3 - Pakistan

Introduction

Urbanization is a transformative process that offers societal benefits but also presents challenges, particularly concerning climate change. In 2021, 56.6% of the global population lived in urban areas, and projections estimate this will rise to 68% by 2050. Cities are developing strategies to address issues associated with rapid urbanization, such as increasing surface temperatures and degraded natural systems.

Rapid urbanization leads to air pollution (Yan et al. 2023), reduced vegetation cover (Gu et al. 2024), changes in land surface phenology (Zhang et al. 2024), and urban heat islands (Elnabawi and Raveendran 2024). These alterations, including vegetation reduction and increased impervious surfaces, compromise urban thermal comfort. Therefore, it is crucial to investigate the scale, temporal trends, and relationship between urban evolution and these issues.

Decreased vegetation cover due to urbanization is a significant impact of human activities, affecting the cooling of urban land surfaces and ecosystem balance (Li et al. 2023). Urban forests and vegetative spaces can reduce noise and air pollution (Tao et al. 2024). Preserving green cover helps prevent soil erosion (Efthimiou et al. 2022) and reduces the greenhouse effect by absorbing CO2 (Gholipour et al. 2022). Geospatial monitoring is essential for mitigating the impact of urbanization on green cover.

Urban Heat Island (UHI)

UHI is a phenomenon where a city's temperature is higher than its surrounding countryside (Cortes et al. 2022). There are two main types: atmospheric UHI, based on in situ air temperature measurements, and surface UHI (SUHI), quantified using satellite-based land surface temperature (LST) observations (Bateman et al. 2023).

Urbanization is a primary cause of UHI, leading to adverse effects on vegetation activity, energy consumption, and climate (Shen et al. 2023). A 1°C increase in air temperature can increase the mortality rate by up to 4.1% (Zafar et al. 2023). Science-based evaluations of the interplay between urbanization, LST, and vegetation cover can help mitigate UHI effects.

Remote sensing and geographic information systems offer tools to examine hydrology (Zafar et al. 2020), agriculture (Nadeem et al. 2023), carbon storage (Zafar et al. 2024b), urbanization (Athukorala et al. 2021), and UHI (Hurley and Heinrich 2024). Research has evaluated the effect of urbanization on vegetation using the Normalized Difference Vegetation Index (NDVI) (Liu et al. 2015). Studies using MODIS-based LST data have examined surface UHI intensity (SUHII) in cities globally, noting that average daytime SUHII is generally greater than average nighttime SUHII (Peng et al. 2012a). For example, one study noted daytime SUHII jumped $1.87 °C$ in 2011 from $0.01 °C$ in 2011 in 32 major cities in China (Zhou et al. 2014).

Studies have shown negative correlations between NDVI and urban area nighttime LST (Quan et al. 2016). Decreased vegetation leads to increased SUHII in outer city parts (Wang et al. 2016). Studies have examined the correlation between vegetation and SUHII in regions like the Yangtze River Basin (Yao et al. 2017) and coastal areas of Pakistan (Rizvi et al. 2021).

Pakistan, with a high urban population (approximately 37%, or 82 million people), is vulnerable to climate change and rapid urbanization (Satti et al. 2023). Karachi showed an 80% increase in urbanization in the first decade of the 21st century (Jabeen et al. 2015). The UN projects that over half of Pakistan's population will be urban by 2050 (United Nations 2017).

Rapid urbanization in Pakistan decreases vegetation, leading to challenges in sustainable landcover/land use (LULC) management, UHI effects, heat-related health impacts, and degradation of natural services. This compromises climate change adaptation efforts. There is a scarcity of studies on SUHII in Pakistani cities, especially concerning long-term effects of urbanization on vegetation cover. Empirical data is needed to understand the spatio-temporal patterns of vegetation loss and its impact on SUHII to inform urban planning and sustainable development.

Study Objectives

The primary objectives of this study on nine major metropolitan cities in Pakistan are:

1. Exhibit the urbanization effect on SUHII and vegetation during 2001–2020.

2. Evaluate the long-term trends for urbanization effects on vegetation cover and SUHII.

3. Study the spatial distributions of the effect of urbanization of vegetation and SUHII of spatial base.

This study assesses the long-term relationship between SUHII and vegetation changes on a regional scale to aid in planning sustainable cities, especially for developing countries with rapid urbanization.

Data and Methodology

Study Area

Pakistan, located between $23.6°$-$38°$ E and $61°$-$78°$ N, has a total area of $796,096 km^2$. Its elevation varies from the Arabian Sea to the Himalayas (Chaudhry et al. 2009). The country's LULC includes 4.2% forest, 26.1% rangeland, 8.7% desert, 1% built-up land, and 0.9% water bodies. Most of the country is arid to semi-arid, except for the northern part, which receives 800 to 2000 mm of rain annually (Jiang et al. 2023). Balochistan is the driest part, with 210 mm annual average rainfall (Ali et al. 2021).

Summer temperatures in plain areas remain above $40 °C$, and winter temperatures range from $4 °C$ to $20 °C$ (Khan 2019). Pakistan is the sixth most populated country with high urbanization and population growth rates (United Nations, 2018). Migration from rural areas to cities pressures land resources in urban regions.

The study selected nine major cities based on their metropolitan nature: Lahore (LHR), Faisalabad (FSD), Multan (MUX), Gujranwala (GJR), Sialkot (SKT), Islamabad/Rawalpindi (ISB/RWP), Peshawar (PEW), Mardan (MAR), and Karachi (KHI) (Fig. 1).

Karachi has a population of over 16 million and contributes 15% to the national GDP (Babar et al. 2021). Lahore has almost 13 million residents (Rehman et al. 2023). Faisalabad has over 9 million and is an industrial center. Rawalpindi and Islamabad have a total population of 8.4 million (N. Khan and Shahid 2024). Gujranwala and Multan have populations of 5.9 million and 5.3 million, respectively. Peshawar has 4.8 million residents (Basit et al. 2024). Sialkot has approximately 4.5 million people, while Mardan houses 2.7 million (Zafar 2024).

Data

Yearly composite LULC data with a spatial resolution of 500 m (MCD12Q1) from MODIS was used, with the IGBP classification layer used to extract urban and rural areas (Zafar et al. 2023). MOD11A2 was used to obtain LST data, with a 1 km spatial resolution (Dian et al. 2020; Sekertekin et al. 2020).

MODIS LST products have retrieval errors usually less than 1 k, with a root means square (RMS) less than 0.5 k (Wan 2008). Errors for MODIS LST products are within $±1 k$ in 39 out of 47 evaluated cases (Rigo et al. 2006).

MODIS monthly composite EVI product with 1 km spatial resolution (MOD13A3) was used to compute vegetation dynamics (Wang et al. 2022; Yao et al. 2018). EVI is more efficient than NDVI for monitoring vegetation changes in urban environments (Zhou et al. 2016).

Data were processed using Google Earth Engine (GEE) to extract urban areas, rural areas, EVI, and LST.

Methodology

Extraction of Urban and Rural Areas

Urban and rural areas were delineated using the MCD12Q1 land cover dataset, with urban areas identified by a digital number (DN) of 13 and water bodies by a DN of 17 (Chen Dan and Keenan Trevor 2021; Keeratikasikorn and Bonafoni 2018; Mehmood et al. 2022; Zafar 2024; Zafar et al. 2023b). Urban growth was obtained by comparing the urban area for the last year of study with the first year.

Computation of SUHII and ΔEVI

The interplay of urbanization, vegetation, and SUHII was examined on seasonal and annual bases. Each year was separated into seasons: winter (December to February), spring (March to May), summer (June to August), and autumn (September to November) (Haroon & Zhang; JiahuaYao, 2016).

SUHII was computed as follows:

$\Delta LST (SUHII) = LST<em>u - LST</em>r$

Where: \LSTu denotes the mean LST of urban areas, and \LSTr represents the mean LST of rural areas.

$\Delta EVI = EVI<em>u - EVI</em>r$

Where: \EVIu denotes the mean of EVI from urban areas, and \EVIr represents each city’s mean of EVI from rural areas.

Trend Analysis

Time series datasets of SUHII and ΔEVI were evaluated using the Mann–Kendall (MK) trend test and Theil–Sen estimator (Sen’s Slope) (Sen 1968). The MK trend test is a nonparametric test widely used for time series analysis (Agbo et al. 2023; Zafar et al. 2023a).

The rates of variations in SUHII and ΔEVI were calculated using Sen’s slope on annual and seasonal basis.

Correlation Analysis

The Pearson correlation test (Havlicek and Peterson 1976) at a 5% significance level was adopted to evaluate the relationship between SUHII and ΔEVI at seasonal and annual intervals.

$r = \frac{\sum<em>{i=1}^{n} (x</em>i - \bar{x})(y<em>i - \bar{y})}{\sqrt{\sum</em>{i=1}^{n} (x<em>i - \bar{x})^2 \cdot \sum</em>{i=1}^{n} (y_i - \bar{y})^2}}$

Continuous Wavelet Transform (CWT) was applied to examine the temporal correlation between SUHII and ΔEVI. The Morlet wavelet was selected for its balance between time and frequency resolution.

The Cross-Wavelet Transform (XWT) is denoted as $W_{xy}(n, s)$. Wavelet coherence is a method employed to detect specific oscillations in signals that are not constant over time.

The wavelet coherence $R^2(n, s)$ can be mathematically denoted as follows (Torrence and Compo 1998):

$R^2(n, s) = \frac{|S(s^{-1} W<em>{xy}(n, s))|^2}{S(s^{-1} |W</em>x(n, s)|^2) \cdot S(s^{-1} |W_y(n, s)|^2)}$

The wavelet transform (S) serves as a smoothing operator, and the local correlation coefficient $R^2(n, s)$, calculates the strength of correlation within the time-frequency domain, ranging from 0 to 1 (Goodell and Goutte 2021). The Monte Carlo method was applied for statistical testing.

Results

Urbanization Trends

The total urban area of selected cities in Pakistan increased from $2126.78 km^2$ to $2230.17 km^2$ during the study period. Every city showed an increment in urbanization area. The average area for all 9 cities increased from $236.36 km^2$ to $247.79 km^2$. Lahore showed the maximum growth of $32.8 km^2$, increasing from $495.05 km^2$ in 2001 to $527.85 km^2$ in 2020. Punjab province cities showed an annual average growth of $12.63 km^2$. Mardan experienced a minimum expansion of approximately $3 km^2$. Karachi, Lahore, and Faisalabad contain larger urban areas.

Spatial Variations of EVI and LST

The EVI results showed significant variations in vegetation cover, particularly between urban cores and peripheral areas. Densely urbanized regions like Karachi, Lahore, and Islamabad/Rawalpindi exhibited lower EVI values (0.01 to 0.05), indicating sparse vegetation. Less urbanized areas of cities like Gujranwala and Multan showed higher EVI values (0.38 to 0.39), indicating better vegetation cover. Urbanization leads to a reduction in vegetation.

LST data demonstrated higher temperatures in urban areas than in rural areas. Highly urbanized cities such as Karachi, Lahore, and Islamabad/Rawalpindi had elevated LST values, reaching up to $33.4 °C$, $33.5 °C$, and $31.8 °C$, respectively, contrasting with rural areas where LST values dropped as low as $17.8 °C$ in Islamabad/Rawalpindi. Cities like Multan and Peshawar exhibited elevated LST values in the urban areas, with temperatures around $36.8 °C$ and $34.5 °C$, respectively. This highlights the impact of urbanization on increasing surface temperatures thus emphasizing the need for green spaces.

ΔEVI and SUHII

All cities showed negative values of average ΔEVI, representing vegetation loss. Gujranwala experienced the minimum average ΔEVI (-0.256) in winter, and Islamabad/Rawalpindi's maximum values were noted in spring (-0.004). On an annual basis, Multan had the highest average ΔEVI, followed by Gujranwala and Faisalabad. The winter season had the highest average ΔEVI, while the spring season exhibited the minimum average ΔEVI among all seasons.

Most cities showed positive values for SUHII in the summer, autumn, and spring seasons, as well as for the annual average, with Karachi being the only city with consistently negative values of SUHII. The maximum average SUHII was in Mardan (4.82 °C) in the summer, while the minimum was observed in KHI (-5.37 °C) in the spring season. Across all cities, the summer season exhibited the maximum average SUHII of $2.47 °C$.

Temporal Trends

Temporal Trends of ΔEVI

Average ΔEVI for all nine cities showed a statistically significant decreasing trend for all seasons and annually. The autumn season demonstrated the maximum statistically significant decreasing trend (-0.00224 year-1). Declining trends for ΔEVI in spring, winter, summer, and annual seasons are also statistically significant. From 2001 to 2020, each city displayed a statistically significant decreasing trend for annual ΔEVI.

Temporal Trends of SUHII

Trends revealed that SUHII is increasing for all seasons and annually. The highest increasing rate was observed for spring (0.041 °C year-1) and summer (0.038 °C year-1). The minimum increase rate was observed in annual SUHII (0.029 °C year-1).

Annually, all cities except Mardan showed increasing trends for SUHII. The most increasing trend was detected for Islamabad/Rawalpindi (0.114 °C year-1). For the winter season, ISB/RWP demonstrates the maximum increasing trend of SUHII, and the minimum is observed in GJR. The increasing trends for spring and summer seasons are observed in the other cities with some showing decreasing trends.

Correlation

Averaged ΔEVI and SUHII across all cities exhibit negative correlations for all seasons and annually. Correlations between ΔEVI and SUHII range from -0.894 in spring to -0.368 in winter. Most cities showed negative correlations annually. Of note, most of the cities showed a significant negative correlation existed between ΔEVI and SUHII, which is an indication of vegetation loss in cities.

Wavelet coherence analysis indicated periods of coherence occurring at approximately 8, 16, and 32 months. A comprehensive understanding of these interactions is important for urban planning and environmental management methods.

Discussion

The SUHII and ΔEVI were analyzed on annual and seasonal bases. The average SUHII for 2001–2020 varied from $0.21 °C$ in winter to $2.84 °C$ in summer. Decreased vegetation cover is a primary reason for the observed increase in SUHII. Forests and croplands are being converted to impervious surfaces, causing increased SUHII and decreased vegetation cover.

The analysis reveals a negative relationship between SUHII and ΔEVI, especially during spring and summer, where vegetation loss amplifies the urban heat island effect. The correlations between SUHII and ΔEVI range from -0.5 to -0.8, with higher effects in the summer, which demonstrates that urbanization exacerbates the heat island effect.

Due to urbanization, the population of cities grew, and many problems have emerged: the reduction of cropland and forest, an increase in the UHI effect, and an upsurge in air pollution. To alleviate the impact of urbanization, it is recommended to prioritize green spaces to mitigate SUHII and promote energy-efficient buildings to decrease SUHII. Climate mitigation techniques should be executed with reflective materials and enhanced public transit systems. Public awareness campaigns should inform individuals about the advantages of green spaces and promote the practice of urban greening.

Limitations and Way Forward

This study has some limitations, primarily the reliance on MODIS data, which may have spatial resolution limitations in densely populated urban areas. The study does not account for other factors such as socioeconomic variables, local climatic variations, and heat emissions from industrial activities, and is limited to a 20-year period, which may not capture long-term climatic trends and urbanization patterns. Future research should incorporate higher-resolution data like Landsat and Sentinel satellites, consider additional influencing factors, and extend the temporal scope for more recent data. Integrating machine learning techniques could improve predictive modeling and provide more insights into the UHI phenomenon.

Conclusion

This study investigated the temporal changes and trends of SUHII and ΔEVI for cities in Pakistan from 2001 to 2020 and demonstrates the need for sustainable management in urban cities:

• Summer Season showed the maximum (2.47 °C), and the winter season showed the minimum (0.21 °C) averaged SUHII.

• Temporal trends for SUHII for an average of all cities showed an increasing trend.

• A negative correlation was observed between SUHII and ΔEVI.

The results from this study will provide references for sustainable LULC management, designing equitable green spaces, and supporting efforts regarding climate change impacts mitigation and adaptation. Future studies should focus on analyzing each city’s behavior separately for a detailed perspective. Additionally, more variables like climatic variables, heat emission from artificial structures, and albedo should be considered in future studies.