Seasonal Urban Heat Island Observation from Remote Sensing by Using Google Earth Engine Over New Capital (IKN) East Kalimantan, Indonesia

1. Introduction

The Indonesian government plans to relocate the Indonesian capital from Jakarta to Ibu Kota Nusantara (IKN) in Penajam Paser Utara (PPU) East Kalimantan starting in 2024. The urgency of relocating the capital is due to the objective condition of Jakarta which is no longer suitable as the capital. This is evident from "burden" Jakarta has to bear: 1) a population density of 16,704 people/km², while Indonesia's population density is only 141 people/km²; 2) the traffic congestion of Jakarta, which is the number 10 most congested city in the world in 2019, although in 2020 it is only 31^st out of 416 major cities in 57 countries (TomTom Traffic Index); 3) acute environmental and geological problems such as flooding that afflicts Jakarta every year and land subsidence that causes parts of Jakarta to be below sea level [1]. The location of the IKN is also said to be ideal because it is close to developed urban areas as buffer cities, namely Balikpapan, Samarinda, and Bontang. The IKN is also a national identity that is a benchmark for a smart, green, beautiful, and modern sustainable capital city with international standards.

The relocation of the capital will certainly entail massive construction works that will affect several sectors. This will greatly change the natural conditions in the surrounding area, affecting climate change and the environment. A strategic intervention program that can respond to the challenges of IKN development is the concept of Forest City and Climate Plus Village (Proklim+). To support the concept of IKN becoming a smart metropolis that guarantees the highest quality of life for its inhabitants, climate and environmental conditions must be maintained.

Anthropogenic activities (land use and land cover change/LULC) alter the surface radiation balance so that surface temperature increases. This is known as the urban heat island (UHI) phenomenon [2,3]. UHI is a phenomenon in which urban areas have higher temperatures than suburban and rural areas. The high land surface temperature (LST) in urban areas is a manifestation of UHI, which gradually decreases in vegetated rural areas [4,5,6]. The modification of LU by adding urban in vegetation areas using Weather Research Forecasting (WRF) will increase air temperature in Jakarta and IKN [7,8].

Research on UHI and land cover using Google Earth Engine (GEE) has been studied by remote sensing experts [9,10,11,12,13]. GEE is a platform that is a powerful application for using big data because it does not require local storage of imagery, but allows the analysis of the satellite imagery in cloud-based processing.

2. Materials and Methods

2.1. Study Area

The location of the study area is centered on the future IKN and the surrounding area as a buffer city is located between latitudes 1.56°S – 0.26°N and longitudes 115.6°E – 117.8°E (Figure 1). The IKN is located in Samboja and Sepaku districts, North Penajam Paser Regency (PPU), East Kalimantan province with a total area of 2560 km² which is four times larger than Jakarta (661.5 km²). There are four buffer cities for IKN, namely Bontang, Samarinda, Balikpapan, and Tenggarong. Samarinda, the capital of East Kalimantan will be the "heart" as the historical center of East Kalimantan in the renewable energy sector. Balikpapan, on the other hand, is the "muscle", that will act as the downstream oil and gas hub and logistics center for East Kalimantan. The site where the IKN will be built is currently planted with Acacia mangium and Eucalyptus sp. Which are used to produce paper raw materials.

East Kalimantan is located at an elevation of 0-6, 4.49 m above sea level, and has a tropical rainforest climate (Af classification according to Kopen and Geiger). East Kalimantan has an annual high temperature of 30.46 °C and an annual low temperature of 25.11 °C, the warmest month in October (30.96 °C) and the coldest month in July (24.79 °C), an average annual rainfall of 97.22 mm, and a humidity of 78.32% [14,15,16,17].

2.2. Data

This study uses the Terra Moderate Resolution Imaging Spectroradiometer (MODIS) LST and Emissivity daily (MOD11A1) version 6.1. This product provides daily LST per pixel with a spatial resolution of 1 km in a 1.200 by 1.200 km grid. The MOD11 Level 2 swath product derives the pixels LST [18].

LST Level 3 Daily is a daily LST product represented in an integrated sinusoidal projection (in V3) or sinusoidal projection (in V4). The MYD11A1 LST daily product is created from the MOD11_L2 product yields per day using scientific datasets (SDS), mapping all MOD11L2 product pixels to the grid/pixel in a sinusoidal or generated sinusoidal projection and averaging the values in each grid [19,20]. The data trimmed by the MODIS sensors cover the study area at latitude 1.56°S to 0.26°N and longitude 115.6°E to 117.8°E. The data consists of 354 columns x 138 rows (note: only 210 pixels in the 138^th row) corresponding to a total area of 48,708 pixels (km²) consisting of 36,435 km² of land and 12,273 km² of sea, the elevation following topography. The data period is from 1 January 2001 until 31 December 2020. The data format is HDF Network Common Data Form [21].

2.3. Methods

Processing MODIS data using Javascript in Google Earth Engine (GEE) platform. GEE is a cloud platform with huge data collections of remote sensing satellite imagery and geospatial data, where the process of remote sensing imagery acquisition and data processing is performed online using Javascript syntax language. Cloud computing at GEE enables faster and easier data acquisition and processing. Remote sensing image data is "called" according to the study area, processed as needed, and the processing results can be downloaded for further analysis, with a much smaller amount of data compared to conventional processing. The capability of GEE is very important at this time, especially for rapid data analysis and monitoring.

In this study, UHI was monitored based on seasonality, in December, January, February (DJF), March, April, May (MAM), June, July, August (JJA), and September, October, November (SON). Seasonal data were set as input data in GEE. The data were calculated for a 5-year average, namely, the average of 2001-2005, 2006-2010, 2011-2015, and 2016-2020, resulting in 16 files in GeoTIFF format. Figure 2 shows an example of the Javascript syntax for data processing in GEE in the period DJF 2001-2005.

In Figure 2 "Geometry 2" is the trimmed study area boundary, "Filter data" is the period from 1 January 2001 to 31 December 2005, and "calendar range" is month 12 (December) to 2 (February). When this syntax is executed, it will produce an output file named LSTDJF01-05 generated in GeoTIFF format. For other running periods and seasons, replace "filter date", "calendarRange", "dataset", and "description" according to the desired period and season.

The output of the running data from GEE (GeoTIFF format) is converted to Network Common Data Form (NetCDF or nc) format using QGIS software. NetCDF is a software library that set data and machine-independent data formats that support access, creation, and sharing of array-oriented scientific data [18]. QGIS is a freely available software and open-source application for cross-platform desktop geographic information systems (GIS) that supports editing, viewing, printing, and analyzing of geospatial data. The post-processing output data GEE runs (NetCDF data format) use Grids Analysis Display System (GrADS) (GrADS). GrADS is an easy-to-use, interactive desktop tool that is freely distributed and used to manipulate and visualize geoscience data [22].

3. Results

3.1. LST Seasonal Spatially analysis

3.1.1. LST Seasonal Spatially analysis

UHI propagation was observed through visualization of the output data spatially as shown in Figure 3. Seasonal LST DJF 2001-2005 is shown in Figure 3 (a), then LST DJF 2006-2006 in Figure (b) DJF 2006-2010, (c) DJF 2011-2015, (d) DJF 2016-2020, (e) MAM 2001-2005), (f) MAM 2006-2010, (g) MAM 2011-2015, (h) MAM 2016-2020, (i) JJA 2001-2005, (j) JJA 2006-2010, (k) JJA 2011-2015, (l) JJA 2016-2020, (m) SON 2001-2005, (n) SON 2006-2010), (o) SON 2011-2015, and (p) SON 2016-2020. Purple to red colors in the legend represent LST from low (≤22 OC) to high ((≥32 OC).

In Figure 3(a) (DJF 2001-2005), it can be seen that the areas of high temperature (≥32 ^OC) in urban areas shown in red, namely Bontang City (purple circle), Samarinda City (black circle), Tenggarong City (green circle), and Balikpapan City (blue circle). This high temperature reflects the occurrence of UHI. While the blue color (at LST around 22 – 25 ^OC) is found in suburban, rural, hilly/vegetated areas (including areas designed as IKN/red circle) and water bodies (rivers and lakes). High-temperature areas (UHI) are increasingly spreading, as shown in Figure 3(b) (DJF 2006-2010), Figure 3(c) (DJF 2011-2015), and Figure 3(d) (DJF 2015-2020). The next season is MAM, Figure 3(e) MAM 2001-2006 shows UHI in urban areas similar to the pattern in Figure 3(a) DJF 2001-2006, also Figure 3(f) is similar to Figure 3(b), 3 (g) is similar to 3(c), and 3(h) is similar to 3(d). The widest UHI occurred in 2015-2020, for both DJF and MAM.

The JJA season in Figure 3(i, j, k, l) shows a decrease in UHI compared to the previous season (MAM). If this is related to the monthly temperatures of East Kalimantan in Figure 4, it is reasonable. Figure 4 shows the average monthly temperature in East Kalimantan from observational data. The lowest temperature is in July and the highest is in October [12,13,14,15]. Moreover, the season SON in Figure 3(m,n,o,p) shows the highest UHI among the other seasons, even the season SON 2001-2005 (Figure 3(m)) has a higher temperature than the other seasons (DJF, MAM, JJA) in 2016-2020 (Figure 3(d, h, l)). This corresponds to Figure 4, where the highest temperatures in East Kalimantan occur in season SON season, and the lowest temperatures in season JJA.

3.1.2. LST Statistical analysis

3.1.2.1. LST (^OC) and area of LST (km²) seasonally

To find out how many temperature areas there are in each season and year, a statistical analysis of LST is performed as shown in Table 1 and Figure 4. Table 1 shows the region with temperatures from ≤ 23 to ≥ 34, and the graph is shown in Figure 5. The area of DJF 01-05 in Table 1 and Figure 4 is the quantitative value of DJF 2001-2005 LST spatial map in Figure 3(a), the area of DJF 06-10 is the quantitative value of 2006-2010 LST in Figure 3(b), and so on to SON 16-20 as the representation of LST SON 2016-2020 in Figure 3(p).

Table 1 and Figure 5 show that the areas with temperatures between 27 – 30 ^OC are larger than the other temperatures. Low temperatures (≤ 26 ^OC) decreased relatively strongly in the area from 2001-2005 to 2006-2010, 2011-2015, and 2016-2020 (DJF, MAM, JJA), while SON was dominated by high temperatures 28-31 ^OC. The temperature composition of DJF like the MAM, namely the temperature of 27 ^OC decreases in the area from 7954 km² (DJF01-05) to 6365 km² (DJF06-10), 5879 km² (DJF11-15), and 2992 km² (DJF16-20); 6981 km² (MAM01-05) to 5131 km² (MAM06-10), 4240 km² (MAM11-15), and 2117 km² (MAM16-20). Similarly, at a temperature of 28 ^OC, the area decreases from 11589 km² (DJF01-05) to 12394 km² (DJF06-10), 10592 km² (DJF11-15), and 6531 km² (DJF16-20); from 13962 km² (MAM01-05) to 13777 km² (MAM06-10), 10427 km² (MAM11-15), and 7224 km² (MAM16-20). The area of 29 ^OC is relatively the same, while the temperature of 30 ^OC is wider, from 2820 km² (DJF01-05) to 4491 km² (DJF06-10), 6317 km² (DJF11-15), and 9639 km² (DJF16-20); from 3248 km² (MAM01-05) to 4449 km² (MAM06-10), 6817 km² (MAM11-15), and 9791 km² (MAM16-20). Similarly, it is larger for a temperature of 31 ^OC, from 551 km² (DJF01-05) to 1338 km² (DJF06-10), 1589 km² (DJF11-15), and 3883 km² (DJF16-20); from 362 km² (MAM01-05) to 930 km² (MAM06-10), 1810 km² (MAM11-15), and 3914 km² (MAM16-20). For high temperatures, namely 31, 32, 33, ≥ 34 ^OC, which is an increasingly common UHI for both DJF and MAM. JJA is dominated by a temperature of 28 ^OC, which tends to decrease in the area from the 2001-2005 mean to 2016-2020. In contrast, it tends to increase when temperatures range from 29 to 34 ^OC. In the low temperature season SON (≤ 28), is less pronounced compared to the seasons DJF, MAM, and JJA. The SON season is dominated by moderate (29 – 30 ^OC) and high (31, 32, 33, ≥ 34 ^OC) temperatures.

3.1.2.2. LST area changes compared to 2001-2005

The LST area changes compared to 2001-2005 at the same temperature is calculated using Equation 1.

dSeason 06-10 = area (Season 2006-2010 – Season 2001-2005),

(1a)

dSeason 11-15 = area (Season 2011-2015 – Season 2001-2005),

(1b)

dSeason 16-20 = area (Season 2016-2020 – Season 2001-2005),

(1c)

where:

d = delta = area changes = (km²)

Season = area of DJF, MAM, JJA, and SON = (km²)

Equation 1 yields the area of LST changes, as shown in Table 2 and Figure 6.

From Table 2 and Figure 6, the area generally decreases at low temperatures (≤ 28 ^OC), whereas it increases at high temperatures (≥ 29 ^OC). LST ≤ 23, 24, 25 ^OC in all seasons and years, the area decreases between 2 and 383 km² except in JJA 06-10, when it increased by 33 km². LST 26 ^OC all seasons recorded a significant decrease in area (57 km² (SON 16-20) – 2295 km² (MAM 16-20)), except JJA 06-10, which expanded by 421 km². LST 27 ^OC all seasons experienced a significant decrease in area (1589 km² (DJF 06-10) to 5837 km² (MAM 16-20)), with the exception of JJA 06-10, which increased by 985 km² and SON (59 – 254 km²). LST 28 ^OC more areas experienced a decrease in area, namely DJF 11-15, DJF 16-20, MAM 11-15, MAM 16-20, JJA 06-10, and JJA 11-15, JJA 16-20, while those that expanded during the season DJF 06-10, MAM 06-10, SON 06-10, SON 11-15, and SON 16-20. LST 29 ^OC, most seasons experience area expansion, except DJF 06-10, DJF 11-15, SON 11-16, and SON 16-20 experience area decline. LST 30 ^OC, all seasons experience area expansion. LST 31 and 32 ^OC, all seasons experience an area expansion except JJA 06-10. LST 33 and ≥ 34 ^OC, all seasons experience an area expansion except SON 06-10.

3.2. LST Seasonal changes at the same location

3.2.1. Spatial analysis of LST Seasonal changes at the same location

To determine the changes of LST in the same location compared to 2001-2005, calculations are performed using Equation 2(a, b, c).

dLST season 06-10 = LST season 06-10 – LST season 01-15,

(2a)

dLST season 11-15 = LST season 11-15 – LST season 01-15,

(2b)

dLST season 16-20 = LST season 16-20 – LST season 01-15,

(2b)

where:

d = delta/LST changes = (^OC)

LST season = LST DJF, MAM, JJA, and SON = (^OC)

The results of the calculation of Equation 2 are shown spatially in Figure 7. Temperature changes during the DJF season are shown in Figure 7(a, b, c), MAM in Figure 7(d, e, f), JJA in Figure 7(g, h, i), and SON in Figure 7(j, k, l). Temperature changes in ≤ -2 ^OC are shown in purple and blue, -1 light blue, 0 ^OC light green, 1 ^OC yellow, 2 ^OC orange, and ≥ 3 ^OC magenta.

Figure 7a (DJF 01-05 to DJF 06-10) shows positive temperature changes of 1, 2, ≥ 3 ^OC (yellow, orange, and magenta) around Samarinda, Balikpapan, Kutai Kertanegara Regency (west of Balikpapan), and North Penajam Paser Regency (PPU) in southern Balikpapan. In Figures 7b (DFJ 01-05 to 11-15) and 7c (DFJ 01-05 to 16-20), it can be seen that the area is expanding. The pattern of temperature increase in MAM (Figure 7d,e,f) and JJA in Figure 7 (g,h,i) is similar to DJF, but the areas of high temperature increase are not as large as in the DJF season. The pattern of temperature changes in SON season has a larger temperature decrease (-2 ^OC, blue color) than other seasonal patterns, which can be seen around Kutai Kertanegara district.

3.2.2. Statistical analysis of LST Seasonal changes at the same location

To determine the number of areas with temperature changes, a quantitative analysis is performed as shown in Table 3 and Figure 8 (dLST positive=increase, 0=constant, negatif=decrease). The locations where the temperature changes from largest to smallest are 1 ^OC (29657 km², yellow color in Figure 8); 0 ^OC (9212 km², gray); 2 ^OC (6826 km², light blue); 3 ^OC (1484 km², yellow), -1 ^OC (1005 km², orange); ≥4 ^OC (459 km², dark blue); and ≤-2 ^OC (65 km², blue).

From 2001 to 2020, during the DJF season, sites that experienced temperature decreases (≤-2 and -1 ^OC) and those that remained constant (0 ^OC), tended to decrease in area, whereas temperature increases of 2, 3, and ≥4 ^OC led to expansion. In the MAM season, sites that experience a temperature decrease (-1 ^OC) and those that remain constant (0 ^OC) tend to decrease in area, while temperature increases of 2, 3, and ≥4 ^OC tended to result in expansion. During JJA season, sites that experience a temperature decrease in (-1 ^OC), as well as those that remain constant (0 ^OC), tend to decrease in area, while temperature increases of 1, 2, and 3 ^OC led to expansion. During the season SON, locations that experience a temperature decrease (≤-2 and -1 ^OC), as well as locations that remain constant (0 ^OC), tend to decrease in area, while temperature increases of 1, 2, 3, and ≥4 ^OC lead to expansion.

3.3. LST analysis in Balikpapan, Samarinda, Bontang, and IKN zero point

3.3.1. LST Seasonal in Balikpapan, Samarinda, Bontang and IKN

To observe the comparison between LST at the IKN zero point (0.97S, 116.7E) and the urban areas as IKN buffer cities, namely Balikpapan City (1.24S, 116.85E), Samarinda City (0.49S, 117.18E) and Bontang City (0.12N, 117.46E), seasonal LST analysis is conducted at the 4 observation points as shown in Table 4 and Figure 9. From Table 4 and Figure 9, it can be seen that the sequence of LST from the lowest to the highest value is respectively at the IKN zero point, Bontang City, Samarinda City, and Balikpapan City. The lowest value of LST at the IKN zero point was reached in the seasons JJA 2001-2005 and JJA 2006-2010, namely 26.45 ^OC, while the highest value was reached in the season SON 2016-2020. Bontang had the lowest LST (29.96 ^OC) in JJA 2006-2010, and the highest (33.09 ^OC) in the MAM 2016-2020. Samarinda had the lowest LST (30.64 ^OC) in the MAM 2001-2005, while the highest (35.53 ^OC) was in MAM 2016-2020. The lowest LST (32.18 ^OC) in Balikpapan occurred in the JJA 2001-2005 season and the highest (37.91 ^OC) in SON 2016-2020.

3.3.2. Cross-sectional analysis of the LST

Cross-section analysis was conducted to observe the UHI pattern in two dimensions. Cross-sections are conducted by drawing lines from north to south (latitudinal) and west to east (longitudinal) through the observation points, namely the cities of Balikpapan, Samarinda, Bontang, and the IKN zero point. In this study, the cases of DJF 2016-2020 season cases were selected as shown in Figure 10.

The results of cross-section data processing are presented graphically in Figure 11. For UHI analysis, LST data is needed in Table 4. The LST of DJF 2016-2020 season for Balikpapan City is 36.84 ^OC, Samarinda City 36.13 ^OC, Bontang City 32.3 ^OC, and IKN zero-point 28.82 ^OC. Figures 11(a) and (b) are latitudinal and longitudinal cross-sections of Balikpapan City. Here it can be seen that Balikpapan City (blue circle) has the highest temperature (36.84 ^OC) compared to the surrounding areas. This indicates that Balikpapan City is the center of UHI. The Samarinda City (black circle) in Figure 11 (c, d) also shows that UHI is present with a high temperature (36.13 ^OC) compared to the surrounding area. In Figure 11(e, f), the UHI is shown in Bontang City, with a high temperature of 32.3 ^OC. Unlike the graphs for the 3 previous cities, Figure 11(g, h) shows a lower temperature (28.82 ^OC) for the IKN zero point. This indicates that UHI has not occurred in the IKN area.

4. Discussions

The highest temperatures in East Kalimantan occur during the SON season, otherwise lowest temperature during JJA season. This is because East Kalimantan Province is one of 38 provinces in Indonesia which is located in the tropics and is crossed by the equator. With this geographical location, the sun will cross East Kalimantan twice a year, the first in February to April and the second in September to November [23,24], on 21 March and 23 September, the sun is directly above the equator.

For high temperatures, namely 31, 32, 33, ≥ 34 ^OC, which represents UHI which is increasingly widespread UHI both in the DJF and MAM seasons. JJA is dominated by a temperature of 28 ^OC which tends to decrease in the area from the average of 2001-2005 to 2016-2020. Conversely, for temperatures of 29 – 34 ^OC it tends to expand. In the low-temperature SON season (≤ 28) it is less extensive compared to the DJF, MAM, and JJA seasons. The SON season is dominated by moderate (29 – 30 ^OC) and high (31, 32, 33, ≥ 34 ^OC) temperatures. UHI in the study area has occurred in three buffer cities in East Kalimantan (Balikpapan City (36.84 ^OC), Samarinda City (36.13 ^OC), and Bontang City (32.3 ^OC)) respectively from high to low LST. In the cities, urbanization replaces vegetation that minimal amounts of water for evapotranspiration in urban areas[25]. In the urban area of Beijing, the measured soil heat flux is 50% of the total net radiation [26]. High LST in urban areas will increase the instability of the air on the asphalt and will directly heat the air and increase the air temperature [27]. UHI is located in the downtown area where there are many buildings. Green surfaces can reduce the effect of UHI, while built-up areas can strengthen UHI [28]. Reducing the latent heat flux in urban areas on impermeable surfaces will reduce the transfer of moisture from the soil to the atmosphere [29]. The seasonal trends of LST with higher LST are mainly concentrated in industrial and business districts [30]. The IKN zero point has the lowest temperature (28.82 ^OC) among the 3 buffer cities of IKN, which means UHI has not yet occurred, this is because during the observation period (2001-2020) IKN is still a vegetated mountain planted with Acacia mangium and Eucalyptus sp. Reduced evaporation will warm the ground and air surface, thereby increasing the sensible heat flux [29] and reducing the convenience of urban living. It is very important to maintain the condition of IKN which is still vegetated and UHI has not yet occurred. As a mitigation effort, for IKN development, it is necessary to choose building materials that can reduce the effects of UHI, including white roofs [29,31,32]

While the other 3 cities are areas that have long been developing. Bontang is an oriented city in the fields of industry, services, and trade, standing 3 large companies in the fields of natural gas, fertilizers, and coal. Samarinda is the capital city of East Kalimantan province with the largest population in East Kalimantan, with two of the busiest ports in East Kalimantan (Samarinda Port and Palaran Port). Balikpapan is the most populous city in East Kalimantan (2190 people/km²), as a center for business and industry (oil, gas, coal, and palm oil), there is an international airport and two seaports. In mountainous areas, the LST is lower than in urban areas because the surface incoming solar radiation will be weaker [33,34].

For future research, several factors related to UHI will be examined, namely Land Use and Land Cover Change (LULCC) [2,29,30], Normalized Difference Built-Up Index (NDBI) [35], Albedo, Normalized Difference Vegetation Index (NDVI) [36,37], anthropogenic heat and Comfortable Index [38,39,40], Normalized Difference Water Index (NDWI) [31,36], and Air Pollution [41,42,43].

5. Conclusions

From the results of the research and analysis above, discussing LST on a seasonal basis will make it easier to observe UHI dynamics. Using GEE as a web-based platform is very helpful for processing remote sensing data [9,10,11,12,13], in this study as a tool for estimating LST. UHI (high LST) propagation has occurred in the urban areas of Balikpapan, Samarinda, and Bontang, East Kalimantan. However, at the IKN zero point, there is no visible UHI, this should be maintained. A more in-depth study of the climate and environmental conditions of the IKN is needed to support the strategic program for the development of the IKN with the concept of Forest City and Climate Plus Village (Proklim+). In the end, IKN hopes to become a smart metropolis that guarantees the highest quality of life for its residents. The condition of IKN that has not yet occurred UHI needs to be maintained so that it supports sustainable development goals (SDGs) 11^th (sustainable cities and communities), 13^th (Climate Action), and 15^th (Life on Land). The results of this study are useful for policy makers and urban planner as one of the scientific considerations in the city development especially in the IKN.