**1. Introduction**

More focus has been placed on global urbanization recently as more people around the globe move to urban areas every year. Today, more than half of the world's population resides in urban areas, and forecasts indicate that an increasing share of urban residents will be responsible for almost all future population increases. The complicated socioeconomic process of urbanization affects the built environment, relocating the population's spatial distribution from rural to urban regions, and converting once rural areas into urban ones. It has an impact on dominant occupations, lifestyles, cultures, and behaviors in both urban and rural regions, altering both the demographic and social structure. The key effects of urbanization include the quantity, size, and density of urban settlements as well as the population share between urban and rural inhabitants [1,2]. By ensuring that cities and human settlements are inclusive, safe, and resilient, SDG 11—one of the United Nations' "17 Sustainable Development Goals"—highlights the importance that cities play in the world's political agenda [3]. In the review of Estoque [4], despite initial efforts, the UN Global Sustainable Development Report 2019 [5] found that the world is not on track to achieve most SDG objectives including indicators related to SDG 11.

**Citation:** Purio, M.A.; Yoshitake, T.; Cho, M. Assessment of Intra-Urban Heat Island in a Densely Populated City Using Remote Sensing: A Case Study for Manila City. *Remote Sens.* **2022**, *14*, 5573. https://doi.org/ 10.3390/rs14215573

Academic Editor: Ronald C. Estoque

Received: 19 September 2022 Accepted: 24 October 2022 Published: 4 November 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Due to urban regions developing more quickly with population expansion, environmental changes ensue [6]. Loss of open space and animal habitats, water and air pollution, transportation, health concerns, and agricultural capacity are a few implications, while changing thermal properties is another result of urbanization and city growth. In terms of the increase in the Land Surface Temperature (LST) of the landscape, ongoing urbanization and the growth of impermeable surfaces are both factors [7,8]. As urban regions expand, the topography changes. Buildings, roads, and other forms of infrastructure take the place of open space and plants, for example, and permeable and moist surfaces eventually become impermeable and dry [9].

As a result of this development, urban heat islands (UHI) occur—a phenomenon in which urban areas experience warmer temperatures than their rural surroundings [10]. In particular, densely packed structures with little greenery develop "islands" with greater temperatures than their surroundings [9,11–13]. UHI may influence the increased risk of health-related conditions, increase in energy consumption, elevated pollutants, and water quality [14]. Urban heat islands (UHI) have the potential to have a detrimental impact on cities and their inhabitants, and as such, available resources and data must be used to detect and quantify these consequences. SDG 11 works toward making societies more sustainable and resilient by giving us a unique chance to make sure that the infrastructure we build today will still be useful in the future. This can be done by investing in parks and green spaces in cities, which will help reduce the "urban heat island effect" [3].

Aside from this, according to a growing body of research [15–17], "intra-urban" heat islands (IUHI), or regions within a city that are hotter than others due to an unequal distribution of heat-absorbing buildings and pavements, as well as cooler zones with trees and greenery, are becoming more prevalent [18]. Intra-Urban Heat Islands (IUHI) detection is of major interest to city planners since high temperatures influence energy usage and human health [16]. In 2015, Martin et al. [19] referred to surface intra-UHI as the detection of hotspots in a metropolis which is made possible by determining temperature thresholds by spatial reference. Consequently, the data can then be used to identify regions of interest in a city and potentially trigger alarms at a finer spatial scale. An example is a study conducted by Igergård et al. [20] in the Stockholm municipality.

In the literature, remote sensing is a good resource to understand the link between urban expansion and the characteristics describing the thermal changes in both geographical and temporal contexts [7,14,21]. Among remote sensing data, satellites are used more to estimate LST due to the thermal and passive microwave sensors aboard them. Although satellite data are very useful, Zhou et al. [14] stressed in their systematic review that retrieved satellite LST and air temperature differences, the effect of clouds, spatial and temporal resolution trade-off, SUHI quantification methods, varying land use land cover methods, and SUHI accuracy assessment are among the current challenges faced by UHI researchers. Worse, the limited availability of datasets for SUHI studies and applications exacerbates the challenges.

The increasing number of publications on the effect of UHI, particularly after 2016, reflects the scientific community's interest in disseminating information about this subject, which investigates its causes and ramifications from several viewpoints, including environmental, social, and economic [22]. The Philippines, like the rest of the world, is experiencing fast urbanization and a population density increase. Furthermore, these densely populated cities are largely clustered in Metro Manila [23,24]. In this context, statistically analyzing satellite data geographically and temporally, Landicho and Blanco [25] confirmed that intra-urban heat islands (IUHI) in Metro Manila are prevalent in 2019 while Alcantara et al. [26,27] conducted UHI studies in Quezon City. Estoque et al. [28], moreover, used satellite-derived surface temperature data and socio-ecological factors to analyze the present health risk in 139 Philippines cities. In addition, cities outside of Metro Manila were part of the Project GUHeat [24], which conducted urban heat island studies in cities such as Baguio [29], Cebu [30], Davao [31], Iloilo [32], Mandaue [33], and Zamboanga [34].

Given prior geographic biases in the literature, greater attention should be placed on understudied areas or cities, as proposed by Zhou et al. [14] and Almeida et al. [22] in their reviews. Furthermore, little published research explores how UHI affects the population because of a lack of fine-scale geographic population data [35]. Consequently, as there is inadequate research about UHI conducted in the country, area-specific assessment in cities like Manila would provide further details on how changes in the landscape impact the city's heat situation and will serve as a basis for urban planners and policymakers for mitigation and improvement. This also supports the goals of SDG 11 to aid the futureproofing of infrastructures for cleaner and greener cities.

The novelty of the present work is the use of space-time pattern mining to assess the presence of intra-urban heat islands using remote sensing data. Although this type of methodology is well established for space-time analysis applications, its usage on remote sensing data such as land surface temperature has not been extensively studied. Moreover, according to the author's knowledge, no work was dedicated to including the population and settlement data in such an assessment method for Manila City or any highly urbanized cities in the Philippines.

Its main purpose is to use satellite-derived and in situ meteorological remote sensing data to assess the presence of intra-urban heat islands in Manila City. Moreover, demographic data such as population and settlement data were used to enhance the assessment. Data represented in a space-time cube were used to carry out a space-time pattern mining approach in generating an Intra-Urban Heat Island (IUHI) map for Manila City. Finally, city-specific strategies to promote outdoor thermal comfort and hotspot interventions were also suggested. This paper is divided into five sections:


#### **2. Materials and Methods**

#### *2.1. Study Area*

As shown in Figure 1, Manila City is located in the northern Philippines archipelago, on the island of Luzon, on the eastern side of the old Manila Bay, with the Pasig River running through it [36,37]. As the Philippines' capital, Manila is considered to have the highest population density among the country's highly urbanized cities, and even among the world's densest cities. In 2020, the Philippine Statistics Authority [38] recorded that 1.84 million population reside in its 24.98 square kilometer land area, which translates to about 74,000 inhabitants per square km. The spatial attributes of the city are shown in Table 1.

**Table 1.** Spatial Attributes of Manila City.


**Figure 1.** Manila City's geographical location (**left**) and administrative boundary (**right**).

According to the Koppen Climate Classification [39] Manila has a tropical rainforest climate (Af). There is no dry season in a tropical rainforest environment, and it rains at least 60 mm per month throughout the year (2.36 in). Tropical rainforest climates do not have distinct seasons; it is hot and humid year-round, with frequent and heavy rains. Manila has an annual average temperature of 27.8 degrees Celsius, or 82.0 degrees Fahrenheit. With an average temperature of 85.0 ◦F (29.4 ◦C), April is the hottest month of the year, while the lowest month is January at 79.0 ◦F (26.1 ◦C) [39].

#### *2.2. Data and Data Sources*

This section enumerates the data and their sources including their descriptions, attributes, and the methods employed to obtain and prepare the data.

#### 2.2.1. Manila City Administrative Boundary

An administrative boundary represents subdivisions of areas, territories, or jurisdictions recognized by governments for administrative purposes [40]. The Philippines follows the Philippine Standard Geographic Code (PSGC) with different geographic levels such as region, province, city/municipality, and the smallest unit, barangay [41]. For the research, we need the shapefiles for Manila City at the city, district, and barangay levels. A published GitHub repository [42] was used since it is complete with all the needed geographic levels projected using the WGS 1984, latitude/longitude projection. These shapefiles were sourced from reliable webpages such as the OCHA Services Website [43] and GADM.org [44].

#### 2.2.2. In-Situ Meteorological Data

Meteorological raw data taken daily from 2014 to 2018 were provided by the weather bureau of the Philippines. The meteorological parameters include rainfall amount, mean temperature, maximum temperature, minimum temperature, wind speed, wind direction, and relative humidity. Since just one synoptic station is in Port Area, Manila (14.5878◦N latitude and 120.9690◦E longitude), only point data are available for Manila City.

#### 2.2.3. Population Data

Population density is a key metric for assessing domestic living circumstances. Due to the statistical approach used, traditional census statistics cannot represent the population's geographical distribution with a high degree of precision [35]. A high-resolution map estimate of the population density inside 30-m grid tiles was supplied by Data for Good Meta, which we used in this research. In this study, the population density demographic data for the year 2018 was used to give an insight into the distribution of people affected by the intra-urban heat island in Manila City. Since the downloaded data represent the whole country, we used ArcGIS Pro software to clip the region of interest based on the administrative boundary of Manila City. Aside from population density, those pixel grids with data are considered settlement areas while empty grids denote non-settlements areas in the city. Each cell's value represents the population density of that pixel/grid. This density may be expressed as a grid's area.

#### 2.2.4. Satellite Data

Satellite-derived remote sensing data in the study were taken from MODIS and Landsat 8 satellite data products. Daily land surface temperatures (day and night) were obtained from MODIS between 2014 to 2018 as complementary data for the meteorological data mentioned above. Consequently, spatial yearly data raster for land surface temperature and spectral indices were downloaded from Landsat 8.

• MODIS Land Surface Temperature Product

Land Surface Temperature data were derived from the Collection-6 MODIS Land Surface Temperature product to complement the available meteorological data at hand. Details of their retrieval were reported in [45]. In this study, data were retrieved to complement the meteorological data since global daily LST data can be obtained with this. Although the spatial resolution is low, the temporal resolution of the MODIS dataset is good and was deemed applicable for the correlation analysis presented in Ref [45].

• Landsat 8 Data Product

The Climate Engine web app (https://app.climateengine.com/climateEngine# (accessed on 9 February 2020)) [46] was used to download analysis-ready Landsat 8 data which were preprocessed using the Google Earth Engine [47] platform. The web application allows easy download of Landsat Bands, Spectral Indices, and Land Surface Temperature aggregated per year of study. In ecological studies, digital numbers and reflectance are the most used while studies involving thermal bands often use digital numbers and temperatures. For this study, we used top-of-atmosphere (TOA) reflectance products to obtain the land surface temperature (LST) and surface reflectance (SR) products for spectral indices such as Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Normalized Difference Built-up Index (NDBI).

• Land Surface Temperature

According to ESA [48], "Land Surface Temperature (LST) is the radiative skin temperature of the land derived from solar radiation. A simplified definition would be how hot the "surface" of the Earth would feel to the touch in a particular location. From a satellite's point of view, the "surface" is whatever it sees when it looks through the atmosphere to the ground. It could be snow and ice, the grass on a lawn, the roof of a building, or the leaves in the canopy of a forest. Land surface temperature is not the same as the air temperature that is included in the daily weather report."

Landsat 8 passes the equator at 10:00 am +/− 15 min (mean local time) [49] so the maps that will be generated are only based on measurements from this specific time of the day. While IUHI can be measured better in Manila City in the afternoon than in the morning, the limitation of satellite data to provide this led the researchers to use such Landsat data for the investigation. Deilami et al. [50] stressed in their review that the popularity of Landsat images for UHI studies can be attributed to factors such as being freely available

to researchers, their worldwide coverage with a reasonable spatial resolution of 30 × 30 m, and the long-term temporal coverage which enables researchers to extract the required information over a long period to monitor changes. Moreover, in their review article, about 22% of the papers reviewed use Landsat 8 data for UHI investigation with data available from 2013 to the present [50].

Raster data of land surface temperature data were taken from 2013 to 2022 on a yearly interval. Because of constraints in maximum cloud cover, LST within the year was obtained to depict maximum temperatures occurrence for that year. The top-of-atmosphere (TOA) product was used to illustrate the presence of cold and hotspots in the yearly intra-urban heat island map generated. Although the actual resolution Landsat 8 LST is 100 m, the analysis product downloaded from the climate engine is provided at 30 m.

• Normalized Difference Vegetation Index (NDVI)

The NDVI is a dimensionless index that describes the difference between visible and near-infrared reflectance of vegetation cover and can be used to estimate the density of green on an area of land. No green leaves produce a value near zero, yet calculations of NDVI for a particular pixel always yield a figure that falls between a negative one (−1) and a positive one (+1). A value of zero denotes no vegetation, whereas a value of close to one (0.8–0.9) represents the greatest potential density of green leaves [51]. The following formula is used to calculate NDVI:

$$\text{NIDVI} = \frac{(\text{NIR} - \text{Red})}{(\text{NIR} + \text{Red})} \tag{1}$$

For Landsat data, NDVI = (Band 5 − Band 4/(Band 5 + Band 4). This can be directly downloaded from the climate engine. Table 2 shows the ranges of NDVI and their corresponding land use land cover (LULC) classification.

**Table 2.** NDVI ranges for LULC Classification.


• Normalized Difference Water Index (NDWI)

NDWI is a measure of liquid water molecules in vegetation canopies that interacted with the incoming solar radiation. It is less sensitive to atmospheric scattering effects than NDVI [52]. This index uses NIR and SWIR bands where the resulting value ranges from minus one (−1) to plus one (+1). Positive values of NDWI correspond to high vegetation water content and high vegetation fraction cover. Negative NDWI values correspond to low vegetation water content and low vegetation fraction cover. In a period of water stress, NDWI will decrease. The following formula gives the NDWI value.

$$\text{NDWI} = \frac{(\text{NIR} - \text{SWIR1})}{(\text{NIR} + \text{SWIR1})} \tag{2}$$

For Landsat data, NDWI = (Band 5 − Band 6)(Band 5 + Band 6). This can be directly downloaded from the climate engine. Table 3 shows the ranges of NDWI values and the corresponding water content classification.

**Table 3.** NDWI ranges for Water Content Classification.

