et al. 2020). Moreover, it is also related to the season period, time of the day, and local climate (Fan et al. 2019;Yu et al. 2018). Compared with other seasons, the nighttime cooling effect spreads more obviously in summer (Hamada and Ohta 2010;Ren et al. 2013). Nevertheless, due to local climatic factors, such as wind speed, the effect may also be weaker at night (Toparlar et al. 2018;Lu et al. 2017). Furthermore, the creation of some green infrastructures around the park will help the park's cooling effect spread into the urban space and connect with the green-blue network (Irie 2022;Shi et al. 2020). It can be seen that there is also a mutual influence of cooling effect diffusion between parks, but there is little research on it.
However, the urban space for new parks is generally limited because the spatial pattern in the urban central area is basically finalized in recent years (Peschardt and Stigsdotter 2013). In addition, the establishment of large-scale urban parks is difficult and costly (Zhang and Han 2021). However, the pocket park, a green infrastructure with a small size, low cost, and flexible layout, is especially suitable for the built-up area with limited green spaces (Lam et al. 2005;Abd El Aziz 2015;Wang 2019). It also has the function of regulating the microclimate (Lin et al. 2017;Grilo et al. 2020;Koudouna and Economou 2020). Therefore, the government began to focus on the pocket park development (Peschardt and Stigsdotter 2013;Zhang 2019). In Xi'an (capital of Shaanxi Province, China), from the beginning of 2019, the government planned to build 65 squares and pocket parks with 1000m 2 at least (Xi'an Government 2019). However, in China, most of the studies were carried out in the east and coastal cities (Giridharan et al. 2008;Lin et al. 2017;Motazedian et al. 2020;Zhang and Han 2021), with few investigations in low-latitude inland cities with a different climate. In addition, the objects of previous studies are mainly aimed at medium or large parks, with few comprehensive considerations of the proportion and layout of park elements on the cooling effect of pocket parks. Therefore, in this study, the field measurements and ENVI-met simulation are used to investigate the influence of pocket parks on the thermal environment at the block scale in Xi'an. This study aims to (1) reveal the actual intensity and diffusion range of the cooling effect of the pocket parks, (2) investigate the relationship between the element proportions and layout characteristics of the pocket parks and the thermal environment, and (3) propose the suitable layout of pocket parks to improve the thermal environment at the block scale.

Study area
This work was performed in Xi'an, an inland city of central China in the cold climate zone. In the summer, the highest mean air temperatures in June, July, and August are 33, 35, and 33 °C, respectively. To conduct this study, two pocket parks, "Fenxiang" park (hereby named F park) and "Nan Yuanmen" park (N park), were selected in the built-up block (400 m × 670 m) in the center of the city (Fig. 1). In Table 1, the two parks have similar size (more than 1800 m 2 ). F park is a square space with a higher tree ratio, surrounded by city roads on three sides. N park is a rectangle space with a higher grass ratio and a higher paving ratio, surrounded by roads on all sides.

Air temperature measurements
To make the data spatially comparable, 30 temperature sensors (HOBO® MX2301A) with built-in data loggers (error: ± 0.2 °C in the range of 0-70 °C) within HOBO RS1 solar radiation shields, were equidistant (every 50 m) installations within blocks based on the size of the pocket park, the scope of the block and the diffusion distance of cooling effect on the main parks in Xi'an was about 200 m (Feng and Shi 2012). To facilitate the management and subsequent analysis, these sites were numbered. Among them, A8 and A4 were located in F park, F25 and E26 were located in N park, and the rest were located in the urban space of the block (Fig. 2). They were installed at a height of about 2 m on a tree or a telephone pole in order to ensure the safety and avoid human interference (Fig. 3). These sensors were programmed to collect the air temperature (AT) data every 3 min. The valid recording time was 24 h, from 00:00 on July 26 to 00:00 on July 27, 2019. The measurements were carried out during clear and windless days in order to minimize the influence of meteorological variables.
Prior to the measurement, the sensors were standardized by measuring in the same outdoor environment for 24 h with the same settings so as to avoid the influence of their ± 0.2 °C differences (Fig. 3). The standardized temperature data were used for subsequent analysis. The formula was as follows: T′ standardized temperature for analysis T a actual measured temperature T measured temperature in the standardized site N total number of measurement sites, i.e., 30

Variable selection and multiple linear regression
The multiple linear regression is to study the influence of urban elements on the diffusion of cooling effect. The Table 2, including floor area ratio (FAR), building height (BH), building density (BD), greening ratio (GR), and road ratio (RR), were chosen based on the controlled detailed planning in the Measures for the Examination and Approval of the Compilation of Controlled Detailed Plans for Cities and Towns issued by the Ministry of Housing and Urban-Rural Development of the People's Republic of China. The calculation range of the elements was a circular buffer zone with a 25 m radius, centered at each measurement point (area = 1962.5 m 2 ). Although some studies employed the buffer area with more than a 100-m radius (Chen et al. 2012;Oke 2004;Yan et al. 2018) to obtain a stronger correlation, the 25-m radius was found to be sufficient to obtain reliable correlations between landscape characteristics and air temperature (Cheung and Jim 2019a;Konarska et al. 2016;Petralli et al. 2014). These element ratios were acquired and calculated by BaiDu Map. The multiple linear regression was applicable to the prediction or estimation of dependent variables by the optimal combination of multiple independent variables (Heusinkveld et al. 2014;Lü et al. 2015;Onishi et al. 2010;Shih 2017;Yu et al. 2020). Prior to the statistical analysis, the Pearson   correlation coefficient (two-tailed) was used to test the multicollinearity. In Table 3, a high correlation coefficient between BD and FAR was considered because the absolute correlation coefficient between them was larger than 0.7 (Brun et al. 2020;Yu et al. 2015). Therefore, one of these two elements was retained in the analysis model. Based on the statistical significance of variables (p > 0.05), the maximum R 2 and minimum standard error principles were estimated, and the best linear regression equation for dependent variables was determined. The standardized coefficient (beta value in Table 6) was used to determine the relative importance of independent variables. All statistical analyses were conducted with IBM SPSS Statistics 23.

ENVI-met simulation
ENVI-met is used to study the effect of proportion and layout of park elements on the thermal environment improvement. It is a simulation tool to recreate urban 3D models by deterministic equations that couple thermal and fluid-dynamics processes (ENVI-met 2020). It was proved to be reliable in many urban studies on the thermal environment with appropriate validation (Wang and Akbari 2014;Lin and Lin 2016;Simon et al. 2018). The 3D model, 468 m × 717 m × 60 m (x-y-z), was built using cubic grids of 3 m (x axis) × 3 m (y axis) × 3 m (z axis). The modeling process was shown in Table 4. Then, the applicability of ENVI-met was verified by using the 3D model. In Fig. 4, the trends between the measured and simulated data were similar, the mean squares of the correlation coefficient (R 2 ) of the linear fitting equations were 0.94 in the park and 0.96 in the block, which means that values simulated with ENVI-met strongly converge with the field measurements. Moreover, the mean root mean square error (RMSE) and mean absolute percentage error (MAPE) were also used to evaluate the validation of ENVI-met. The RMSE is a commonly used index to evaluate simulation accuracy. The MAPE is a percentage to measure the model error and is not affected by the value range of the original data and is suitable for comparisons between different datasets (Jin et al. 2017;Zhang et al. 2022). For the model in this study, the RMSE was 1.36 ℃, within an acceptable range of 1.31-1.63 ℃ (Chow et al. 2011), and MAPE was 3.31%, less than the effective range of 5% proposed by some studies (Gou et al. 2021;Zhang et al. 2022). Therefore, the ENVI-met model was considered to have sufficient accuracy to use in the study. Based on the requirements of the "Code for the design of public park of China" (2016) for the proportion of land types of parks with an area of less than 2000 m 2 (greening > 65%; paving 15 ~ 30%), the 3D model of the block and the interior layout models of the two parks were built in Table 4. The models were set up on the basis of minimum and maximum limits of paving ratio (minimum: paving 15%, greening 85%; maximum: paving 30%, greening 70%). Different greening layouts (tree in the center, grass around; grass in the center, trees around; tree and grass are evenly distributed; only tree; only grass), and different paving layouts (concentrated; dispersion; only road) were designed. ENVI-met supports 2D and 3D plant options. 2D plants need the basic information of plants, while 3D plants allow the user to set the Leaf Area Density (LAD) for each cell and analyze  individual trees at the scale of a typical crown geometry with the resolution of a single branch. 3D plants provide more detailed settings and analysis of the plants, but it will cost much more time than 2D plants to run the simulations. Because this study focuses on the layout of plants rather than the characteristics of their own, 2D   Fig. 4 The trend throughout the day and linear fitting between measured and simulated average air temperatures plants met the demands and were used in this study. The models were simulated in the block model with the input parameters in Table 5.

Data processing
In field measurement, the hourly air temperature was calculated as the average of the following 20 temperature values (recorded every 3 min) within 1 h. For example, the 8:00 air temperature was derived from 8:00 to 8:57. In ENVImet simulation, direct results at the height of 2.1 m at the same points of the measurement including the air temperature (AT), mean radiation temperature (MRT), wind speed (WS), relative humidity (RH), and physiological equivalent temperature index (PET) (Cheng et al. 2015;Vellei et al. 2017;Xie et al. 2018) were adopted to analyze the outdoor environment. In this study, based on the research purposes and the relevant literatures, PET was selected as one of the thermal indexes with some advantages. Firstly, PET is a commonly used human thermal index which had a widely known unit (°C) as the measurement of thermal stress (Höppe 1999;Gulyás et al. 2006). Secondly, some studies reported that PET facilitates application and makes thermal stress understandable and comprehensible for the users not familiar with modern human biometeorological terminology, including urban designers, landscape architects, policy makers, and the lay people (Lin et al. 2010;Cohen et al. 2019).
Thirdly, PET could be calculated using available software packages, such as Biomet of ENVI-met and the RayMan software ( Matzarakis et al. 2007;Fang et al. 2018). Fourthly, in the cold climate area, PET has been used as a thermal comfort indicator in previous studies (Chen et al. 2020a, b;Yu et al. 2022), which was proved markedly more suitable in the warm months of China (Yang et al. 2021). In addition, due to the different thermal environment in the daytime and nighttime, these data were processed into diurnal temperature (from 6:00 to 18:00) and nocturnal temperature (from 19:00 to 5:00) and analyzed separately.

Data and analysis
The cooling effect of the parks At present, various indexes have been proposed to quantify the park cooling effect, such as the cool-island intensity ( Chang et al. 2007), the park cooling intensity (Feyisa et al. 2014), and the cooling distance (Toparlar et al. 2018). However, the cooling distance did not use a common definition of the indexes to measure the park cooling effect (Peng et al. 2021). Therefore, according to the purpose of this study and the definition of relevant indicators, park cool island was calculated by the difference between the average park temperature and the average urban temperature in this study (Yan et al. 2018). In Fig. 5, the air temperature (AT) changes showed that the variation trends inside the two parks were lower than those outside the parks. The trend of the relative humidity was opposite. At 10:00, there was a small change in temperature, which is due to a sudden increase in humidity caused by the sprinkling of urban roads and the irrigation of parks (photos in Fig. 5). The average difference of 0.43 °C indicated that pocket park had a cooling effect in the block. During the whole day, the AT inside F park was lower than that outside the park, and the AT difference inside and outside N park was more obvious before 6:00 and after 18:00. The AT difference inside and outside F park was 0.53 °C higher than that of N park. Furthermore, the results indicate that at 10:00, 15:00, 19:00, and 23:00, the temperatures inside the park were significantly lower than the temperatures outside the park. For F park, the mean cool island intensity was 0.61 °C, 0.9 °C, 0.65 °C, and 1.13 °C, respectively. For N park, the mean cool island intensity was 0.09 °C, 0.22 °C, 0.2 °C, and 0.14 °C, respectively. Therefore, F park had stronger cooling intensity than N park. This may be caused by the different ratios of elements in the two parks (Table 1).

Diffusion and connection of park cooling effect
In Fig. 6, as the distance from the park margin increased, the air temperature continuously increased and the cooling effect extended to 100 m. From the park boundary to a distance of 100 m, a gradual increase in AT was observed with 0.24 °C in the daytime and 0.39 °C in the nighttime on average. From a distance of 100 m to that of 150 m, a decrease in AT was observed with 0.34 °C in the daytime and 0.18 °C in the nighttime. Furthermore, in Fig. 7, the cooling effect was more obvious at 23:00, when the coefficient of determination R 2 was 0.663. For F park, at 15:00, 19:00, and 23:00 the mean AT increased by 0.27 °C, 0.3 °C, and 0.41 °C. The cooling effect extended 150 m. For N park, at the same time, the mean AT increased by 0.08 °C, 0.19 °C, and 0.48 °C, respectively. The cooling effect extended 100 m. The diffusing distance of the cooling effect of the two parks is not only due to different internal factors, but also related to the surrounding urban elements. Therefore, multiple linear regression was applied to the analysis. In Table 6, in the daytime, the variable with a significant effect (p < 0.05) was RR. According to the standardized coefficient (Beta), the increase in AT of the block space by 0.06 °C was accompanied with every 10% increase in RR, indicating the impedance of the diffusion of the park cooling effect by the roads. For the temperature in the nighttime, the variables with significant influence were GR and RR.  At night, the decrease in AT by 0.05 °C was accompanied with the increase in GR by 10%. However, the increase in RR by 10% was accompanied with the increase in AT by 0.03 °C. Compared with other factors, the influence of roads on temperature is more obvious during the whole day. It is because urban roads are mostly made of dark concrete or asphalt pavement materials with high specific heat capacity and high absorptivity and emissivity, which can store solar energy and anthropogenic heat energy generated by vehicles. And then, the energy is released as sensible heat, raising the air temperature (Li et al. 2013;Cheung and Jim 2019b). In addition, the boundary of two pocket parks is close to the  urban roads, which also can increase the influence in the results. From the result, the cooling effect of trees was more significant and the warming effect of roads was weakened at night. For the trees, although the shade can hinder the warming effect of solar radiation in the daytime, it is difficult to form a wide range of shade that has an obvious cooling effect on pedestrian height. At night, the temperature under the tree is lower than that in other spaces, and the airflow can take away the released heat more quickly (Erell et al. 2012).
For the road, it can receive more solar radiation to raise the temperature during the day, and its ground materials can release heat concurrently. At night, only the heat release of surface materials exists, so the warming effect is weaker than that in the daytime. According to the results of multiple linear regression, the relationship between the block space and the diffusion of the cooling effect was further analyzed. In Fig. 8, the increasing distance from the F park boundary was accompanied with the decrease in GR and the increase in the RR of the block. Around N park, the increasing distance from the park boundary was accompanied with the decrease in RR. Compared with N park, within 50 m, GR in F park was 3.56% higher and the road ratio was 6.81% lower. Within 100 m, GR in F park was 4.29% lower and RR was only 0.39% lower. Combined with the cooling effect diffusion of the two parks in Fig. 7, GR and RR within a radius of 50 m away from the park boundary had a relatively important influence on the temperature.
In addition, the cooling effect between the two parks was analyzed by temperature changes at B9, C10, and D14. In Fig. 10, the mean AT of F park and N park was found to increase with increasing distance from the park boundary and reach the maximum value at C10 (100 m away from both parks) in the daytime and at D14 (150 m away from F park and 50 m away from N park) in the nighttime. However, the mean AT of the three points between the two parks was higher than that of the other block spaces. In Fig. 9, the GR at B9, C10, and D14 were significantly lower than that of other block spaces, while the RR was higher. Meanwhile, the RR of B9 and D14 were higher than that of C10. Based on the previous statistical analysis (Table 5), the urban elements of the three points were not conducive to the diffusion of the park cooling effect. Combined with the AT variation trend in Fig. 10, it indicated the influence of the diffusion and superposition of the cooling effect between the two parks.

Influence of the design elements on the thermal environment
The different ratio of the tree and the grass Figure 11 displays an increase in the tree ratio by 50% accompanied with the PET decrease by 2.09 °C on average in the "Paving 15%; Greening 85%" (P 15 G 85 ) and 1.66 °C on average in the "Paving 30%; Greening 70%" (P 30 G 70 ) in the daytime, and a decrease by 0.51 °C on average in the P 15 G 85 and 0.55 °C on average in the P 30 G 70 at night. This suggested that the increasing tree ratio had a more significant effect on PET in the daytime. The increasing tree ratio by 50% had a greater effect on PET in the P 15 G 85 in the daytime and in the P 30 G 70 in the nighttime. For AT, the increasing tree ratio decreased the AT by 0.16 °C on average in the daytime, larger than that at night. In the P 15 G 85 and the P 30 G 70 , from "G100%" (G 100 ) to "T50%G50%" (T 50 G 50 ), the AT of the block decreased by 0.03 °C on average in the daytime and was unchanged at night. It was also unchanged at night from T 50 G 50 to "T100%" (T 100 ). For MRT, the increasing tree ratio by 50%, was accompanied with a decrease in the daytime and an increase at night. In the daytime, from G 100 to T 50 G 50 , it decreased by 0.63 °C on average in the park, larger than that from T 50 G 50 to T 100 . For the block, in the P 15 G 85 , the MRT decreased by 0.15 °C, larger than that in the P 30 G 70 . However, at night, from G 100 to T 50 G 50 , it increased by 0.57 °C on average in the park, larger than that from T 50 G 50 to T 100 . In the block, in the P 15 G 85 and the P 30 G 70 , the MRT decreased by 0.02 °C. For wind speed, from G 100 Fig. 8 The greening ratio (GR) and road ratio (RR) in the block at different distances block spaces Fig. 9 Comparison of the urban elements of the three points (B9, C10, and D14) between the parks and others from their boundaries Fig. 10 The mean air temperature (AT) between the two parks and that in other block spaces during the daytime and nighttime Fig. 11 PET, air temperature (AT), mean radiation temperature (MRT), wind speed, and relative humidity (RH) of the two parks and block within the block in the different ratio of the tree and the grass during the daytime and nighttime to T 50 G 50 , it decreased by 0.14 m/s on average in the park and lower than 0.1 m/s in the block. From T 50 G 50 to T 100 , the wind speed in the park and block had no noticeable changes. At last, in the daytime, from G 100 to T 50 G 50 , the RH increased by 0.64% on average in the park, larger than that from T 50 G 50 to T 100 . In the block, the RH increased by 0.04% on average, larger than that from T 50 G 50 to T 100 . However, in the nighttime, the RH difference was 0.05% in the park and unchanged in the block. Therefore, increasing the tree ratio had a greater effect on RH in the park, and the change was greater in the daytime than that at night.

The different layout of the tree and the grass
In the case of T 50 G 50 , the impact of their layout was further analyzed. In the park in Fig. 12, the PET of "Trees (T) around" pattern was the highest, followed by the "Trees (T) evenly distributed," and the "Trees (T) in the center" was the lowest. The average PET difference between each layout was 1.72 °C in the daytime. However, in the nighttime, the change of PET was opposite to that in the daytime, with an average difference of 0.32 °C between each layout. In the block, the three layouts of PET showed the Fig. 12 PET, air temperature (AT), mean radiation temperature (MRT), wind speed, and relative humidity (RH) of the two parks and block of the block in the different layout of the tree and the grass during the daytime and nighttime same trend as the park in the P 15 G 85 , while the PET of "T around," and "T evenly distributed" in the P 30 G 70 were the same, and that of "T in the center" was the lowest. For AT, the trend in the daytime was similar to that of PET, with an average difference of 0.06 °C between each layout. Meanwhile, the change of AT in the nighttime was similar to that in the daytime, with an average difference of 0.03 °C between each layout. In the block, the difference between each layout was very small, with an average difference of 0.01 °C. For MRT, the trend was similar to that of PET, with an average difference of 3.18 °C in the daytime and 0.79 °C in the nighttime between each layout in the park. In the block, in the P 15 G 85 , the MRT of "T in the center" was the highest, followed by the "T around," and the "T evenly distributed" was the lowest. However, in the P 30 G 70 , the MRT was the same for the three layouts. For wind speed, the difference for the three layouts was small in the park and block, with an average difference of 0.01 m/s. For RH in the park, the trend was also similar to AT, with an average difference of 0.28% in the daytime and 0.12% in the nighttime between each layout. However, the RHs of the three layouts were slightly different in the block. The different layout of the paving In Fig. 13, in the daytime park, the PET of "Dispersion" was the highest, followed by "Concentrated", and the "Only road" was the lowest. The average difference was 0.07 °C between the three layouts. However, in the nighttime, the PET of "Concentrated" was the highest. In the block, in the P 15 G 85 , the PET of "Only road" was the lowest, with an average difference of 0.13 °C in the daytime and 0.02 °C in the nighttime for the three layouts. In the P 30 G 70 , in the daytime, the PET of "Only road" was the lowest, with an average difference of 0.07 °C for the other two layouts. For AT, in the park, it was the lowest in the "Concentrated" during the daytime, with an average difference of 0.02 °C for other two layouts. In the block, the AT of "Only road" was the lowest in the daytime, while that of the "Concentrated" was the lowest in the nighttime. However, the difference for the three layouts was not significant, with an average difference of 0.02 °C in the daytime and 0.01 °C in the nighttime. For MRT, in the park, it was the lowest in the "Only road" during the daytime and in the "Concentrated" at night. Since the "Only road" could be shielded by more trees to reduce the direct solar radiation in the daytime, the "Concentrated" had more open space to allow the ground long-wave radiation heat dissipation at night. In the block, the MRT of the "Concentrated" in the daytime was the lowest, with an average difference of 0.1 °C for the other two layouts. At night, it was the lowest in the "Only road", with an average difference of 0.02 °C for the other two layouts. For wind speed, the overall change was not significant, with an average difference of 0.02 m/s between the three layouts. For RH, in the park during the daytime, it was the highest in the "Dispersion". At night, the RH of "Dispersion" was the highest in the P 15 G 85 , and that of the "Only road" was the highest in the P 30 G 70 . In the block, in the P 15 G 85 , the average difference was only 0.01% on average between the three layouts in the daytime. In the P 30 G 70 , the RH of "Dispersion" was the lowest, and the other two layouts were the same. In the nighttime, the RH of "Only road" was the lowest.

Comprehensive analysis of different layouts
The simulation results showed the effects of different ratios and layouts of elements on various factors of the thermal environment. A comprehensive evaluation method was designed in Fig. 14, mainly based on the main function of the pocket park during the daytime and nighttime. In the daytime, pocket parks were mainly used for personal outdoor activities. Therefore, the PET, an indicator of body thermal comfort, was selected as the daily evaluation index. At night, it makes more sense to reduce the building's energy consumption by lowering the outdoor temperature, so the AT was the main consideration index. Figure 15 presents the comprehensive evaluation results. Among the layout patterns with different ratio of the tree and the grass, in the P 15 G 85 , from G 100 to T 50 G 50 , the combined value of daytime PET and nighttime AT decreased by over 0.35 °C in the park and over 0.06 °C in the block, compared to the value change from G 100 to T 50 G 50 . In the P 30 G 70 , although the combined value decreased more from T 50 G 50 to T 100 in the park, it decreased more from G 100 to T 50 G 50 combined with the change value of the block. As to the different layout of the tree and the grass, in the P 15 G 85 and the P 30 G 70 , the combined value of the "T in the center" in the park and block was lower than other two layouts. Therefore, the "T in the center" was more beneficial to the improvement of the thermal environment. For the layout of the paving, the combined value of the "Only road" was the lowest in the P 15 G 85 and the P 30 G 70 , which was more beneficial to the improvement of the thermal environment.

Discussion
In this study, lower temperatures in the pocket park with a small size than that in the surrounding block were confirmed by the field measurement. Although some parks smaller than 5000 m 2 did not affect the surrounding microclimate in London ( Monteiro et al. 2016), small green spaces with only 10 m 2 could produce a cooling island effect by changing the tree coverage ratio in Hong Kong (Giridharan et al. 2008). For this study, two pocket parks, about 2000m 2 , were measured and simulated to produce a limited range of park cooling island intensity. In different cities, the park with 2000m 2 had a cooling range from 1 to 6.9 ℃ (Spronken-Smith and Oke 1998;Oliveira et al. 2011;Cohen et al. 2012). Thus, the size was not the only determinant of the cooling effect of a park (Cheng et al. 2015), and it's the regional climate and local environment might be other factors. This study not only explored the cooling effect of a single pocket park with an area of about 2000 m 2 but also obtained the connection of two parks of this small size that have cooling effects within a 200-m interval in the block. In addition, the cooling effect of pocket parks was found to extend 100 m at pedestrian height. The extension distance of the cooling effect was not clearly quantified due to the complex process of heat transfer between the urban elements. Therefore, the researchers could only estimate the distance by comparing the cooling effects among all scenarios (Lin and Lin 2016; Yan et al. 2018). In previous studies, there were few studies on the cooling distance of the pocket park , so this study also used this method to pay attention to it.
In this study, the extension of the cooling effect was mainly affected by the surrounding greening and roads. The temperature could be lowered by the greening, while it could be raised by the roads (Armson et al. 2012;Li et al. 2013;Lee and Jim 2018). The capacity of pocket park to diffuse the cooling effect depended on the space with dark surfaces around it, such as roads, because they could absorb more energy and get hotter to affect the temperature (Grilo et al. 2020). Therefore, in built-up areas, adjacent wider urban roads should be avoided when choosing the location of pocket parks. However, there was no significant correlation between the index elements of the spatial form (building density, average building height, and floor area ratio), because the difference in spatial form was not particularly obvious around measuring points located in the same block with high density. These results can have a guiding effect on the location selection of the pocket parks in the built-up areas.
This study also confirmed the cooling effect connection between two pocket parks in the straight line distance of 200 m, and this could be used as the spacing reference. Generally speaking, pocket parks had the advantage of being small enough to be built in the built-up block with high density. Thus, the total park area could not be expanded without limitations. To achieve a more dramatic cooling effect within a limited total park area, the distance from the two Fig. 15 The PET in the daytime and AT in the nighttime of the park and block park arrangements should be considered. Furthermore, as seen from the cooling effect diffusion in the space between the two parks, F park could reach 150 m at night. Therefore, through some changes inside and outside the park, the distance between the two was estimated to be larger than over 200 m, and this will be further studied in the future.
For the layout of the interior of the park, by using the ENVI-met simulation, it was found that the thermal environment was changed more by the increasing trees in the daytime than in the night. Meanwhile, the temperature was lowered more by adding 50% more trees to a grass-only site than to the site with half trees and half grass. This was due to the overlapping of previous shade by the trees newly added. However, in general, trees and grasses were present together in the greening of a park. Therefore, the influence of the layout of trees and grasses was also explored in this study. The "tree in the center, grass around" had a positive impact on the thermal environment, because the cluster trees would generate better cooling effect than grass and scattered trees (Amani- Beni et al. 2018). For the paving layout, the "Only road" had a positive impact on the thermal environment. Although, in the "Concentrated," the larger open space promoted the heat dissipation from the park at night, resulting in a slight temperature decrease (Rosenberg et al. 1983). In the "Only road," the linear layout of the paving allowed the trees to shade more effectively against the floor's absorption of solar radiation, thereby keeping the temperature low. These results suggested that the tree layout and paving layout of pocket parks were relevant in view of optimizing the cooling effect.
For other cities, the layout and interior design of the pocket park will be different according to the urban space form. However, this study confirmed that the pocket park has a cooling effect and a certain range of cooling distance from the perspective of measurement and simulation. In Shanghai, another city in China, the average cool island intensity of pocket park is 1.2 ℃, and the average cooling distance is 132 m ( Wu et al. 2021). This study also confirmed that the cooling distance of the pocket park can reach 100 m. It can be seen that although the two cities have different climate and urban space forms, there are some similarities in the cooling effect of pocket parks. Therefore, the results can also be used for reference to other cities to a certain extent. Moreover, a more accurate and detailed planning reference for other cities can also be obtained by the method of this study.
This study also demonstrated some limitations and could be improved in the future. For the field measurement, due to the management and logistic limitations, the monitoring period of this study was restricted to 24 h in the summer of 2019. Therefore, the year-to-year variability in climate was not considered in the analysis. For the two pocket parks with similar sizes, the relationship between the size and the cooling effect was not clear. Simultaneously, for the location of the two parks, the same block with a similar urban spatial form may affect the result of the regression model. For the models, the plant diversity was not fully considered since only two typical plants, trees and grasses were selected. These aspects could be combined for further exploration in the subsequent research.

Conclusion
In this study, the field measurement and ENVI-met simulation were employed to evaluate the relationship between pocket parks and the thermal environment in the built-up block. The cooling effect of the pocket park was found at the block scale under weather conditions in the Northwest of China and could be spread to 100 m beyond its boundary. A cooling effect connection was found between the two parks at 200 m from the straight line, and this could be used as a reference distance for setting up pocket parks in high-density built-up areas. The diffusion of cooling effects was inhibited by the roads around the pocket park. For a 50% increase in the area of the surrounding road, the daytime AT was increased by 0.3 °C, twice as large as the increase at night. Thus, the urban space near the wider roads was recommended to avoid when setting up the pocket park. By comprehensively considering daytime PET and nighttime AT, for the concentrated trees in the pocket park in the middle layout, it was 2.79 °C lower on average than the other two layouts in the park and 0.06 °C in the block. For the paving with more road space, it was 0.14 °C lower on average than the other two layouts in the park and 0.07 °C in the block. This indicated that both the increase in the green coverage ratio and the optimization of the layout structure within the green space contributed to the alleviation of the UHI effect. These results offered a reference for urban planners to increase pocket parks in built-up areas with higher density based on the improvement of urban thermal environment.