Optimizing Urban Greenery for Climate Resilience: A Case Study in Perth, Australia

Abstract

Urban vegetation plays a pivotal role in mitigating the Urban Heat Island (UHI) effect and enhancing ecological resilience amid accelerating global urbanization. This study investigates the spatiotemporal dynamics of vegetation coverage and its interplay with climatic factors and surface thermal patterns in Perth, Australia, from 2014 to 2023, leveraging multi-source remote sensing data, geostatistical modeling, and spatial analysis. Utilizing Landsat-derived Normalized Difference Vegetation Index (NDVI), Land Surface Temperature (LST), and Land Use/Land Cover (LULC) datasets, combined with meteorological statistics, the research quantifies vegetation trends, evaluates seasonal and annual climate correlations, and stratifies UHI intensity zones. Key findings reveal the following: (1) Perth’s vegetation cover has decreased over the past decade, and LST has increased, with a negative correlation between the two. (2) NDVI demonstrated a strong negative correlation with annual maximum temperature (r = −0.754) and a positive correlation with precipitation (r = 0.779). (3) Seasonal analysis of NDVI-LST relationships showed intensified cooling effects in summer (r = −0.527) compared to winter (r = −0.180), aligning with evapotranspiration dynamics in Mediterranean climates. (4) Spatial stratification of LST identified “low-temperature green islands” in forested regions, contrasting sharply with high-temperature zones in built-up areas. This study suggests that vegetation optimization—particularly preserving urban forests and integrating green infrastructure—can effectively mitigate UHI impacts, thus reducing surface temperatures. In particular, it shows that urban greenery is a more significant factor towards lowering UHI than urban density. This research advances the understanding of how vegetation optimization can mitigate thermal stress in growing urbanization and provides quantitative evidence for climate-adaptive urban planning.

1. Introduction

1.1. A Review of Current Literature on UHI, NDVI and LST

The Urban Heat Island (UHI) effect, characterized by elevated temperatures in built environments compared to surrounding rural areas, has emerged as a critical environmental challenge affecting over 54% of the global population in cities [1,2]. Many studies have been published over the past few decades, covering a variety of backgrounds and greatly improving our understanding of UHI dynamics. The main causes of the UHI effect are the large amount of heat generated by urban buildings, the use and re-radiation of solar energy, artificial heat sources and reducing greenery cover. Other natural factors, such as weather and geography, have also been shown to significantly impact the UHI effect [3].

From the perspective of the urban landscape, urban greenery plays a crucial role in regulating both local and global climate by influencing climatic factors, which can mitigate the UHI effect [4]. Bowler et al. [5] conducted a systematic review and concluded that urban greenery through parks, green roofs, and other forms of vegetation can effectively cool urban environments, especially during heat waves. Based on this, Zölch et al. [6] extended this by evaluating heat mitigation strategies that were more localized, such as prioritizing tree planting in identified urban hot spots, increasing green facades and green roofs on urban buildings, etc. They demonstrated that strategically placed greenery in cities can significantly reduce air and land surface temperatures, providing climate protection benefits to urban areas. These studies underscore the broader benefits of urban greening towards climate protection. Vegetation cover contributes significantly to the management of the Earth’s climate, especially temperature and precipitation patterns [7]. Vegetation releases water vapor during transpiration, which increases air humidity and leads to cloud formation leading to increased precipitation [8]. Research has shown that vegetation greenery can influence regional climatic conditions by affecting wind and moisture, leading to different rainfall distribution and intensity [9]. Bowler et al. [5] highlighted the importance of green infrastructure for mitigating urban heat waves, suggesting that urban infrastructure projects can help offset some of the warming associated with dense urban environments. All of these studies emphasize the potential of urban greening to reduce local warming, providing an ecological solution for cities to meet rising temperatures due to climate change.

Vegetation’s cooling capacity, quantified through the Normalized Difference Vegetation Index (NDVI), has been extensively documented as a key UHI moderator [10]. Comparative analyses reveal that NDVI generally outweighs precipitation and background temperature in UHI mitigation [11]. NDVI-LST inverse correlations have been reported across diverse urban settings [12]. The study by Dewan and Corner [13] revealed the long-term impact of land cover changes on the LST and UHI in Dhaka from 1990 to 2011. Through correlation and regression techniques, research exploring the relationship between LST and biophysical parameters (e.g., NDVI) indicated that persistent loss of green space results in increased urban temperatures. However, vegetation’s cooling capacity exhibits diurnal and seasonal variability. Buyantuyev and Wu [11] compared diurnal and seasonal data and confirmed the important role of vegetation and land type cover in explaining the spatiotemporal variability of temperature in Pheonix. Experimental data show that LST exhibited strong positive correlations with impervious surfaces (e.g., buildings, roads; r > 0.7) and negative correlations with vegetation (NDVI) and water bodies (r < −0.6). As global climate change continues to intensify, understanding how vegetation (both in natural and urban environments) affects local and regional climate is essential for mitigation and adaptation strategies.

LST derived from thermal infrared sensors has become the gold standard for UHI quantification, overcoming traditional air temperature measurements’ spatial limitations [14]. Validation studies show strong agreement between MODIS LST and in situ measurements in urban settings [15]. However, challenges persist in Mediterranean regions where rapid daytime heating and complex surface emissivity may introduce errors [16]. Different climate regions may produce different LST dynamics due to seasonal differences, affecting the accuracy of judgment data.

1.2. Research Questions and Objectives

A full understanding of the effect of urban development and greenery on UHI taking into account urban development, land use/landcover, and geography lies outside the scope of this paper. The intention of this paper is to process satellite data, using ArcGIS Pro 3.0.2, to study the relationship and causes of variations in NVDI and UHI from an urban design perspective—specifically, looking at the effects of LULC, greenery cover, climate and LST:

• How did urban development impact urban greenery (measured in terms of NDVI) and surface temperature (LST) over a given period of time (2014 to 2023)?
• Is NDVI more or less significant than urban development in influencing UHI over these 10 years?
• How does climate affect NDVI and is there a trend over the study period in terms of climate change?
• Is the UHI effect dependent on seasonal variations in temperature and rainfall and if so, in what way?
• How do land use and landcover (LULC) affect UHI?

In this study, we selected Perth, a city in Australia with a Mediterranean climate, and developed the research in four stages:

• Investigate the trend of NDVI and LST in Perth over a 10-year period (2014–2023)
• Study the relationship between NDVI and climate, in particular temperature and precipitation.
• Understand the relationship between NDVI and LST across the seasons, in particular between summer and winter.
• Study the effect of LULC on NDVI and LST

1.3. Study Area—Perth

As shown in Figure 1, Perth is the capital city of Western Australia, with most of Perth’s metropolitan area on the Swan Coastal Plain between the Indian Ocean and the Darling Scarp. It characterized by a Mediterranean climate with hot, dry summers and mild, wet winters. Perth is one of the world’s most liveable cities and consistently ranks in the top 10 of the Economist Intelligence Unit’s World’s Most Liveable Cities survey [17].

Perth has experienced significant urban expansion over the past two decades, driven by population growth and suburban sprawl, with its population increasing by approximately 30% between 2011 and 2021, becomes the fourth most populous city in Australia [18]. This growth has transformed natural landscapes into urbanized areas, reducing vegetation cover and exacerbating environmental challenges such as UHI effect—a critical concern given the region’s vulnerability to climate change [19].

1.4. Data and Pre-Processing

The maximum temperature, average temperature, and cumulative precipitation data are derived from (1) the Global 0.5° climate dataset released by the Climatic Research Unit (University of East Anglia) and Met Office and (2) the Global High-resolution Climate Dataset released by WorldClim. Climate data for some selected years is referenced from the Australian Bureau of Meteorology. The spatial resolution is 30 s (~1 km²). The DEM data are derived from the Copernicus GLO-30 Digital Elevation Model (COP30) Dataset, which is provided by OpenTopography. The NDVI data are derived from the Landsat-8 dataset provided by USGS, with the metadata filters set (cloud cover under 5%). The date is chosen between June and July from the years 2014 to 2023. LST values through a series of calculations and original data are derived from ESRI Sentinel-2 10-Meter Land Use/Land Cover.

In ArcGIS-Pro 3.0.2, all the data are pre-processed by projection transformation, raster calculation, zonal statistics, fishnet, and reclassification to generate images and exported data.

2. Methodology

2.1. Stage 1: Trend of NDVI and LST in Perth over a 10-Year Period (2014–2023)

Building on the foundational question of how urbanization reshapes urban ecosystems, this stage quantifies the decadal trajectories of vegetation health (NDVI) and thermal conditions (LST) in Perth.

2.1.1. NDVI

Monitoring the growth of vegetation coverage has always been an important issue. One widely recognized metric is the NDVI. Rouse et al. [20] proposed a simple band conversion: near-infrared (NIR) radiation minus red radiation, divided by NIR radiation plus red radiation. This formula generates a synthetic image that allows researchers to analyze vegetation coverage.

Vegetation strongly reflects NIR light while absorbing RED light for photosynthesis, and NDVI can take advantage of this contrast to measure the “greenness” or biomass of an area [21]. NDVI is popular and widely used in land studies because it can quickly characterize vegetation and simplify complex data. Since the 1970s, the interest in NDVI has increased dramatically. Earth observation satellites can now generate this index at various spatial and temporal scales [22]. NDVI values are often categorized into different ranges to indicate different degrees of vegetation health, density, and land cover type. NDVI values typically range from −1 to +1, with negative values indicating non-vegetated surfaces, such as water. NDVI values approaching +1 indicate areas with dense and healthy vegetation. NDVI values closer to +1 reflect higher photosynthetic activity and therefore indicate a lush vegetative coverage [23]. In recent years, many scholars have used satellite datasets to monitor NDVI and obtained a series of factors affecting NDVI values [22]. In the Australian region, NDVI has been applied to track vegetation changes, monitor land degradation, and study the impact of drought. It has been observed that NDVI values often correlate with temperature and precipitation, providing valuable insights into vegetation dynamics in response to climate factors [24].

In recent years, several studies have demonstrated that there is a relationship between the NDVI and leaf area index (LAI), which can be used to estimate the LAI from NDVI [25,26]. It quantifies the leaf area per unit of ground surface area, providing more detailed insights into vegetation structure. The traditional method of calculating LAI is to collect a certain area of vegetation and estimate the total leaf area based on selected calculations, which is time-consuming and difficult [27]. Studies have also shown that although NDVI and LAI are correlated, the strength of this relationship varies depending on vegetation type, seasonal variation, and environmental conditions [28]. In the

Table 1. Land cover-specific NDVI to LAI converting algorithms [29].

Land Cover	Regression	${\mathbf{R}}^2$	RMSE
Coniferous forest	LAI = 1.8 (NDVI + 0.069)/(0.815 − NDVI)	0.346	0.599
Mixed forest	LAI = 4.686NDVI/(1.181 − NDVI)	0.592	0.536
Temperate deciduous shrub	LAI = 8.547NDVI − 0.932	0.583	0.384
Alpine evergreen shrub	LAI = 9.174NDVI − 0.648	0.715	0.500
Broadleaved forest	LAI = 7.813NDVI + 0.789	0.732	0.461
Crop land	LAI = 6.211NDVI − 1.088	0.760	0.529
Alpine meadow	LAI = 3.968NDVI + 1.202	0.871	0.470

below, the regression equations are calculated differently for different vegetation types.

2.1.2. LST

As an essential metric in environmental science and climate research, LST reflects the thermal radiation emitted by the Earth’s surface layers. While air temperature captures atmospheric conditions near the ground, LST specifically measures heat released by soil, vegetation, and built environments, serving as a cornerstone for evaluating phenomena such as UHI, water cycle dynamics, and ecological stability [30]. Satellite-based thermal infrared (TIR) sensors, including instruments aboard Landsat, MODIS, and ASTER platforms, enable the acquisition of LST data through spatially comprehensive and temporally stable observations across diverse geographic scales [12]. Accurate derivation of LST necessitates meticulous calibration of raw sensor outputs using radiometric adjustments and atmospheric compensation techniques, requiring a rigorous application of physical models and empirical algorithms.

Key steps to calculate LST [31,32]:

Step 1: Top of Atmosphere (TOA) Radiance:

$$\mathrm{L}\mathsf{\lambda }={\mathrm{M}}_{\mathrm{L}}\times {\mathrm{Q}}_{\mathrm{c}\mathrm{a}\mathrm{l}}+{\mathrm{A}}_{\mathrm{L}},$$

(1)

where Lλ is TOA spectral radiance (Watts/(m² ∗ sr ∗ μm)), $M_L$ is Radiance multiplicative Band (No.), $A_L$ is Radiance Add Band (No.), $Q_{cal}$ is Quantized and calibrated standard product pixel values (DN).

Step 2: Top of Atmosphere (TOA) Brightness Temperature:

$$\mathrm{BT}={\displaystyle \frac{{\mathrm{K}}_2}{\ln \left[\left({\mathrm{K}}_1/\mathrm{L}\mathsf{\lambda }\right)+1\right]}}-272.15,$$

(2)

where BT is Top of atmosphere brightness temperature (°C), Lλ is TOA spectral radiance (Watts/(m² ∗ sr ∗ μm)), $K_1$ is $K_1$ Constant Band (No.), $K_2$ is $K_2$ Constant Band (No.).

Step 3: NDVI:

$$\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}={\displaystyle \frac{\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{R}\mathrm{e}\mathrm{d}}{\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{R}\mathrm{e}\mathrm{d}}},$$

(3)

where $Red$ is DN from the $Red$ band, $NIR$ is DN from Near-Infrared band.

Step 4: Land Surface Emissivity (LSE):

$${\mathrm{P}}_{\mathrm{v}}={\left({\displaystyle \frac{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}-{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{m}\mathrm{i}\mathrm{n}}}{{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{m}\mathrm{i}\mathrm{n}}}}\right)}^2,$$

(4)

where $P_v$ is Proportion of Vegetation, NDVI is DN from NDVI Image, ${NDVI}_{min}$ is Minimum DN from NDVI Image, ${NDVI}_{max}$ is Maximum DN from NDVI Image.

$$\mathrm{E}=0.004\times {\mathrm{P}}_{\mathrm{v}}+0.986,$$

(5)

where E is Land Surface Emissivity, $P_v$ is Proportion of Vegetation.

Step 5: LST:

$$\mathrm{LST}={\displaystyle \frac{\mathrm{B}\mathrm{T}}{\left[1+\mathrm{W}\times \left(\frac{\mathrm{B}\mathrm{T}}{\mathrm{14,380}}\right)\times \ln \mathrm{E}\right]}},$$

(6)

where $BT$ is Top of atmosphere brightness temperature (°C), W is Wavelength of emitted radiance (Band 10 is 10.8 µm), E is Land Surface Emissivity.

2.1.3. Process and Results

In Landsat-8,

$$\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}={\displaystyle \frac{{\mathrm{B}\mathrm{a}\mathrm{n}\mathrm{d}}_5-{\mathrm{B}\mathrm{a}\mathrm{n}\mathrm{d}}_4}{{\mathrm{B}\mathrm{a}\mathrm{n}\mathrm{d}}_5+{\mathrm{B}\mathrm{a}\mathrm{n}\mathrm{d}}_4}}.$$

(7)

Using the NDVI calculation formula and processed data, ArcGIS Pro 3.0.2 was used to draw the NDVI maps of Perth City between 2014 and 2023.

As shown in Figure 2 and Figure 3, Perth’s NDVI values showed a slight downward trend between 2014 and 2023, with values ranging from 0.175 to 0.257. The lowest NDVI was 0.175 in 2019, and the highest was 0.257 in 2016.

In order to better show the changing trend of NDVI over time, this study used Linear Regression Analysis to statistically analyze the NDVI value changing trend between 2014 and 2023 in Perth.

ArcGIS Pro 3.0.2 calculated the data and found that the slope value of Perth’s trend change was between −0.085 and 0.081. The data were reclassified and assigned values, using −0.01, −0.001, 0.001, and 0.01 as four interval values, and the obtained tendency was divided into five categories: Significantly decreased, Slightly decreased, Basically unchanged, Slightly increased, and Significantly increased. The percentages of different categories were calculated for Perth.

According to Figure 4 and

Table 2. Perth’s NDVI Tendency Slope percentage between 2014–2023.

	Perth
Significantly decreased (−0.085 ≤ s ≤ −0.01)	5.6%
Slightly decreased (−0.01 ≤ s ≤ −0.001)	55.4%
Basically unchanged (−0.001 ≤ s ≤ 0.001)	20.1%
Slightly increased (0.001 ≤ s ≤ 0.01)	17.7%
Significantly increased (0.01 ≤ s ≤ 0.081)	1.1%

, the areas with a slight decrease in vegetation cover in Perth between 2014 and 2023 account for the majority (55.4%) of the total area. Areas where vegetation coverage has declined significantly due to urbanization and urban spread, where vegetation needs to be removed for development, are mainly concentrated in the northern and southern fringe areas and densely built-up areas, accounting for 5.6%. The areas where it significantly increased are relatively small, accounting for 1.1%.

In Stage 1, we also selected 2014 and 2023 to see how LST has changed over these ten years.

As shown in Figure 5 and

Table 3. Perth’s maximum, minimum, and average LST changes.

LST (°C)	MAX (°C)	MIN (°C)	MEAN (°C)
2014	43.079	12.6	31.732
2023	49.986	16.895	34.771

, by calculating and extracting data from the obtained map, the maximum LST in Perth was 43.079 in 2014, which increased to 49.986 in 2023. The minimum LST was 12.6, which increased to 16.895 in 2023. The mean LST increased from 31.732 to 34.771. The overall data shows that Perth’s LST has been growing over the past decade.

As shown in Figure 6 and

Table 4. Correlation coefficient values between NDVI-LST.

	Year 2014	Year 2023
Correlation coefficient (r)	−0.456	−0.396

, according to the relationship curves between LST and NDVI in 2014 and 2023, and the calculated correlation coefficient (2014’s correlation coefficient (r) is −0.456, and 2023’s correlation coefficient (r) is −0.396), it is proved that there is a negative correlation between NDVI and LST, that is, the higher the NDVI value, the lower the LST value.

Through the summary of all the data and charts in Stage 1, Perth’s NDVI showed an overall downward trend between 2014 and 2023, while LST showed an upward trend, and there was a clear negative correlation between NDVI and LST.

2.2. Stage 2: Study the Relationship Between NDVI and Climate, Specifically Temperature and Precipitation

Having identified the overarching decline in vegetation coverage, this stage shifts its focus to disentangling the dual pressures of urbanization and climate variability.

According to

Table 5. Perth’s precipitation and maximum temperature values between 2014 and 2023.

Year	Perth Monthly Precipitation (mm)	Perth Mean Maximum Temperature (°C)
2014	62.45	25.4
2015	50.11	25.6
2016	67.26	24.2
2017	64.74	25
2018	58.91	25.3
2019	49.1	25.7
2020	54.86	25.3
2021	74.99	25
2022	58.4	25.2
2023	49.7	25.5

and Figure 7, when the NDVI value of Perth increases, the mean maximum temperature decreases, and when the value decreases, the mean maximum temperature increases. In 2019, when the NDVI was the lowest at 0.175, the maximum temperature reached the highest value of 25.7 °C. The relationship between precipitation and NDVI is the opposite. When precipitation increases, the NDVI also increases, and when precipitation decreases, the NDVI also decreases. In 2019, Perth reached the lowest precipitation, 49.1 mm.

According to the Australian Bureau of Meteorology, 2019 was the driest year on record for Australia; the annual mean temperatures for 2019 were above average and the highest on record for a large area. The main reason was a very strong positive Indian Ocean Dipole (IOD), which contributed to very low rainfall across Australia; many areas experienced insufficient rainfall [18].

To measure the correlation coefficient between NDVI with temperature and precipitation, this study uses the Pearson correlation coefficient (r) to measure linear correlation [33]. r is a number between −1 and 1, which is used to measure the strength and direction of the relationship between two variables. When r is negative and tends to −1, it means that there is a negative correlation between NDVI and another variable, and when r is positive and tends to 1, it means that the correlation between the two variables is positive.

Below is a formula for calculating the Pearson correlation coefficient (r):

$$\mathrm{r}={\displaystyle \frac{\sum \left({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}\right)\left({\mathrm{y}}_{\mathrm{i}}-\overline{\mathrm{y}}\right)}{\sqrt{\sum \left({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}}\right)\sum {\left({\mathrm{y}}_{\mathrm{i}}-\overline{\mathrm{y}}\right)}^2}}},$$

(8)

where ${\mathrm{x}}_{\mathrm{i}}$ and ${\mathrm{y}}_{\mathrm{i}}$ are the NDVI value and the maximum temperature/precipitation value in year i, $\overline{\mathrm{x}}$ and $\overline{\mathrm{y}}$ are yearly average values.

According to Figure 8 and

Table 6. Correlation coefficient values.

	Perth Monthly Precipitation (mm)	Perth Mean Maximum Temperature (°C)
Correlation coefficient with NDVI (r)	0.779	−0.754

, there is a positive correlation between precipitation and NDVI. When precipitation increases, NDVI increases. Perth has a pretty high correlation coefficient, which is 0.779. It also shows that there is a negative correlation between maximum temperature and NDVI. When the maximum temperature increases, NDVI decreases.

2.3. Stage 3: Understand the Relationship Between NDVI and LST Across the Seasons

While Stages 1 and 2 established macro-scale patterns, this stage was refined through seasonal lenses, adding temporal nuance.

In this stage, in order to understand the seasonal differences in the correlation between NDVI and LST, we selected January (summer) and August (winter) in 2021 for comparison and extracted the LST mpas, as shown in Figure 9.

As shown in Figure 10, according to the relationship curves between LST and NDVI in January and August 2021 and the calculated correlation coefficient (January Correlation coefficient (r) is −0.5272, August Correlation coefficient (r) is −0.1796), there is a negative correlation between NDVI and LST, that is, the higher the NDVI value, the lower the LST value. Moreover, by comparing the months of summer (January) and winter (August), the correlation between the two is higher in summer.

In order to make a more detailed comparison, we drew five sections from north to south across the entire urban area of Perth, as shown in Figure 11. The land use types covered by these five sections are also different, more detailed information is included in

Table 7. Five sections of information.

Number	Crossed LGAs	LU/LC
Section 1	Joondalup, Swan, Mundaring	Suburban Residential, Mariginiup Lake, Rangeland, Nature Reserve, Rangeland, Walyunga National Park, Rangeland
Section 2	Cambridge, Netlands, Subiaco, Perth, Belmont, Kalamunda, Mundaring	Suburban Residential, Park, Perth CBD, Commercial, Perth Airport, National Park, Suburban Residential
Section 3	Fremantle, Melville, Canning, Gosnells, Kalamunda	Park, Suburban Residential, Whaleback Golf Course, Central Commercial, Korung National Park, Mundaring State Forest
Section 4	Kwinana, Serpentine-Jarrahdale, Armadale	Suburban Residential, Park, Rangeland, Residential, National Park, Wungong Reservoir, National Park
Section 5	Rockingham, Serpentine-Jarrahdale	Park, Suburban Residential, Rangeland, Commercial, National Park

. ArcGIS Pro 3.0.2 was used to calculate the correlation between the NDVI data and LST data of all covered points in January and August 2021.

Through all the correlation charts of the five sections of NDVI and LST in January and August, as shown in Figure 12, most of the data show that NDVI and LST change in opposite trends. Through the calculation of the Pearson correlation coefficient (r), as shown in Figure 13, all the correlation coefficient results between NDVI and LST are negative, indicating that there is a negative correlation between the two. That is, the higher the NDVI value, the lower the LST value.

2.4. Stage 4: Study the Effect of LULC on NDVI and LST

Synthesizing insights from prior stages, this final stage interrogates how spatial configurations of land use amplify or alleviate UHI.

In the last stage, we integrated Perth’s Land Use/Land Cover data and LST data and used the Mean-Standard Deviation Method to reflect the degree of deviation of the LST from the average temperature in the study area. Moreover, we characterized the temperature variation due to different ground features. This method is defined by the distance each class deviates from the average of all elements. A combination of different standard deviation multiples of the mean value of the LST was used to divide the land surface heat field to define a UHI [34,35]. We used a 1/2 standard deviation as the classification interval to form the thresholds of different land surface heat levels concerning the average in the study, the classification details are shown in

Table 8. Heat island temperature classification using mean-standard deviation method.

Classification	Interval of Temperature Classification
Strong heat island area	T > μ + 1.5σ
Heat island area	μ + 0.5σ < T ≤ μ + 1.5σ
Medium temperature area	μ − 0.5σ ≤ T ≤ μ + 0.5σ
Secondary low temperature area	μ − 1.5σ ≤ T < μ − 0.5σ
Low temperature area	T < μ − 1.5σ

.

$$\mathrm{T}=\mathsf{\mu }\pm \mathrm{k}·\mathsf{\sigma }$$

(9)

where μ is the normalized surface temperature mean; σ is the standard deviation; k is the multiple of the standard deviation, which is taken as 0.5, 1.5 in this study.

As shown in Figure 14, the surface temperature classification results based on the mean-standard deviation method show that in January, strong heat islands areas, heat island areas, and medium temperature areas in Perth are mainly distributed in urban building-intensive areas, rangeland areas, and industrial areas, indicating that human activities are also one of the important factors affecting the formation of the heat island effect. The secondary low temperature areas and low temperature areas are mainly concentrated in the eastern and southern forest areas and the urban green space, indicating that areas with a large amount of vegetation cover have a significant contribution to the surface temperature decrease. By combining the heat island area distribution map in August, it can be seen that in terms of overall spatial distribution, the forest area in the east of Perth always maintains the cold island classification (secondary low temperature area, low temperature area), while the central urban area and densely populated areas maintain the medium temperature area classification or above.

As shown in Figure 15 and Figure 16, we have integrated LU/LC data into the calculation. According to the land cover composition of heat island areas at all levels, in January and August, the low temperature area and the secondary low temperature area are mainly dominated by trees, accounting for more than 80%. Built Areas and Rangeland, land use types with low vegetation coverage, account for the largest proportion in the strong heat island area, heat island area and medium temperature area. Moreover, in January, the strong heat island area is mainly dominated by Rangeland. January is summer in Australia, with high temperatures and long sunshine hours. Rangeland is usually dominated by sparse shrubs, grasslands or bare soil with low vegetation coverage, which can absorb more solar radiation, resulting in rapid surface warming and weak transpiration, and cannot effectively dissipate heat through water evaporation. Through the data and diagram obtained, areas with high tree and vegetation coverage significantly reduce the surface temperature and become “low-temperature green islands” in the city, while built-up areas (high hardened surface ratio) and low-vegetation rangeland become the core components of high-temperature areas due to their high heat absorption and slow heat dissipation.

3. Discussion

This research takes Perth, Australia, as the case study. Based on a series of remote sensing data and satellite data, a four-stage analysis provided important insights into the complex interactions between urban development, vegetation dynamics, climate change, and UHI. ArcGIS Pro software was used to visualize the data, and the research conclusions are as follows:

(1)

From Stage 1, the observed decline in NDVI (55.4% of the study area) and concurrent rise in LST over the decade align with global trends where urban sprawl replaces green spaces with impervious surfaces, exacerbating UHI effects. The strong negative correlation between NDVI and LST (supported by previous work [36]) underscores vegetation’s role as a natural thermal regulator. Urban design that prioritizes NDVI enhancement, even in dense areas, can mitigate thermal stress, offering actionable strategies for sustainable city planning.

(2)

From Stage 2, the significant correlations between NDVI and climatic variables (temperature: −0.754; precipitation: +0.779) highlight climate change as a dual-threat amplifier. Rising temperatures and altered precipitation patterns may compound vegetation loss caused by urbanization, creating feedback loops that intensify UHI. While these findings align with regional studies [37], they emphasize the urgency of integrating climate resilience into urban greening policies, particularly in semi-arid regions like Perth.

(3)

From Stage 3, the stronger NDVI-LST correlation in summer (−0.5272) vs. winter (−0.1796) suggests seasonal vegetation activity and water availability mediate thermal regulation. Summer dryness reduces vegetation cooling efficiency, amplifying UHI during peak heat periods. This seasonality implies that UHI mitigation strategies, such as targeted irrigation or drought-resistant greening, will account for temporal variability in climate-vegetation interactions.

(4)

Urban density is not correlated with NDVI; Perth city has a lower NDVI than the suburbs. At the same time, in this study, LST is found to be correlated with NDVI, suggesting that increasing urban density is preferable to urban sprawl if greenery is used as a mitigating factor. This will allow the population to grow whilst reducing the urban footprint and maintaining a high NDVI within the city.

(5)

In Stage 4, the medium and high temperature areas were mainly concentrated in urban dense areas and areas with low vegetation coverage, while the low temperature areas were mainly concentrated in areas with high vegetation coverage, such as forests. The results confirmed that areas with large vegetation coverage contributed significantly lower UHI values [38]. It was also confirmed by the LU/LC composition of heat island areas at all levels that areas with high tree and vegetation coverage have significantly lower surface temperatures (LST) and can provide “low temperature green islands” in the city, while built-up areas and low vegetation areas became the core components of high temperature areas.

(6)

Notably, rangelands emerged as unexpected contributors to summer UHI in Perth due to dry-season desiccation, challenging conventional views that non-urban vegetated areas uniformly cool cities. This finding calls for nuanced LU/LC management particularly in cities like Perth with long dry summers and arid natural landscapes.

4. Conclusions

This decade-long investigation provides useful insights into vegetation–climate–thermal dynamics in Mediterranean cities. The identified NDVI–climate relationships demonstrate vegetation’s dual role as both a climate regulator and a climate-dependent system. Moreover, integrating LST zoning with LU/LC data can identify urban forests and highly vegetation coverage areas as “low-temperature green islands”. From this research, it can be seen that vegetation coverage decline is an important driver of UHI intensification in Perth, with 61% of its metropolitan area experiencing NDVI reduction (2014–2023). This study demonstrates that the strategic integration of vegetation into urban fabric can counteract UHI effects even as cities grow.

In terms of the questions raised above, there are several uncertainties. It is not clear whether precipitation and temperature affected NDVI or the other way round, which UHI values can be attributed to various parts of Perth, to what extent surrounding greenery affects UHI of urban areas, and whether urban spread contributes more (or less) to UHI than urban densification. Further research is needed. However, this study substantiates vegetation optimization as an essential pathway for climate-adaptive urban development. In terms of natural heat islands (e.g., rangelands in this study), we need to consider interventions to increase their greenery whilst preserving and maintaining its ecological balance. Increasing urban density can be effectively mitigated with sufficient greenery. Appropriately, one of Perth’s Urban Greening Strategies 2023–2036 is to add more green roofs and vertical greening to high-density development projects [39]. Quantifying the thermal regulation services of green infrastructure provides actionable evidence for policymakers.