A Numerical Study on Mitigation Strategies of Urban Heat Islands in a Tropical Megacity: A Case Study in Kaohsiung City, Taiwan

: In recent years, with the rapid increase in global warming and urbanization, urban heat island effects (UHI) have become an important environmental issue. Taiwan is no exception, with previous studies demonstrating serious UHIs in megacities. Although existing UHI research has utilized computer simulations to analyze improvement scenarios, there are few cooling strategy studies in actual blocks of Taiwan. Therefore, this study selected a block of a megacity in a tropical region of Taiwan as a case study by ENVI-met. Five improvement strategies were tested and compared to the current situation (B0): (1) Case C1 changed to permeable pavement, (2) Case C2 increased the green coverage ratio (GCR) of the street to 60%, (3) Case C3 changed to permeable pavement and increased the GCR in the street to 60%, (4) Case C4 changed to permeable pavement, increased the GCR in the street to 60%, and increased the GCR in the parks to 80%, and (5) Case C5 changed to permeable pavement, increased GCR in the street to 60% and parks to 80%, and set the GCR on the roof of public buildings to 100%. The results showed that the average temperature of the current thermal environment is 36.0 °C, with the comfort level described as very hot. Among the five improvement schemes, C5 had the greatest effect, cooling the area by an average of 2.00 °C. Further analysis of the relationship between the different GCRs of streets (SGCR) and the cooling effects revealed that for every 10% increase in the SGCR, the temperature of the pedestrian layer was reduced by 0.15 °C.


Introduction
Since the Industrial Revolution, the rapid shift in land use from natural to agriculture or urban areas has caused the accumulation of artificial matter over the land surface, and the subsequent highrise construction, human activities, and energy consumption have led to global climate change [1][2][3]. According to United Nations 2018 statistics, 55% of the world's population is concentrated in urban areas, which is projected to increase to 68% by 2050, and 90% in high-income countries such as in North America [4]. Therefore, urban environmental problems will inevitably become more serious and important, especially the urban heat island (UHI) effect, when the temperature in the urban center is higher than that of the surrounding suburbs [5].
In Taiwan, a survey of the heat island effect in Taipei Metropolitan Area (25°08′ N, 121°30′ E) in 1999 found that the UHII (heat island intensity) comparing the city center and peripheral suburbs can reach 4-5 °C (day and night maximum values) in summer [6] compared with a 2 °C (night mean value) difference in other high-latitude countries, such as Japan [7]. More recently, in the summer of the actual blocks, while Taiwan, which is located in a tropical island, has rarely conducted research on the improvement strategies of its actual blocks. Although the theoretical mechanism of the urban structure is mature, it is difficult to implement to actual policy. Therefore, this study selected large tropical cities with large heat island effects in Taiwan [6] and used actual CFD simulation tools to analyze their current thermal environment conditions, evaluating the spatial elements of three types of blocks: buildings, streets, and parks. Improvement measures in different heat islands, thermal environment improvement benefits under water permeability, planting species and green coverage, and related mechanisms, are discussed. Finally, suggestions for the improvement of the existing rules on urban planning and design are put forward.

Simulated Area and Site Characteristics
The study site is located in the megacity of the tropical region of Taiwan, the actual center block area of Kaohsiung City (22°6′ N, 120°00′ E). The site area is 23.9 ha, and it is a mixed commercial district. The ratio of land use types and the ratio of buildings on different floors in the study area are shown in Table 1 and Figure 1. The highest proportion of land use type is residential and commercial mixed buildings, which account for about 57%, followed by roads (40%). Building heights of below 5 stories is approximately 74%, 6-14 story buildings account for 22%, and the average building height is 5.2 stories.  The CFD simulation software used in this study was a CFD simulation tool developed by a German research group that suitable for outdoor environments, ENVI-met Version 4.43 (www.envimet.com). The simulation tool is a prognostic non-hydrostatic model composed by a threedimensional main model and one-dimensional atmospheric boundary layer model. Based on the fundamental laws of fluid dynamics and thermodynamics, it can reproduce the outdoor microclimatic and physical situation of urban or rural spaces by accounting for the interaction of surface, plants, and air to evaluate the thermal conditions [51,52]. Several studies have shown that ENVI-met can simulate both spatial and temporal temperature and wind speed for the evaluation of microclimate in both simple and complex urban areas [20][21][22][23][24][25][26][27][28][29][30][31][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48]. A recent review by Tsoka et al. [53] also provided evidence of its suitability for urban climate analysis and examined mitigation strategies.

Parameter Setting
The space setting was as follows: simulation grids 55 × 65 × 40, horizontal grid resolution = 10 m, and vertical grid resolution = 3 m (Table 2) The initial setting of the meteorological environment was based on the days on which the highest summer temperature (from June to September) of the Central Meteorological Administration's Sinsing station [8] from 2014 to 2019 exceeded 35 °C. The initial simulation time was set as 0:00 [54], the initial temperature was 29.7 °C, wind speed 1.9 m/s, wind direction was west wind, and humidity was 52%-76%. The total simulation time was 48 h and the dates after 24 h of simulation were used for analysis and evaluation (Table 3). Other settings of surface materials and planting are shown as Table 4.

Simulation Items and Accuracy Verification
The simulated thermal environment items included air temperature, surface temperature, wind speed, humidity, MRT (mean radiation temperature), and finally, a comprehensive thermal comfort assessment based on the simulated thermal environment results. The thermal comfort indicators used in this research, SET (standard effective temperature) and PET (physical equivalent temperature), include human physiological and environmental factors to evaluate the physiological thermal comfort produced by the heat exchange between the human body and various factors in the surrounding environment, providing a comprehensive assessment of the comfort of the thermal environment. SET is a general comfort index in the field of architecture, proposed by Gagge et al. [55]. It is a comprehensive index with the same thermal sensation obtained by the subjective responses of subjects to different air temperatures, relative humidity, wind speed environment, and solar radiation intensity, and is applicable for outdoor thermal comfort evaluation. PET was proposed by the German scholar, Höppe [56] and is a comfort index suitable for outdoor spaces. It considers all the physiological functions of the human body, including vasoconstriction and expansion, physiological sweating, and physiological thermal balance. It is mostly used for outdoor thermal environment evaluation. The value of SET and PET, which corresponds to the human body's thermal sensation, is shown in Table 5.
The simulation results of the thermal environmental factors and comfort on the selected dates at the hottest time (2:00 p.m.) of the day were comparatively analyzed. The simulation results of each grid, including the average value, the maximum value, and the minimum value of each analysis item, were imported into Excel to calculate.
In the verification of simulation results, because the correctness of the CFD radiation calculation affects the accuracy and reference of the final simulation results, the surface temperature generated by the CFD radiation calculation and the air temperature result obtained by the final fluid calculation were used as reference indicators and to compare the results with similar land cover properties or urban spaces to examine the accuracy and referentiality of the simulation results [14,15]. Table 5. Thermal perception of standard effective temperature (SET) and physical equivalent temperature (PET) [57].

Simulated Configurations
Based on previous research regarding the heat island mechanism, it is known that the land cover type and urban geometry formed by a building group are the main influencing factors [58,59]. However, because changes of building geometry need to change building height and street width, it is very difficult in terms of policy implementation and cost to make changes for a mature existing block. Therefore, the improvement strategies analyzed in this study were mainly on land cover types, including (A) street planting of trees; (B) roof greening strategies, which have been considered in previous studies; and (C) permeable paving with both water retention and heat reduction effects to reflect the flooding caused by global changes. Simultaneously, according to the three most common urban space units, streets, building areas, and parks in actual urban blocks, different degrees of greening or permeable pavement improvement strategies were implemented. The five practical management policies that we used and are easy to implement were: (1) C1 changed to permeable pavement, (2) C2 increased the GCR of the street to 60%, (3) C3 changed to permeable pavement with a green coverage ratio (GCR) of 60%, (4) C4 changed to permeable pavement with a GCR of 60% and a park GCR of 80%, and (5) C5 changed to permeable pavement, GCR was 60%, park GCR was 80%, and public building roof GCR was 100%. They were compared with the current situation (B0) to understand the UHI improvement benefits of each case as a reference for urban planning or design. The detailed simulation conceptual architecture is shown in Figure 2. The detailed description and related settings of each scheme are shown in Table 6 and Table 7.
In addition, due to the degree of street GCR, there should be a significant difference in the cooling effect of the UHI [22][23][24][25], however, there are few studies on real blocks. Therefore, this study further explored the impact on the thermal environment of the different GCRs of trees near streets by adding cases C2a, C2b, and C2c. Simultaneously, due to the previous research of the strategy of roof greening, comprehensive greening roofs were assumed in the study area. Case C5 and C5a were added to compare the research results of this study with other studies and understand the difference between roof greening of total block area (restrictions on private buildings to be greened).

Analysis at Different Levels
As the average number of floors in the building in the simulation area is 5 and the overall average height of the Cianjin District is 21 m, the divisions were pedestrian height (1.5 m), 1F (3 m), 2F (6 m), 5F (15 m), 8F (24 m) (height of less building obscuration) to explore the difference of the improvement of the microclimate environment at the near-surface pedestrian level and different height levels between the cases (Figure 3 and Figure 4). In addition, due to the limitation of the simulation results of the vertical grid, the actual vertical floor plan heights were 1.

Verification of Simulation Results (2:00 p.m.)
The surface temperature simulation results of all schemes are presented in Figure 5, showing that the surface temperature of the artificial pavement was between 50 °C and 60 °C, the surface temperature of the permeable pavement was 30-35 °C, and the ground surface temperature under the canopy was 28 °C, consistent with the trend and numerical range scale of the measured results of related land cover types [60][61][62].
In addition, in all schemes, the average temperature of the streets on the upwind side of the park was 37.5 °C (representing urban temperature conditions without the influence of green space), dropping to around 33-34 °C after entering the park. This effect extended to the adjacent or leeward street ( Figure 6), which matches the trend and numerical range scale obtained from the actual research results inside and outside the park [63,64]. Therefore, the simulation results have a certain degree of reference.

Thermal Environment Status in the Block
The simulation results for B0 of the current situation of the block at H = 1.5 m is provided in Figure 7, showing that the average surface temperature of the street was 50.0 °C, the average wind speed was 0.76 m/s, and the street parallel to the wind direction was east-to-west. The street with the highest wind speed reached an average wind speed of 1.8 m/s more than 2.3 times the block's average. The average air temperature of the block was 36.0 °C. Regarding the comfort results, under the existing environmental conditions, the SET of the block was 38.3 °C on average, which reached the very hot range of thermal sensation. In addition, the average value of PET was as high as 50.51 °C, that is, within the very hot thermal sensation range, only decreasing to 40.4 °C under the canopy of trees on the street (Table 5).

Pedestrian Activity Layer Benefits of Each Improvement Strategy (H = 1.5 m)
3.3.1. Analysis of the Effect of Temperature Mitigation and Mechanism The temperature simulation results are shown in Figure 8a; the maximum cooling effect of C3, C4, and C5 was -1.93, -1.96, and -2.00 °C, respectively, followed by -1.37 for C1, and, the lowest was -0.44 °C for C2. This means that a GCR of the street of 60% and the strategy of changing the pavement to permeable (C3-C5) had the most obvious cooling effects on the overall street. Secondly was the strategies of changing the street to the permeable pavement without planting (C1), and changing the street GCR to 60% (C2) had the lowest cooling effects. Simultaneously, because the proportion of parks and public buildings in the block is only about 2%-7%, increasing the GCR of the park or the roof greening of the public buildings only increased the cooling effect of the overall block by 0.03-0.04 °C. The main reason was the difference in the reduction effect of the surface temperature, as shown in Figure 8b. The C1 case can reduce the average surface temperature of the overall block by as much as 7.90 °C, while C2 can reduce it by 3.24 °C, indicating that surface temperature has an important influence on the pedestrian temperature near the surface layer. If the water can be effectively retained through the sponge city strategy to cause latent heat to evaporate, the improvement in the thermal environment of the block will be greater than the case of a GCR of 60% at street level. Whether the street can maintain good ventilation is also a decisive factor to determine whether the urban surface can discharge heat after absorbing solar radiation. According to the wind speed simulation results in Figure 9, because C1 had no plants, the wind speed at the pedestrian layer increased by about 0.2 m/s, thereby improving the heat emission effect. The other four strategies (C2-C5) included planting trees, reducing the overall wind speed by about 0.1 m/s, but there was not much difference between each of those strategy.

Analysis of the Effect of Comfort and Mechanism
The simulation results of comfort SET and PET are shown in Figure 9 and Figure 10. The differences and characteristics of the various schemes on the SET are similar to those of the air temperature. The reduction effects of C3, C4, and C5 were -2.45, -2.49, and -2.50 °C, respectively, followed by -1.80 °C for C1, and the lowest was -0.61 °C for C2. However, there were different changes in PET, with C3, C4, and C5 all having similar mitigation effects. The largest mitigation effects were -6.83, -7.04, and -7.07 °C,for C3, C4, and C5, respectively, followed by C2 at -2.62 °C, and the lowest was -2.46 °C for C1. In addition, on average, the PET mitigation effect was larger than SET, especially for the most effective C3, C4, and C5 situations, with a difference of 4.2 °C or more. This effect can ease the comfort from very hot to hot and from cold to warm (Table 5).
Although there are many factors that affect human comfort, MRT indicates how much average long and short wave radiation the human body absorbs [64]. The simulation results are shown in Figure 11, demonstrating the maximum reduction effects in MRT in C3, C4, and C5 of -11.31, -11.70, and -11.71 °C, respectively, followed by -5.27 °C for C2, with the lowest being -2.81 °C for C1. This means that C1 receives higher solar radiation because it has no plantings, so the reduction in MRT cannot be higher than the shading effect achieved by planting (C2). However, due to the lower surface temperature achieved by the permeable pavement, which decreases the long-wave radiation, the comprehensive result can still cause a -2.81 °C lower MRT than the current B0 situation. Furthermore, the difference in the reduction effect of C1 on MRT is lower than that of C2, so will not impact its overall effect on comfort. In addition, the effect of this MRT will be directly reflected in PET, and the effect on SET will not be more obvious than PET.

Temperature Mitigation Effects and Wind Speed Changes
The comparison of the temperature of the vertical schemes is shown in Figure 9a, with C3, C4, and C5 having the largest cooling effects, followed by C1 and C2, in other words, the higher the temperature, the smaller the temperature improvement.
In addition, the comparison of the ventilation status of each scheme in each layer is shown in Figure 9b. The changing trend of the schemes from the height level of H = 1 m to H = 7.5 m in the pedestrian layer was consistent. Among them, due to no planting, the difference in C1 above H = 2.7 m is larger than the pedestrian layer. The wind speed increased by 0.25 m/s compared to B0. For heights above H = 16.5 m, there was no significant difference between schemes.

Analysis of the Effect of Comfort and MRT
A comparison of the comfort levels SET and PET in the vertical layers are shown in Figure 9c,d. In terms of SET, the vertical change trends of all schemes were the same, with C3, C4, and C5 achieving the largest reduction, followed by C1 and C2, in other words, the higher the height, the smaller the SET improvement effect. It is of note that the improvement effect of C1 was greater than the result near the ground (H = 1 m) when H = 2.7 m. In terms of PET, the vertical change trend of each scheme was consistent with H = 1 m, with the largest reduction achieved by C3, C4, and C5, followed by C2 and C1, in other words, the higher the PET, the smaller the improvement. It is worth noting that the reduction of C1 at H = 7.5 m suddenly decreased, and the characteristics of this change were consistent with the MRT results (Figure 12e), which shows the criticality of the impact of MRT on the comfort index PET.

Comparison of the Effects Between Permeable Pavement and Greening at Streets (C1 vs. C2)
A comparison of the differences in air temperature and surface temperature between the permeable paving (C1) and tree planting (C2) strategies is shown in Figure 10, with the average reduction in temperature achieved by C1 being 0.94 °C lower than that of C2 (Figure 10a). The main difference was that the surface temperature of the open spaces or non-planted pavement in C1 had a greater effect in reducing the temperature, which was reduced by as much as 21 °C (Figure 10b  A comparison of different green coverage rates in the park area, scheme C3 (total green coverage ratio 60%/arbor 50%) vs. scheme C4 (total green coverage ratio 80%/arbor 75%), is shown in Figure 11. The average improvement in the air temperature and comfort in the park was small, having little impact on the overall block, but the impact of the air temperature was extended to the downwind block.
In addition, under the same park green coverage (total green coverage ratio 60%/arbor 50%), the difference in changing the park's outer block to permeable paving is shown in Figure 12. The strategy of scheme C3 reduced the average temperature in the park by 1.31 °C and SET by 1.73 °C compared to C2, indicating that increasing the green coverage of the park by 20% had a stronger mitigating effect. It is speculated that the possible reason for this is that the cool air generated by the water retention on the street up-wind of the park can flow into the park (average wind speed = 0.69 m/s, westerly).

Influence of Different GCR on Streets
In order to understand the effect of the increase in the GCR on the cooling effect in streets, a linear regression analysis was performed on the green coverage and cooling value (average value) of the four schemes (C2, C2a, C2b, C2c), showing that for every 10% increase in arbor GCR of the street, the temperature of the pedestrian layer was reduced by 0.15 °C (R 2 = 0.98), which was a highly correlated linear relationship (Figure 13a), and the cooling area distribution extended from the tree planting area to the periphery ( Figure 14). Accordingly, the amount of tree planting required to reduce the pedestrian air temperature by 1 °C was 26% of the site area (= 67% street GCR × 39% street area ratio); when the street green coverage ratio (arbor) was 100%, the maximum cooling effect was estimated to reach 1.1 °C.
In terms of comfort SET and PET results, linear regression analysis (Figure 13c,d) showed that when the arbor green coverage of the street increased by 10%, SET and PET reduced the comfort level  Table 5), it was estimated that tree planting of approximately 222% (= 570% street GCR × 39% street area ratio) and 160% (= 410% street GCR × 39% street area ratio) of the GCR is required. Considering that humans living in tropical areas have a high degree of heat resistance [65], and the degree of comfort falls into the warm temperature and cold feeling (SET = 34.0 °C, PET = 34.0 °C; Table 5), the amount of tree planting required is about 117% of the total site area (= 300% street GCR × 39% street area ratio) and 98% (= 250% street GCR × 39% street area ratio) green cover amount. This means that for people in tropical areas, if they want to walk in the urban street space on a hot summer afternoon, the overall street tree GCR (at LAD = 1.3) must reach about 100% of the total site area.
In contrast, the changes in the vertical coverage of different streets in the vertical direction are shown in Figure 13, indicating a negative linear relationship between temperature and GCR at all altitudes (R 2 > 0.95), and the temperature decreases as the vertical altitude rises. This means that the higher the vertical height, the smaller the temperature difference caused by different GCRs. The wind speed results show that, except for the height layers = 7.5 m and 25.5 m, the other height layers have a linear relationship that is negatively related to the GCR (R 2 > 0.99) below the H = 2.7 m layer. As the vertical level increases, the decreasing slope of the wind speed increases, and above H = 16.5 m, the opposite is true, in particular, it may be affected by the planting canopy (canopy height 3-10 m). In terms of the MRT, there was a linear relationship (R 2 > 0.98) that was negatively correlated with GCR at each layer, and the decreasing slope value of the MRT was higher below the height layer of H = 7.5 m and at 16.5 m. It is possible that the height below 7.5 m is affected by the shadowing of the planting canopy (canopy height 3-10 m), so that the radiation in/under the planting canopy decreases as the GCR increases, indicating that the reduction in radiation caused by planting the canopy will be greater than the shadow shielding effect produced by the height of the building. The comfort value SET results, except for the height layer of H = 2.7 m, had a linear relationship that negatively correlated with the GCR (R 2 > 0.94). The trend in the PET results was similar to the MET results, suggesting that PET is greatly affected by MRT.

Influence of Different Building Cover Areas on Roof Greening
A comparison of the effects of comprehensive roof greening (C5a) of the site and the roof greening (C5) of public buildings at H = 1.5 m is shown in Figure 15. In terms of temperature, the average difference was only 0.22 °C. When converted to the difference in the proportion of unit areas (public buildings: all buildings = 7.1:56.6 = 12.5%), C5a increased the area of roof greening by 87.5%, then every 10% increase in roof greening area reduced the temperature by 0.03 °C. It is speculated that the reason for this is that the average roof layer of the site is about 2-4 floors (47.6%, Table 1, Figure 3), so the effect of roof greening is most obvious near the roof. Therefore, further analysis of the results of various meteorological factors in the vertical direction (temperature, humidity, wind direction, wind speed, MRT, PET, SET) revealed that the temperature difference between C5a and C5 reached a maximum of -0.33 °C at H = 16.5 m (5F high), followed by the upper and lower layers, where the temperature difference also gradually decreased ( Table 8). The impact of roof greening is most affected by the average height of the building roofs in the block, mainly manifesting in temperature. At the same time, as the distance in height from the roof is higher, the impact is smaller, therefore, for a pedestrian activity layer with a height of 1.5 m, the temperature effect range was approximately 0.2 °C in the case of a fully greened block roof (thin layer greening H = 0.2 m, LAD = 0.3).
Alcazar et al. [28] used a simple block model and ENVI-met by LAD = 0.3 and 1.5 hypotheses of roof greening (grass flower), to obtain air temperature values at the pedestrian level of approximately -0.2 °C and -1.0 °C, similar to those obtained in this study. Razzaghmanesh et al. [44] performed a simulation analysis on an actual block and found that when H < 20 m, the roof GCR increased by 10%, reducing the air temperature at the pedestrian level by -0.02 °C, which is equivalent to the value scale of this study.

Conclusions
This study focused on the actual central business areas of a megacity in the tropical region of Taiwan, using CFD simulation tools to analyze and understand the thermal environment caused by the current physical environment (B0). Three main street space units and three UHI mitigation strategies feasible for policy control, street arbor greening, roof greening (grass), and permeable pavement, were simulated to explore the effects of different mitigation strategies (C2~C5) on the thermal environment, particularly with regard to temperature and comfort. The vertical thermal environment difference analyses were performed at five levels of a representative height in the block of the site, further analyzing the influences of different proportions for comparison with previous research, to provide important reference results for future government urban planning or design regulations.
According to the simulation results of B0, the current thermal environment, the average temperature in the summer at 2:00 p.m. can reach 36.0 °C, with a maximum street temperature of 38.5 °C. Due to the influence of the high radiation environment (MRT = 62.5 °C) comfort of SET and PET reached a very hot level of human cold and heat experience, which is a poor thermal environment for people to move around in in outdoor spaces. Strategies such as improving the pavement of streets, building areas, and parks, and increasing the green coverage ratio and arbor planting types can reduce the air temperature and increase comfort, thereby improving the thermal environment of the pedestrian layer. Of the five improvement schemes simulated, C5 achieved the most improvement, with the effect of C2 being marginal. It is of note that the cooling effect of water retention on the street pavement was three times greater than that of tree planting on the street, mainly because it effectively reduced the surface temperature of the street pavement (about -11~-20 °C), further reducing the temperature above the pavement (-0.5~-3 °C). As the park occupies a small area within the test area, the impact on the thermal environment caused by changes in the green coverage of the park had little effect on the overall block. A summary of the important results is shown in Table 9. Regarding the vertical layer results, the changing trends of each layer and scheme were consistent with the pedestrian layer, in other words, the higher the altitude, the smaller the temperature or comfort improvement effect. It is of note that the height and thickness of the street canopy will have an impact on the height layer where it is located, especially on the effects of wind speed and radiation MRT. The GCR of different streets and their cooling phenomena was highly correlated and it was estimated that the amount of tree planting required to reduce the pedestrian temperature by 1 °C was 26% of the total site area; if the street green coverage ratio (arbor) was 100%, it is estimated that the maximum cooling for the site as a whole could reach 1.1 °C. By comparing the effects of green roofs on the different building areas, the overall roof greening was only 0.2 °C higher than that on the public buildings on the pedestrian floor, with the cooling effect being similar to the average building height of the site (H = 16.5 m) (0.33 °C). This range was comparable to previous research, indicating that if the setting of the LAD value was different, the temperature effect would be significantly different.
To improve the existing urban thermal environment in Taiwan in the future, it is recommended that (1) the green coverage ratio of the buildings should be at least 60%, (2) if there is no planting in the street, the pavement must be permeable, but if there is planting, the planting species should be mainly arbor, with a green coverage ratio of at least 60%, and (3) the green coverage ratio of the parks must be at least 60% (including more than 50% trees).

Conflicts of Interest:
The authors declare no conflicts of interest.