*3.6. Intra-Urban Heat Island Map Assessment and Mitigation Strategies* 3.6.1. Location Assessment

Using the IUHI Map of Manila City, areas classified as "preserve" and "intervene" were examined visually using high-resolution maps from Google Earth Pro.

From the IUHI map, areas that need intervention were assessed by visually inspecting the locations to see the morphology of the areas exhibiting consistent surface temperatures during the study period. Based on the inspection, most of these areas fall within the Sampaloc district, which is part of Manila City's university belt shown in Figure 14E–H catering to Manila's academic population. The area's abundance of hotels and boarding houses makes it ideal as a dormitory and as a commuting town [36]. Moreover, there are also a few areas situated in Tondo District (A, B, and C) which is among the biggest urban poor communities in Manila City. Area D, on the other hand, mainly points toward a commercial location in Paco District.

Looking at the high-resolution satellite images, the areas shown in Figure 14 represent commonality in terms of their urban structure. It is noticeable that these areas (A, B, C, E, F, G, and H) are mostly residential and is characterized by predominantly settlement and housing locations with narrow streets and sidewalks. Although there are attempts to introduce urban soft scape via trees and vegetation, these are few and sparsely distributed within the areas of concern. In general, roads and walkways are mainly built with asphalt and concrete which might contribute to higher surface temperatures. There is also commer-

cial space identified, such as (D), which seemed to have establishments and buildings and parking spaces made of either asphalt or concrete as well.

**Figure 14.** Some areas with the "Intervene" Class of Action. (**A**–**H**) are the areas highlighted to show their morphologies.

The same approach was applied in examining the areas to be preserved shown in Figure 15. Aside from the stretch of Pasig River amidst Manila City, the Intramuros district including Rizal Park Complex (part of Ermita district) as shown in (D) shows large areas with relatively lower surface temperatures. It is the historic core of Manila and is described as the "walled city" where walls surrounding the area are present until today. The Intramuros area has evident low surface temperature due to its strategic location. Aside from being situated near a body of water (Pasig River), the area is surrounded by greenery (mostly grass and some shrubs and trees) which is part of a golf range. On the other hand, the Rizal Park complex is one of the largest urban parks in Asia wherein the area is a combination of vegetation and trees, gardens water features, and shaded areas.

**Figure 15.** Some areas with the "Preserve" Class of Action. (A-H) are the areas highlighted to show their morphologies.

Predominantly, most of the areas shown in Figure 14 exhibit common morphological characteristics. For instance, areas shown in A, F, and H are either surrounded or akin to bodies of water and other water features, while areas shown in C, D, and G contain substantial vegetation and green areas. In addition, areas like B and E, although residential, also contain a decent quantity of trees spread within the area.

In this visual inspection, the two areas have distinguishable features which relate to the surface temperature in the area. Understanding the morphological characteristics of the cold spots (preserve) can help in planning the mitigation strategies needed to improve the thermal condition of the hotspots (intervene).

#### 3.6.2. IUHI Class of Action and LULC Indicators Assessment

Overlaying the 2022 maps with the IUHI Map, the average values per class of action are shown in Table 13. It can be observed that the average NDVI values do not provide a clear distinction among the classes of action since the expected cold spots (water bodies and vegetation) have values at the extremes of the index. On the contrary, NDWI and NDBI average values convey the results. For instance, for "preserve", the average NDWI translates to higher water content while the average NDBI shows non-built-up areas. A similar remark can be drawn for "intervene" values where the average NDWI means low water content and the average NDBI falls in the built-up area category.

**Table 13.** Average values of LULC indicators per IUHI class of action.


Using the same data, we also investigate how the individual index classification is distributed among the IUHI class of action to validate it with the literature. Table 14 provides the distribution of NDVI-based LULC per class of action. It can be observed that areas considered as "preserve" have a higher proportion of water bodies and vegetation while areas considered as "intervene" mostly fall into the urban built-up category.


**Table 14.** Distribution of LULC per IUHI class of action based on NDVI.

Table 15 shows the distribution of water content category per IUHI class of action based on NDWI. Based on the proportions, most parts of the areas considered "preserve" have high water content while those for "intervene" have low water content. This shows that the water content of the area has an impact on its surface temperature.

**Table 15.** Distribution of Water Content category per IUHI class of action based on NDWI.


Table 16 shows the distribution of built-up categories per IUHI class of action based on NDBI. As shown about two-thirds of the "preserve" area occupy non-built-up locations while almost all parts of the "intervene" area are built up. This illustrates the effect of built-up areas such as infrastructures, roads, and buildings that contribute to higher surface temperatures in the city.


**Table 16.** Distribution of Built-up category per IUHI class of action based on NDBI.

Based on the observations above, LULC indicators allow us to assess the IUHI maps according to different aspects of the indices. By understanding such categories and how they are related to the IUHI map class of action, the areas can be quantitatively described and later can be used to incorporate mitigation strategies.

#### 3.6.3. IUHI Class of Action and High-Resolution Settlement Layer Assessment

The high-resolution settlement layer which consists of population per pixel and settlement categories was also used to assess the IUHI map. The demographic data represent the year 2018 which is the latest available during the conduct of the study.

By superimposing the generated IUHI Class of Action Raster and High-Resolution Settlement Layer containing population per pixel and settlement class, an attribute table is generated. From this attribute table, statistics about the population data and settlement information are taken and summarized in Tables 16 and 17. An example of the attribute table is shown in Figure 16. The object ID represents the corresponding pixel where values related to the attributes are provided. In the Population/Settlement column, population per pixel is shown while those that indicate zero mean a non-settlement pixel.

**Table 17.** Distribution of affected population per IUHI class of action.



**Figure 16.** Excel Sheet of the superimposed IUHI Class with Population/Settlement Data.

In Table 17, although the percentage of "intervene" areas is small compared to the other IUHI categories, there are still about 61 thousand of the population affected by higher surface temperatures. As Manila is a densely populated city, the population despite its small percentage is still not negligible.

In Table 18, the distribution of settlement categories (from the high-resolution settlement layer data) with IUHI class of action is presented. We can see that about three-fifths (1.70%/2.65%) of the "intervention" area falls on settlement areas. This implies that most of these areas are inhabited by people, which was backed up by the visual inspection in Section 3.4.1. For the "preserve" class of action, most of the areas are non-settlement areas which are mostly vegetated locations, parks, and those near the water features.


**Table 18.** Distribution of settlement category per IUHI class of action.

3.6.4. IUHI Class of Action and Land Surface Temperature

To compare the variation of temperature between the cold spots (preserve) and hotspots (intervene), the yearly land surface temperature was calculated for each class of action.

A summary table of the average LST per year per class of action is shown in Table 19. As can be seen, the average difference between the warmest and coldest areas in Manila City is 6.13 ◦C. The difference through the years has a small deviation wherein the lowest is recorded in 2013 while the highest is in 2017. To better see the trend, a graphical representation of Table 18 is shown in Figure 17.

**Table 19.** Average LST (◦C) per year per IUHI class of action.


**Figure 17.** Average LST per year per class of action.

3.6.5. Mitigation Strategies for Areas That Need Intervention

With the assessment done in Sections 3.6.1–3.6.4., the differences in temperatures at different urban morphologies were tackled. SDG 11, with its aim to make cities and human settlements inclusive, safe, resilient, and sustainable, can only be realized by not only understanding the city's current situation but also providing means to identify vulnerable areas and implementing solutions to solve existing problems. While the assessment provides information about the presence of intra-urban heat islands in Manila City, this also offers insights into which area in the city policymakers can focus on in offering mitigation strategies. In the analysis, for example, urban settlement and residential areas with narrow streets and sidewalks, asphalted roads and walkways, and concrete commercial spaces can contribute to high surface temperatures, while areas surrounded by and near bodies of water/water features, substantial green spaces/vegetation/trees, and residential areas with decent quantities of trees are places of lower surface temperature. With this in mind, the following mitigation strategies are suggested to help ameliorate the effect of urban heat islands, some of which were adapted from the compendium of strategies by the U.S. Environmental Protection Agency [9].

As part of the local institutional mechanism to address SDG 11, the government can include the following in their priority development initiatives, especially in the identified areas for intervention:


Additionally, the current densely populated city cannot accommodate extra large-scale trees and vegetation anymore, so the following alternatives can be employed:


These are just some of the mitigation strategies applicable to Manila City in its current state. For the attainment of SDG 11 and to address the ill effects of UHI that would result in a sustainable and livable city, a holistic approach is necessary for implementing such strategies. It should be highlighted that the local government unit including its population plays an important role in this.

#### **4. Discussion**

The result of this study shows evaluation methods using multiple sources to understand the presence of Intra-Urban Heat Islands in Manila City, Philippines. The satellite data retrieved from Landsat 8 provided distribution maps from 2013 to 2022 which include land surface temperature and LULC indicators such as NDVI, NDWI, and NDBI. More satellite data from MODIS Terra were also obtained to provide point data for land surface temperature data for both day and night. In addition, in-situ data were obtained at Port Area, Manila City, with meteorological data measurements from 2014 to 2018. Finally, raster data containing population density and urban settlement category for 2018 were acquired to represent demographics data for Manila City.

The LST and air temperature data show that beginning in March and continuing through April and May, there is an increasing tendency in the values, whereas values begin to decline in October and continue through January and February, which is similar to the observations in [28,65]. This trend is because March to May is the hot dry season in the Philippines while October to January is rainy and December to February is the cool dry season. In addition, it was found that there is a significant linear relationship between air temperature and land surface temperature based on daily data, while relative humidity shows a weak correlation with the LST data.

In terms of outdoor thermal comfort, a limited analysis was done due to limitations provided by the point measurements of meteorological data in Port Area Manila, City from 2014 to 2018. Despite these limitations, we used the meteorological parameters to estimate the Physiological Equivalent Temperature (PET) thermal index using the RayMan microclimate model. With the calculated PET thermal index values, corresponding physiological stress levels were provided to understand the outdoor thermal comfort. We observed that mild heat stress may be routinely experienced in May, and at certain times in April and June. From July through December, moderate heat stress was seen; however, the thermal comfort zone, where there is no heat stress, did not emerge until January and February. Understanding the thermal comfort in this location may also help us predict the outdoor thermal comfort in other areas of Manila City. It should be noted that the location of Port Area, Manila City is near Manila Bay, which may indicate that the meteorological parameters may not be representative of the whole of Manila City. The calculation of thermal index is calculated based on the meteorological parameters while these meteorological parameters were correlated with land surface temperature. With this, we have associated thermal comfort indirectly with the land surface temperature such that while Port Area, Manila City is not considered as an area for intervention, it still experiences heat stress. Therefore, other areas which are considered areas for intervention are more likely to experience worse thermal stress than Port Area, Manila. This observation and the generated IUHI map can be the basis for selecting additional meteorological stations in areas that may experience worse heat stress, so it can be monitored and provided by mitigation strategies in the future.

Land Use Land Cover (LULC) indicators such as NDVI, NDWI, and NDBI were very useful in understanding the morphological characteristics of Manila City, while their relationship with land surface temperature was also considered. Results of the multivariate analysis show that clusters can be generated based on combinations of these LULC indicators relative to land surface temperature. The clustering findings reveal that values with low NDWI, moderate NDVI, and high NDBI are grouped in the high LST cluster. Low NDWI corresponds to low water content, and high NDBI corresponds to urbanized zones; therefore, this is also predicted. Correlation between LULC indicators and LST shows the link between LST and LULC indicators with their respective slope of linear fit and frequency distribution chart. The data demonstrate a direct association between LST and NDBI at r = 0.361, meaning highly built-up regions have high reported temperatures. The multivariate analysis supports this finding. LST and NDVI (r = 0.064) and NDWI (r = 0.365) have indirect relationships. A Low Pearson correlation between LST and NDVI implies low temperatures for water bodies and vegetation, whereas mid values imply built-up areas. High water/moisture locations exhibit lower surface temperatures using LST and NDWI. Based on these data, it can be argued that NDWI is a better indication than NDVI for land surface temperature, which agrees with Alexander et al. [66]. NDBI is a good indication for LST, according to the data.

The creation of a space-time cube for LST made spatiotemporal pattern analysis easier. Using the space-time mining tools in ArcGIS Pro, Emerging Hotspot Analysis and Local Outlier Analysis were performed. The resulting reclassified maps of EHSA and LOA were respectively used as input to the suitability analysis model to generate an easy-tounderstand Intra-Urban Heat Island (IUHI) class of action map between 2013 to 2022. Such a map contains the class of action (preserve, monitor, and intervene) as well as the administrative boundaries at the city, district, and barangay levels.

In the location assessment, the focus was given to areas to preserve and intervene. Understanding the morphology of "preserve" locations helps in the provision of mitigation strategies for the "intervene" locations. The results show that the highest temperatures are in areas with a concentration of urban settlement areas, buildings, and establishments while those with low temperatures are areas with enough vegetation and near bodies of water. Visual inspection revealed that most "intervene" areas are in the Sampaloc district and university belt. Such an area has a high concentration of universities and colleges while within it are settlement areas, establishments, and concrete roadways which are deemed contributory to the high surface temperature. Knowing this is crucial because aside from its residents, the population in this area swells due to students and employees coming from the nearby province during the daytime. Other intervention areas can be found in the Tondo district, which is home to urban poor communities, while there are also hotspots in the Paco district, which mainly points toward a commercial location. These regions are largely residential, with small streets and sidewalks and a concentration of settlements and dwelling sites. In the regions of concern, initiatives to create an urban soft scape employing trees and plants are limited and scarce. Roads and sidewalks are often constructed with asphalt and concrete, which may contribute to greater surface temperatures. There is also an identifiable commercial area, which seems to have asphalt or concrete companies, buildings, and parking spaces.

On the other hand, "preserve" areas are mostly located in Intramuros, Rizal Park, and sites near the Pasig River banks. Most of the regions have similar physical characteristics. For example, these places are either next to or resembling bodies of water and other water features, while other areas have extensive vegetation and green landscapes. Additionally, residential neighborhoods feature a significant number of trees. Noting these characteristics, mitigation strategies appropriate to the "intervene" areas can be established.

The IUHI class of action was also assessed relative to the corresponding LULC indicator values. While NDVI does not provide a clear distinction among the classes of action, NDVI and NDWI convey their results. For example, the average NDWI for "preserve" indicates a greater water content, but the average NDBI indicates undeveloped lands. Similar observations may be made for "intervene" values when the average NDWI indicates a low water content and the average NDBI falls under the category of "built-up area." Using the same data, we also investigate how the individual index classification is distributed among the IUHI class of action to validate it with the literature. It may be noticed that regions designated as "preserve" have a greater percentage of water bodies and vegetation, higher water content, and occupy non-built-up locations while regions designated as "intervene" are in urban built-up areas with lower water content.

With the high-resolution settlement layer (HRSL), the distribution of the affected population including the settlement category for 2018 was assessed. Upon superimposing the HRSL with the IUHI class of action map, about 61 thousand of the population are affected by higher surface temperatures as indicated in the "intervene" areas. Despite the small percentage of "intervene" locations compared to the entire Manila City; it is evident that such a small percentage is not negligible due to the city's dense population. In terms of the settlement category, the "intervene" locations are mostly located in settlement areas while the "preserve" locations are in non-settlement areas. Such observation is aligned with what was observed in the visual inspection of locations using high-resolution satellite images.

Summarizing the LST values per year per class of action reveals an average LST for "preserve", "monitor" and "intervene" as 34.43 ◦C, 38.51 ◦C, and 40.56 ◦C, respectively. The result of this study clearly shows differences in temperature within Manila City. With these data, the average difference between cold and warm areas is about 6 ◦C, just as in the discussion in [20]. As the LST statistics are based on the highest LST readings for each site, it should be understood that the highest LST recorded differentiates 6 ◦C between specific urban areas. We avoided pixel-based comparison in the overall analysis to evaluate clusters of warm and cold regions appropriate to a city viewpoint and to make the analysis more significant.

Finally, applicable mitigation strategies based on the assessment of cold spots and hotspots in the city were proposed. These strategies support the attainment of SDG 11 in making cities and human settlements inclusive, safe resilient, and sustainable. Such strategies are (1) water mist/dry-mist sprayer in pavements and pedestrians, (2) provision of shade structures, (3) using cool materials for pavements and roofs, (4) provision of cooling center, (5) conversion of regular walls to green walls, and (6) plants in plant boxes, road isles, and indoors.

### **5. Conclusions**

This study presents the use of satellite-derived data and meteorological data to assess the presence of an intra-urban heat island in Manila City, Philippines. To address SDG 11 and provide better insights to make cities and human settlements inclusive, safe resilient, and sustainable in terms of UHI, different assessment methods were used and established. The assessment includes (a) understanding the temporal variability of air temperature measurements and outdoor thermal comfort based on meteorological data, (b) comparative and correlative analysis between common LULC indicators (NDVI, NDBI, and NDWI) to LST, (c) spatial and temporal analysis of LST using spatial statistics techniques, and (d) generation of an intra-urban heat island (IUHI) map with a recommended class of action using a suitability analysis model. Finally, the areas that need intervention are compared to the affected population, and suggestions to enhance the thermal characteristics of the city and mitigate the effects of UHI were established. Results show that there exists a clear difference between cold and warm areas within Manila City. Overall, residential areas, asphalted and concrete roads and walkways, and some commercial establishments and buildings exhibit higher surface temperatures compared to areas with vegetation and near bodies of water. Based on the results, mitigation strategies applicable to Manila City were proposed to improve the areas which need intervention.

In the future, we plan to realize these strategies by partnering with the local government unit to implement these proposed measures. We also advise providing additional meteorological stations to some of the hotspots, to understand outdoor thermal comfort in Manila City better. In addition, the methods used in this study can also be used in other cities as well as municipalities that require assessment due to the presence of intra-urban heat islands.

**Author Contributions:** Conceptualization, M.A.P., M.C. and T.Y.; methodology, M.A.P., M.C. and T.Y.; software, M.A.P.; validation, M.A.P., M.C. and T.Y.; formal analysis, M.A.P., M.C. and T.Y.; investigation, M.A.P.; resources, M.A.P.; data curation, M.A.P.; writing—original draft preparation, M.A.P.; writing—review and editing, M.C. and T.Y.; visualization, M.A.P.; supervision, M.C. and T.Y.; project administration, M.A.P.; funding acquisition, M.A.P. and M.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded and supported by Adamson University, Kyushu Institute of Technology, and the Department of Science and Technology—Science Education Institution (DOST-SEI) through the STAMINA4Space Program of the University of the Philippines-Diliman.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors would like to thank Adamson University, Kyushu Institute of Technology, and the Department of Science and Technology—Science Education Institution (DOST-SEI) through the STAMINA4Space Program of the University of the Philippines-Diliman for supporting and funding this research. Moreover, we thank the following who one way or another helped in the realization of this work: Climate and Agrometeorological Data Section (CADS), Climatology and Agrometeorology Division (CAD), Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) for the weather data; Maria Fe B. Abalos, Knowledge Management and Communications Division, Information Technology and Dissemination Service, Philippine Statistics authority for Philippine statistical data including the population for Manila City; Julius M. Judan, SSED-Ground Receiving Station, DOST-Advanced Science and Technology Institute for giving access to remote sensing data and satellite images; Joven Javier, for assisting in communicating with the government agencies such as PAG-ASA and DOST-ASTI; Joseph Ronquillo and Anna May Ramos for the statistical analysis, Ronnie Serfa Juan and Evelyn Raguindin for revision comments; and BIRDS-4 satellite project members for the support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

