Unpacking Park Cool Island Effects Using Remote-Sensed, Measured and Modelled Microclimatic Data

Abstract

There is increasing interest in the role of parks as potential cool refuges in the age of climate change. Such potential refuges result from the Park Cool Island (PCI) effect, reflecting the temperature differential between the park and surrounding urban areas. However, this study of different park typologies in Perth, Australia, illustrates that while surface temperatures are 10–15 °C lower in parks during summer afternoons (much less than at other times), air temperatures are generally no different from the adjacent streetscape for the smaller parks. Only the largest park in the study had 1–2 °C lower morning and mid-afternoon air temperature differentials. The study illustrates that while the PCI is a real phenomenon, the magnitude in terms of air temperature is small, and it is of less relevance to the conditions felt by humans in average summer daytime conditions than the direct effects of solar radiation. Many studies have assessed the PCI effect, an indicator that has shown a wide range across different studies and measurement techniques. However, this novel paper utilises satellite remote-sensed land surface temperatures, on-ground measurements of surface temperatures, air temperatures, and humidity, as well as modelling using the microclimatic simulation software ENVI-met version 5.0. A reliance on land surface temperature, which in isolation has a marginal correlation with human experience of thermal comfort, has led some researchers to overstate the PCI effect and its influence on adjoining urban areas. The research reported in this paper illustrates that it is the shade provided by the canopy in parks, rather than parks themselves, that provides meaningful thermal comfort benefits. Accordingly, adaptation to increasing temperatures requires the creation of a continuous canopy, ideally over parks, streetscapes, and private lots in an interconnected network.

1. Introduction

Many parts of Australia already have extensive exposure to extreme summer conditions. Perth (the capital city of Western Australia and the focus of this paper) has summer days that often exceed 40 °C, which is projected to worsen with climate change. Recent updates on global warming [1] project that, with current real-world policies and actions (as opposed to national pledges and targets), global temperature increases of 2.2–3.4 °C above pre-industrial levels are likely by the end of the century and will continue to rise thereafter. These conditions will have dire implications for Australian cities, particularly for residents of suburban areas with minimal tree canopy cover, whether middle-ring suburbs experiencing urban infill [2] or compact new suburbs on the urban fringe [3].

1.1. Urban Heat Islands and Park Cool Islands

The Urban Heat Island (UHI) is a well-recognised phenomenon in cities. It arises from retaining solar radiation-induced heat in ‘impervious’ urban surfaces with high solar absorptivity and heat capacity, such as roads, pavements and building surfaces [4]. The UHI effect can lead to land surface temperatures 10 °C higher than those of their peri-urban surroundings [5], and it is more pronounced at night than during the day [6].

In contrast, vegetated surfaces in parks and gardens absorb and re-radiate less heat, resulting in lower land surface temperatures. This phenomenon has come to be known as the Park Cool Island (PCI) effect [7] (Figure 1). Urban designers are increasingly interested in understanding the effects of UHI and PCI on improving summer thermal comfort conditions in urban settings, particularly the role of parks as refuges. Moreover, parks are also described as mitigating UHI effects in high-level policy documents, such as Australia’s recent National Urban Policy, which describes how ‘Green spaces play a key role in mitigating the urban heat island effect by providing shaded pathways, cooler environments and more comfortable conditions’ [8].

Many studies have assessed the PCI effect, which reflects the temperature differential between the park and surrounding urban areas, which are typically hotter in the middle of the day, an indicator that has shown a wide range across different studies and measurement techniques [9].

The literature includes studies based on the measurement of land surface temperature by remotely sensed surface temperature data [10] or thermographic cameras [11]. As noted by Liu et al. [12], satellite-derived surface temperature is the most prevalent method identified in the literature to identify the PCI effect. Other studies involve the measurement of air temperature [13,14]. These measures reflect the urban energy balance but have different implications for human thermal comfort. Additionally, the literature also includes studies suggesting that the PCI contributes to the mitigation of UHI through the cooling of surrounding urban areas [7,15], which can even reduce energy for cooling in buildings [16,17]. An explanation of the urban energy balance is necessary to resolve the importance of these observations in the study of urban microclimatic conditions.

1.2. The Urban Microclimate

The UHI effect represents the difference in temperature between urban areas and their surrounding non-urbanised areas [4]. At the metropolitan scale, the difference is between the ‘city’ and the adjacent ‘countryside.’ At the urban precinct scale, the differential of interest lies between urban streetscapes and vegetated areas, whether these are landscaped parks or conservation areas. When referring to UHI, it is important to define what is meant by both ‘temperature’ and the location of the temperature measurement. Regarding UHI studies, air temperature near ground level is the most relevant, noting that air temperature varies with elevation. This measure of UHI is known as the Canyon Urban Heat Island (CUHI), which is the layer of air between the ground and the tops of buildings and trees [18]. When the measurement is land surface temperature, this is known as the Surface Urban Heat Island (SUHI) [4]. In urban areas, the near-surface air temperature is usually very different from that of the surface, as it is influenced by factors such as wind, shading, turbulence, and the three-dimensional structure of the urban environment [18].

In essence, the UHI arises when the heat energy stored through absorbed solar radiation during the day within urban elements with high thermal mass (such as roads and buildings) is released to heat the near-surface air during the late afternoons and evenings. The magnitude of the UHI effect depends on the amount and rate of energy stored and released, which are functions of solar exposure, albedo, heat capacity, thermal conductivity, and emissivity of the various surfaces. Measurements of near-surface air temperature during a heat wave in Birmingham, in the United Kingdom, reported by Stewart and Mills [4], indicated that the city centre was hotter than an urban park from around 4 pm, peaking at 5 °C around 10 pm. Air temperature differentials (ΔT_a) were less than 1 °C during most of the daytime (Figure 2).

This behaviour is because it takes time for ground surfaces exposed to solar radiation to heat, only emitting long-wave radiation later in the day to heat the near-surface air. The higher heat capacity surfaces that have absorbed the heat continue to re-radiate heat well after (direct and diffuse) short-wave solar radiation has declined in intensity, but the effect diminishes as heat is lost from both the ground and the near-surface air overnight at similar rates [11]. The CUHI effect, therefore, has little impact on peak daytime near-surface air temperatures in summer conditions (which typically occur in the mid-afternoon) and is largest in the evenings. Importantly, the CUHI also affects air-conditioning loads in buildings (as average daily air temperatures are higher in affected locations) [16].

Many UHI articles are based on land surface temperature, which measures the surface urban heat island (SUHI) effect [15,19,20,21]. As noted above, SUHI generally reflects CUHI effects but not quantitatively. When exposed to radiation, built-form elements that absorb more energy than vegetated surfaces develop much higher land surface temperatures. Measurements of surface and air temperatures in Phoenix, Arizona (also reported by Mills and Stewart [4]), show peak land surface temperatures in the desert and over asphalt pavement of around 60 °C in the middle of the day, compared to 45 °C on a suburban irrigated grass surface (ΔT_s ≈ 15 °C). (Δ denotes Delta or differential). However, at the same time of day, the near-surface air temperatures of all three surfaces were similar (38–40 °C) (Figure 3).

Other studies involving the measurement of air temperature differences in urban areas also illustrate that ΔT_a is much less than ΔT_s. Eliasson [22] found that overnight air temperature differentials (ΔT_a) were seldom more than 2 °C. Spronken-Smith and Oke reported average August time ΔT_a values in Vancouver parks of 1.9 °C (night) and 2.2 °C (day) and in Sacramento parks of 2.4–3.3 °C (night) and 1.2–2.4 °C (day). In an extensive study in Tokyo, Sugawara [23] reported measured ΔT_a values of 0.5–1.8 °C (composite of 72 fair-weather days in August).

Many studies on the PCI utilise remotely sensed land surface temperature data [15,19,20,21]. Many of these sources use the land surface temperature data to assess the apparent cooling effect of parks on adjacent urban areas, often referred to as the ‘park cooling distance’ (PCD) [24]. The data analysis methodologies of these studies vary, but they all rely on Landsat Thermal Infrared Sensor (TIRS) data. The spatial resolution of the raw data from Landsat 8 is 100 m, and the data is ‘resampled’ to 30 m—a process known as ‘Cubic convolution’, which averages the 16 closest 100 m data points to the nominal 30 m grid [25]. This process is generally acceptable for delineating large-scale temperature differences in the landscape.

However, as is clear from ground-based land surface temperature data, there is a sharp delineation in T_s between park surfaces such as grass and adjacent roads under exposure to short-wave solar radiation (as identified in thermal imaging, Spronken-Smith and Oke [11]). If land surface temperature data were accurate to 5 m, that would be evident in the images. However, the resampling ‘averages out’ the differential between the actual land surface temperature in the park and the actual land surface temperature in the surroundings and will vary according to where the original image is taken relative to the park boundary. Accordingly, land surface temperature sources are not a reliable guide to assessing small parks’ PCI or the ‘park cooling distance’ in general.

However, measurements of air temperatures [11,13,23] support the notion that the lower temperatures in parks influence temperatures in adjacent urban areas. Vanos et al. found that cooler air ‘extends outside the park in the direction of prevailing winds’ [26]. This process is known as advection (which is analogous to convection). Sugawara et al. [27] similarly found that ‘turbulent mixing transfers colder air from the park to urban areas in its vicinity’ with a greater distance in the downwind direction; however, cooling also occurred in calm conditions. Sugawara refers to this phenomenon as a ‘gravity current driven by the pressure gradient between the colder park and the warmer town’ [23]. Oke [28] hypothesised a similar mechanism, with cooler, denser air from the park flowing outwards in a circulation pattern that brings ascending warmer air to converge at higher elevations above the park. This mechanism is analogous to the mechanism by which daytime sea breezes occur at much larger scales in coastal regions. The process would be disrupted at the park/urban scale by any significant ground-level wind speeds.

A focus on surface or air temperatures neglects the full effects of summer conditions on human thermal comfort. The human energy balance that controls thermal comfort (for a particular level of physical activity) depends on the combined effect of air temperature, humidity, air velocity, and, most importantly for outdoor environments, the mean radiant temperature. In outdoor settings, the mean radiant temperature is a function of the intensity of solar radiation, including its effects on re-radiation from adjacent surfaces. This is the major factor influencing heat stress levels for people directly exposed to the sun in locations with high direct normal radiation in peak summer conditions, such as Perth. In this study, the Physiological Equivalent Temperature (PET) is used as the measure of thermal comfort, which includes the full range of variables of interest.

1.3. Research Questions

In order to identify the relevance of this previous work for evolving climate conditions in Perth, Australia, and their implications for urban and landscape design, in this article, we seek to resolve the following questions:

• What are the factors that control the urban microclimate of Perth within the existing urban form?
• What are the variables in Perth that are relevant to urban design and human thermal comfort?
• To what degree do parks in Perth provide cooling, either internally or in surrounding areas, in hot summer conditions?

To answer these questions, this study compares the results of a commonly used method to evaluate the PCI (satellite-derived land surface data) with field measurements of surface, air temperature, and relative humidity, and the results of microclimate simulation software (ENVI-met).

2. Materials and Methods

This study examined the likely PCI effects in park case studies (n = 3) in Perth, Western Australia, a city of 2.4 million people stretching some 125 km from north to south on a coastal plain. Perth has a hot-summer Mediterranean climate (Köppen climate classification Csa) and receives moderate, highly seasonal rainfall, predominantly during winter. Summers are generally hot and dry, lasting from December to March, and can cause extreme heat events that pose risks to exposed populations’ health. Research has indicated that Perth has experienced more frequent, longer, and hotter heatwaves in the twenty-first century [29].

The case study parks across Perth (Figure 4) represent different prevailing theories for POS provision and neighbourhood design in the surrounding area. In brief, eligible park case studies were situated in predominantly residential areas. The case studies are discussed in relation to their respective landscape and urban design theories below.

2.1. The Case Study Parks

2.1.1. Garry Meinck Park, Butler

The suburb of Butler in Perth’s north has a population density of 2450 people per km² [30] and reflects the current ‘Liveable Neighbourhoods’ design code governing greenfield development in Western Australia [31]. Following New Urbanist thinking, this policy calls for a compact suburban layout comprising small lots (averaging 300 m²) with high site coverage by houses at a density of about 30 dwellings per hectare, interspersed with smaller yet accessible neighbourhood parks [31]. Due to high site coverage by buildings, small lots and juvenile tree specimens, these case studies have minimal urban canopy cover on private land. Comparatively narrow road sections and the constraints presented by service infrastructure have also resulted in minimal tree canopy cover in streets.

Within Butler, Garry Meinck Park (Figure 5) is designated as a ‘Neighbourhood Park’ within the Western Australian classification framework for public open space [32]. The landscaping of these neighbourhood parks consists of areas of irrigated and non-irrigated lawns, mulch, and paving with some fringing trees. Garry Meinck Park is around 1.3 ha, with a canopy cover of less than 5% and a grass cover of around 20%. Typical tree types in Garry Meinck Park included native species such as Eucalyptus gomphocephala, Agonis flexuosa, Corymbia calophylla, and Melaleuca quinquenervia.

2.1.2. Macnaughton Park, Kinross

The 1990s suburb of Kinross has a population density of 2395 people per km² [30]. Kinross reflects planning for semi-compact suburbs, comprising housing at a density of 25 dwellings per hectare, with an average lot size of 350 m² and high site building coverage, resulting in minimal tree canopy cover on private lots. Due to comparatively narrow road sections and the constraints presented by service infrastructure, the streets in the case study also have minimal tree cover. Macnaughton Park (Figure 6) is a designated ‘Neighbourhood Park’ within the Western Australian classification framework for public open space [32]. The case study park has, in part, resulted from government policy aimed at protecting bushland in the Perth region [33]. In addition to retained bushland to the west and south, and stands of remnant trees to the east, the 4.6 ha park also contains an irrigated oval and clubrooms. Typical trees are native, including Agonis flexuosa, Xanthorrhoea preissii, Allocasuarina spp. and Eucalyptus, with a canopy cover of 13.8% and a grass cover of 33%.

2.1.3. Hyde Park, Highgate

The inner mixed zone of Highgate has a population density of 4008 people per km² [30] and comprises late nineteenth-century inner-city suburbs that fringe the Perth Central Business District. Various housing types include late nineteenth and early twentieth-century workers’ cottages, grand Federation houses and significant apartment developments from the 1960s and 1970s. Tree canopy cover on private lots is generally low due to high building cover, but it is reasonable on the streets. This urban form surrounds Hyde Park, a product of the public parks movement, which saw municipal parks (such as Olmsted’s and Vaux’s design of New York’s Central Park) being established in the mid-nineteenth century in major cities as a way of improving living conditions in overcrowded and quickly growing industrial cities [34]. Hyde Park reflects such elaborately landscaped parks of the late nineteenth century and comprises walking paths shaded by avenues of trees encircling a central lake (and once a wetland) (Figure 7). The 16 ha park has a canopy cover of 59.5% and a grass cover of 18.7%. Trees are typically exotic and include Platanus acerifolia, Ficus macrophylla, Washingtonia filifera and Jacaranda mimosifolia. Hyde Park is a designated ‘Regional Park’ within the Western Australian classification framework for public open space [32].

2.2. Microclimatic Case Study Data

Microclimate data for the case study sites were collected and analysed from three contrasting sources: remote-sensed land surface temperature data [35], on-site measurements, and (for Hyde Park) ENVI-met microclimatic modelling [36] (

Table 1. Microclimate data sources utilised.

Microclimate Data Source	Garry Meinck Park, Butler	Macnaughton Park, Kinross	Hyde Park, Highgate
Measured land surface temperature	✓	✓	✓
Measured air temperature	✓	✓	✓
Measured relative humidity	✓	✓	✓
Landsat 8 land surface temperatures	✓	✓	✓
ENVI-met land surface temperatures			✓
ENVI-met air temperatures			✓
ENVI-met Physiologically Equivalent Temperatures (PET)			✓

). The ENVI-met model simulated an average summer day and provided outputs for land surface temperature, air temperature, and the Physiological Equivalent Temperature (PET) thermal comfort index. As PET includes radiation, a factor not directly captured by air or surface temperature, it offers a more comprehensive measure of human thermal comfort.

2.2.1. Remote-Sensed Land Surface Temperature Data

Land surface temperature maps were generated for all Australian cities using Landsat 8 thermal infrared sensor (TIRS) band 10 data over the summer of 2018–2019 [35]. Land surface temperatures were collected for Perth on the 14 December 2018 and the 21 January 2019; both dates were cloud-free and had maximum temperatures of 39.1 °C and 29.8 °C, respectively [37].

2.2.2. Measurements of Land Surface Temperature, Air Temperature, and Relative Humidity

Land surface temperature, air temperature, and relative humidity were monitored inside and adjacent to each park. Measurements were taken every 25 m, forming a transect that extended approximately 200 m on either side of the park. Each set of measurements was taken on approximately average summer days, with an overnight minimum below 20 °C, a maximum around 30 °C, and clear, sunny conditions. Three sets of measurements were taken: at 7:00–8:00 am, 2:00–3:00 pm, and 7:00–8:00 pm (hereby referred to as morning, afternoon, and evening measurements).

Instruments were mounted on a trolley, which was used to ensure that measurements could be completed quickly, minimising the effects of changing conditions as the day progressed. Temperature and relative humidity were measured with a Sauermann Si-HH3 thermo-hygrometer manufactured in Montpon-Ménestérol, France, while land surface temperatures were measured with a Topdon TC-003 thermal camera, manufactured in Shenzhen, China. The thermo-hygrometer was shielded from solar radiation to ensure direct sunlight on the probe did not impact air temperature readings.

Temperature and humidity values from Bureau of Meteorology (BOM) measurement sites in Perth were used to adjust for changes in the macro weather as each collection event progressed.

2.2.3. ENVI-Met Microclimate Modelling of Land Surface Temperature, Air Temperature and Physiological Equivalent Temperature

Microclimate modelling was conducted for Hyde Park using the ENVI-met software version 5.0 [36]. ENVI-met is a three-dimensional software model for simulating urban environments that incorporates computational fluid mechanics, thermodynamics, and atmospheric physics. All buildings, surfaces, and vegetation were digitised in ArcMap 10.8.2. This process involved identifying the material properties, comparing them with the software’s default options, and creating any required custom materials. The assigned IDs for these materials were then used during the digitisation process in ArcMap. Vegetation modelling included the digitisation of grass, shrubs, and trees. Since the default software library does not include trees from the Southern Hemisphere, these were classified and created using the ENVI-met Albero programme based on specific attributes such as height, crown height and width, trunk height, tree calendar, and leaf type and positioning, resulting in 30 different tree types (

Table 2. ENVI-met modelling and simulation input parameters.

Domain Size	500 × 500 × 50 (Height)
Grid spatial resolution:	2 m
Meteorology:	Full forcing using EPW file + enabled IVS radiation.
Modelling features:	Buildings: (default building) moderate insulation wall, between and below buildings: Pavement (concrete) Soils: Default unsealed soil (Sandy soil), sandy soil placed below simple and 3D vegetation Surfaces: Asphalt road, wood chip mulch, light concrete pavement Vegetation (2D): non-irrigated grass: 2 cm Trees (3D): Simplified to 11 different trees in different sizes (in total 30)
Duration:	14 h

). The final model was exported using QGIS version 3.34.8.

The weather conditions set as inputs to the ENVI-met models were ‘epw’ files produced by the Commonwealth Scientific and Industrial Research Organisation (CSIRO). CSIRO typical meteorological year (TMY) files are based on historical weather data drawn from the years 1990 to 2015. The simulation runs for 14 h on an ‘average’ summer day, selected to statistically best match the combination of hourly average dry bulb temperature and relative humidity. The selected day is 9 January 1998.

Surface and air temperature maps were generated using the ENVI-met Leonardo programme to compare the microclimatic conditions within parks and their surrounding environment. The human thermal comfort index was calculated using the ENVI-met BioMet programme, a post-processing tool that calculates human thermal comfort indices from ENVI-met output files. PET reflects ‘…the air temperature at which, in a typical indoor setting (without wind and solar radiation), the heat budget of the human body is balanced with the same core and skin temperature as under the complex outdoor conditions to be assessed’ [38]. The PET index accounts for the combined effect on human physiology arising from air temperature, mean radiant temperature, wind speed, and vapour pressure (i.e., a measure of humidity), depending on clothing and metabolic rate (Figure 8) [38]. ENVI-met produces calculated values of PET from a refined set of equations derived from Walther and Goestchel [39].

3. Results

Microclimate data for the case study sites are set out below, including on-site measurement of land surface temperature, air temperature, relative humidity (%), remotely sensed land surface temperature data [35], and ENVI-met microclimatic modelling for the final Hyde Park case study [36].

3.1. Weather Data

The field data was collected during summer days that were broadly representative of average conditions.

Table 3. Weather conditions during field data collection.

		Average Summer Day (9 January 1998)	Garry Meinck Park (28 February 2025)	MacNaughton Park (14 February 2025)	Hyde Park (1 March 2025)
Dry bulb temperature (°C)	Morning	18.9	18.2	21.3	19.1
	Afternoon	30.3	25.2	23.2	29.6
	Evening	27.1	25.2	22.3	23.1
Relative humidity	Morning	78	63	66	67
	Afternoon	40	49	63	42
	Evening	50	78	83	65

compares the air temperature and relative humidity data from the nearest Bureau of Meteorology weather station with the selected average summer day during the data collection period. Morning temperatures were broadly representative of average conditions, while afternoon temperatures were marginally cooler than average for the Garry Meinck and MacNaughton Park data collection days. Wind speeds in all cases represented average Perth summer days with east–southeast breezes in the morning, which swing to coastal sea breeze from the south west in the afternoons.

3.2. Microclimate Data for Garry Meinck Park, Butler

The figure below shows the measured land surface temperature, air temperature and relative humidity for the Garry Meinck Park case study (Figure 9). The land surface temperatures of the irrigated turf within the park are substantially cooler in the middle of the day, and to a lesser extent in the morning and evening. Air temperatures are generally consistent between the park and the urban catchment. Relative humidity levels are consistent between the park and the urban context in the morning and evening, with a marked increase in humidity in the park (and to the south of the park) in the middle of the day due to the arrival of a sea breeze during the measurement process.

The figure below compares the remotely sensed and measured land surface temperatures [35] for the Garry Meinck Park case study (Figure 10). The land surface temperature is approximately the average of the daytime measured temperatures in Figure 9. A negligible land surface temperature difference is recorded between the park and the urban catchment in the remotely sensed data.

3.3. Microclimate Data for Macnaughton Park, Kinross

The figure below shows the measured land surface temperature, air temperature and relative humidity for the MacNaughton Park case study (Figure 11). The irrigated turf of the oval (left-hand green box on graph) provides substantial cooling of land surface temperatures in the middle of the day; however, this cooling effect is substantially lower in the morning and evening. Notably, the area of non-irrigated bushland (right-hand green box on the graph) is hotter in the morning and midday than the surrounding urban context. The irrigated oval area of the park provides some cooling of air temperatures in the middle of the day, and none in the evening. Notably, the air temperatures in the irrigated section of the park are slightly warmer in the early morning relative to the urban context. The non-irrigated bushland area of the park experiences warmer air temperatures in the afternoon compared to the surrounding urban context. The irrigated oval in the park results in a substantial increase in relative humidity during the middle of the day; however, the non-irrigated area is similar to the adjacent streets.

Figure 12 compares the remotely sensed and measured land surface temperature data for the Macnaughton Park case study. Again, the land surface temperature is approximately the average of the daytime measured temperatures in Figure 11. A minor reduction in land surface temperature is recorded on the irrigated oval.

3.4. Microclimate Data for Hyde Park, Highgate

The figure below shows the measured land surface temperature, air temperature and relative humidity for the Hyde Park case study (Figure 13). In terms of land surface temperatures, the park provides substantial cooling during the middle of the day; however, it has little effect on cooling in the morning or evening. In relation to air temperatures, the park provides the greatest cooling in the afternoon and to a lesser extent in the early morning. In the evening, the park does not provide cooling in relation to air temperatures. Levels of relative humidity are elevated in the park, particularly during the midday hours, due to the lake. Evaporative cooling from the lake is also likely to contribute to the lower air temperatures.

Remote-sensed land surface temperatures for the Hyde Park case study reveal lower temperatures in the park (Figure 14), but a considerably smaller temperature difference between the park and urban catchment compared to those measured during the middle of the day (Figure 13). Again, the remotely sensed data generally reflects the average of the measured data.

The following figures compare microclimatic data from on-site measurement, remote sensing [35], and ENVI-met modelling [36].

Figure 15 compares the remotely sensed land surface temperature data, the measured land surface temperatures, and the ENVI-met modelled land surface temperature data for an average summer day in the Hyde Park case study. The results are similar within the park. However, ΔT_s (between the unshaded areas of the park and street) is less than that in the measured land surface temperature data. The measured land surface temperatures show greater variability but generally correlate with those modelled in ENVI-met.

Figure 16a–c presents the Hyde Park ENVI-met map of air temperature, incorporating measured data throughout the day. The figure shows that ENVI-met projects cool air to the north of the park along a wind path through advection (the wind at 07:00 is a light breeze from the south at 2.5 m/s), also noting that warm air enters the park from the south. However, the air is only fractionally cooler (less than 1 °C), and the effect diminishes during the day. By evening, there is no detectable effect. These results confirm that while the PCI is a real phenomenon, the magnitude in terms of air temperatures is minor, and the PCD effect is limited. Figure 16d–f compares the ENVI-met simulated air temperature.

Figure 17 sets out the ENVI-met PET outputs for Hyde Park. PET levels are generally consistent between the park and the urban context, except under tree canopies.

4. Discussion

4.1. Park Cool Island Effects

The data presented above indicate that while the case study park surfaces (particularly grass) have cooler land surface temperatures in the middle of the day than surrounding areas (positive ΔT_s of 10–15 °C) when direct solar radiation is most impactful, the effect diminishes by evening, and by morning, it is not evident. ΔT_s values in the evening are similar to those reported by Spronken-Smith for a park in Vancouver [11].

Regarding air temperatures, the afternoon ΔT_a is significantly smaller than ΔT_s, even in the afternoon (approximately 2 °C), and is largely absent in the evenings and mornings. At MacNaughton Park, morning T_a is slightly higher than the park’s surroundings, as is afternoon T_a on the sandy soil surface of the park.

Spronken-Smith [11] reported a ΔT_a of around 2 °C at both daytime and nighttime for 10 parks in Vancouver, and 1.2–2.4 °C (daytime) and 2.4–3.3 °C (nighttime) for 10 parks in Sacramento, with both sets of data from August. Sugawara [23] reported ΔT_a values ranging from 0.5 to 1.8 °C (nighttime) and 1 to 1.6 °C (middle of the day) over 72 days in summer conditions in Tokyo. Those results are very similar to those reported here for daytime ΔT_a but vary for other times.

The Hyde Park ENVI-met data show that the afternoon T_a in most of the park is similar to that of the adjacent streetscape, and in some exposed areas, it is even higher. Afternoon T_a is around 2–3 °C lower to the north of the lake, where the water leads to some evaporative cooling. Also, that morning, T_a was relatively warmer in this location due to the retained heat in the lake.

Other studies have identified that the cool air in parks produces lower air temperatures in adjacent streetscapes. Spronken-Smith [11] reported a PCD of approximately 75 m in the Vancouver study, and Sugawara [23] reported PCDs ranging from 31 to 120 m in the Tokyo study. That effect is not evident in the datasets reported here for any case study park at any time of day, irrespective of their size. Indeed, Figure 16 shows that ENVI-met projects cool air to the north of the park along a wind path through advection (the wind at 07:00 is a light breeze from the south at 2.5 m/s), also noting that warm air enters the park from the south. However, the air is only fractionally cooler (less than 1 °C), and the effect diminishes during the day. By evening, there is no detectable effect. These results confirm that while the PCI and PCD are real phenomena, their magnitude in terms of air temperatures identified in this study is minor.

4.2. Correlations Between Remotely Sensed, Measured, and Modelled Climate Data

4.2.1. Remote-Sensed and Measured Land Surface Temperatures

Comparisons of remotely sensed land surface temperatures (which reflect afternoon conditions) with measured data (Figure 15) illustrate that remotely sensed land surface temperatures tend to slightly underestimate streetscape land surface temperatures (T_s) and slightly overestimate T_s within parks. Remote-sensed land surface temperatures also do not accurately delineate park boundaries for the smaller parks. Cooler land surface temperatures evident through measurement in Garry Meinck Park, which is only around 75 m in width, are not reflected in the remotely sensed land surface temperature data (Figure 10).

These observations are consistent with the fact that the raw data from which the remotely sensed land surface temperature maps are produced have a resolution of 100 m and hence cannot identify small-scale land surface temperature differentials. For example, remotely sensed land surface temperatures underestimate generic urban form T_s values due to the averaging of higher-temperature impervious streetscape surfaces and lower-temperature vegetated spaces in verges and yards.

Therefore, the use of remotely sensed land surface temperature is of limited value in urban studies at the neighbourhood scale due to the limitations of the raw TIRS data’s resolution [10]. The resampling process used to produce 30 m resolution images inevitably averages actual T_s values, resulting in underestimates in general urban form and overestimates in (all but large) parks.

4.2.2. ENVI-Met Surface and Air Temperatures

The Hyde Park ENVI-met model input is based on an ‘epw’ file representing average summer conditions, so macro-scale air temperatures are not identical to those measured from the nearest Bureau of Meteorology weather station during data collection. However, they are very similar during the day (see

Table 4. T_a (°C) measured and ENVI-met input weather data for the Hyde Park case study (averaged for the hours in which data were collected).

	Measured	ENVI-Met Weather File
Morning	19.1	18.9
Afternoon	29.3	30.3
Evening	23.1	27.1

). Table 3 indicates that the measured T_a is similar to the ENVI-met weather file in the mornings and afternoons, but slightly higher in the evenings.

The Hyde Park ENVI-met model shows reasonable agreement between modelled and measured data in respect of both air and surface temperature. Outside the park perimeter, the modelled T_s values (Figure 15) are somewhat lower than the measured data. There is slightly more variance within the park, but this can be attributed to the model surface being grass rather than the actual surface, which is asphalt paving. The ENVI-met and measured Ts results within the park are similar in the afternoon, illustrating the overriding impact of surfaces shaded from direct solar radiation.

The air temperatures (T_a) (Figure 16d–f) are similar between the datasets. The ENVI-met results are slightly lower in the morning, almost identical in the afternoon and slightly higher in the evening. This situation primarily reflects the macro-scale differences outlined in Table 3.

The differences between parks and surroundings with respect to surface and air temperature results are broadly similar to the findings of other researchers, including those referenced in the introduction to this article [22,23]. The key question is their relevance to human thermal comfort.

4.2.3. Thermal Comfort Modelling

As described, the Physiological Equivalent Temperature (PET) index accounts for the combined effect on human physiology arising from air temperature, mean radiant temperature, wind speed, and vapour pressure (i.e., a measure of humidity) [38]. Mean radiant temperature (MRT) is a summation of radiative effects from both direct and diffuse (short-wave) and indirect (long-wave) radiation. Direct solar radiation is by far the largest contribution to MRT in summer daytime conditions in Perth, and hence, PET is dominated by this exposure. This situation is starkly evident from the afternoon PET results in Figure 17, which show strong to extreme heat stress in the park, except in the shade of the trees. Except in those locations, the PET is no different from the surrounding urban areas. Exposure to direct solar radiation completely overwhelms any minor reduction in air temperature in the park. The thermal comfort benefits offered by the park are evident in the evenings, where the PET is lower than in the surrounding area due to lower T_a in the absence of solar exposure.

This analysis raises questions about the relevance of T_a and T_s in relation to the urban microclimate and their impact on outdoor human thermal comfort. As illustrated here and in other studies, the temperature of urban surfaces (T_s) is not a good indicator of near-surface air temperature (T_a), nor is it a good indicator of human thermal comfort, which reflects exposure to direct and indirect radiation, humidity and wind and at the hottest time of day in summer is dominated by direct exposure to solar radiation. These results confirm that while the PCI is a real phenomenon, the magnitude is comparatively minor (ΔT_a ≈ 2 °C) and is of limited relevance to human thermal perception in solar-exposed locations during average summer daytime conditions in climates such as that of Perth. By far, the major benefit offered by parks is the shade provided by trees, as is evident from the PET maps of Hyde Park (Figure 17).

Similarly, PCD is of little relevance to the planning and design of urban areas. Although it was not evident in the measurement and modelling reported here, even when identified in other studies, the extent of influence on Ta was minor and limited to approximately 120 m.

4.3. Research Findings

The study identifies that it is the direct and indirect effects of solar radiation in the form of mean radiant temperature (MRT) that have the major influence on the urban microclimate in Perth summer daytime conditions (research question 1). This variable has a greater impact on human thermal comfort (as measured by PET) than air temperature in solar-exposed locations. Surface temperature influences MRT but in itself is not a good indicator of thermal comfort (research question 2). In relation to measured air temperatures, only the Hyde Park case provides some cooling in the afternoon and, to a lesser extent, in the early morning. While PCI is briefly observable in this study in the Hyde Park simulation, it is not measurable in the hottest part of the day or in the data collected from field measurements (research question 3).

Regardless, urban parks are very important elements of the urban fabric for a multitude of reasons [40]. As this study of different park typologies in Perth, Australia, illustrates, it is largely the continuous shade provided by trees that provides any benefit, as illustrated by the corridor of trees surrounding Hyde Park Lake (see Figure 17b). Therefore, the adaptation of parks (and urban areas more generally) to ameliorate daytime summer conditions requires the creation of a continuous tree canopy, ideally on parks, streetscapes and private lots in an interconnected network. One of the dangers of a singular focus on Park Cool Island effects is that urban planners and designers discount the importance of outdoor thermal comfort in streetscapes and private lots, as well as the need for connected urban greening more generally.

4.4. Limitations

The findings of this study, focusing on Perth, Western Australia, are of most relevance to locations with similar Mediterranean climates and high levels of solar irradiation in summer. Other locations will exhibit a different balance between air temperatures, humidity and mean radiant temperatures and hence result in different levels of thermal comfort in park environments.

While the ENVI-met modelling is only reported for Hyde Park in this paper, it has been used by the authors to model several other parks in Perth and other locations in Western Australia, which will be reported in a future article (presently under review). The results from those studies mirror the findings reported here. Hyde Park exhibited the largest PCI effect and was chosen for further analysis in ENVI-met to understand the dynamics of the effect in more detail, given that the other case studies had negligible PCI effects.

Finally, the measurement of PCI effects in this study is based on three time points. Future research in this area could utilise continuous diurnal profiles to illustrate the temporal evolution of the PCI effect.

5. Conclusions

The major factor that gives rise to the microclimatic conditions in urban areas is the modification of the macroclimatic impacts of solar radiation, air temperatures, and wind speeds and direction by buildings, artificial surfaces, and vegetation. The temperature of surfaces and near-surface air in any particular location results from the complex interaction between these factors. While T_a is important for human thermal comfort when solar radiation is absent, it is much less influential at the hottest times of summer days in warm to hot climates such as Perth. Accordingly, variables such as the PET are much more relevant to estimating the impact of urban design and parks on human thermal comfort in summer conditions. Given the importance of solar radiation to human thermal comfort, the adaptation of parks (and urban areas more generally) to ameliorate daytime summer conditions requires the creation of a continuous tree canopy, ideally over parks, streetscapes, and private lots in an interconnected network.