Time series analysis of landsat data for urban heat island monitoring in Semarang City

The urban heat island (UHI) condition has received significant attention because it has the potential to negatively impact life quality and comfort, especially livability and level of comfort. Semarang City, the capital of Central Java, has been attracting many people to live and causing land use conversion to built-up areas. This study aims to observe the land surface temperature (LST) trend and to analyse the UHI spatiotemporal pattern in Semarang based on satellite time series data within 10 years. Land cover and LST were extracted from Landsat data of 2008, 2013, and 2019. Based on the results, it was observed that the vegetated area has reduced from 20.347 Ha to 7.241 Ha, while the average LST value has increased from 29.24°C to 31.31°C in 2008 and 2019 respectively. We also found that the UHI pattern, which was initially detected only in a few areas in 2008 and was focused in the city centre, has gradually spread to the east and south of Semarang in 2019. The result also shows the spatiotemporal trend of land cover conversion and UHI areas, which can be beneficial for mitigation planning in the future.


Introduction
Land use/cover changes (LULC) from vegetation to built-up area is one of the causes of increasing air temperature in urban areas.Changes in land use are closely related to the level of urbanization [1].One of the problems in urban areas related to LULC is the urban heat island (UHI).This UHI phenomenon happens when the air temperature in a densely built city is higher than the temperature around it, both in the village and on the outskirts of the city [2].Several factors affect the level of UHI, one of which is buildings that radiate heat faster than green fields or lakes [3].Urban areas usually have a more significant impact on the UHI phenomenon when compared to suburban areas.UHI indirectly affects the local climate and community activities, especially in urban areas [4].Other effects that can occur from UHI include decreasing the degree of public health and reducing air and water quality, especially in surface water [5].
Semarang is the capital city of Central Java Province.Semarang is the center of industry, urbanization, and government.The domination of built-up areas and the lack of vegetated areas is a contributing factors to the rising surface temperatures in the metropolitan area of Semarang.Many attempts have been made to remotely sense data for disasters [6], especially UHI detection and monitoring due to its advanced capability of sensing the thermal features of the earth's surface [7].The revisit time of remote sensing satellites also enables earth monitoring and change detection.In particular, the Landsat project is the longest constellation in remote sensing technology.The Landsat project has had a significant impact on understanding and addressing the UHI phenomenon in Semarang city.The project of remote sensing capabilities have provided invaluable insights into the spatial and temporal patterns of surface temperatures within the urban landscape.By capturing thermal infrared data, Landsat has enabled researchers to identify and analyze the intensity of UHI effects, where urban areas encounter higher temperatures than their surrounding rural areas [2].This information has been instrumental in quantifying the extent of UHI in Semarang, pinpointing the areas most affected by heat accumulation, and assessing the factors contributing to temperature variations.Landsat's continuous monitoring and long-term data archives have facilitated the identification of trends and changes in UHI patterns over time.This temporal analysis has been crucial in understanding the relationship between urban development, land use changes, and UHI intensity.Moreover, the availability of historical Landsat data has allowed researchers to explore the correlation between urban growth, vegetation cover, and surface temperature, providing valuable insights into potential mitigation strategies.
Armed with this knowledge, awareness of the environment is needed [8], particularly urban planners and policymakers in Semarang have been better equipped to implement evidence-based interventions to mitigate the adverse effects of UHI.This might include strategies like urban greenery expansion, heatreflective building materials, and optimized urban design.The Landsat project's contribution to UHI research in Semarang serves as a testament to the power of remote sensing in fostering sustainable urban development and improving the overall quality of life in rapidly growing cities.
Previous studies conducted by [9,10] revealed the relationship between green area cover and surface temperature levels.Research by [11] compared the conditions of three big cities across South East Asia which are Bangkok, Manila, and Jakarta.More analysis however is needed in other urban areas, especially in Semarang due to its distinct characteristic of coastal urban areas.Based on the described research background, the objectives of this study are to observe the land surface temperature (LST) trend and to analyze the UHI spatiotemporal pattern in Semarang based on satellite time series data within 10year periods.First, land surface temperature (LST) and land use/cover data were extracted from Landsat data.UHI was then determined in the study area.

Research areas and data used
Figure 1 shows the administrative map of Semarang City, the capital of Central Java Province.Geographically, it is located between 6 0 50' -7 0 10' South latitude and 109 0 35' -110 0 50' East longitude of 373.70 Km 2 .It has an area of Famous for its coastal tourism destinations, Semarang City has a coastline length of 13.6 Km.Administratively, Semarang City covers 16 sub-districts.It is the center of the Kedungsepur region which is a region covering Demak Regency, Ungaran Regency, Semarang Regency, and Purwodadi Regency.
In this study, a time series of Landsat data were used.Three Landsat data with the acquisition date of 2008 (Landsat 5), 2013 (Landsat 8), and 2019 (Landsat 8) were obtained for processing land cover, LST, and UHI.From the USGS data source, Landsat 8 data was obtained starting in 2013, so it requires Landsat 5 data which has been in orbit for a longer time and the data can be accessed.Landsat 5 and Landsat 8 both have thermal and infra-red bands that are used for LST analysis, apart from that it also depends on the condition of the percentage of cloud cover at the research location.Different Landsat data were used due to the different operational periods of Landsat satellites.Figure 2 shows the flow chart of this research.

Land cover from supervised classification
Land cover data were extracted from Landsat for the three observation periods.We used a superviseddigital classification technique called parallelepiped.It is a classification method that uses a simple decision rule to classify multispectral remote sensing data.Existing spectral values will be simplified and categorized according to the sample of land cover classes that have been created before.In the next step, an accuracy test was conducted to validate the interpreted land use cover from satellite images.We compared the classified maps with the high-resolution image provided by Google Earth.

Land surface temperature and UHI extraction
For the LST calculation, the spectral radiance values in the thermal band in Landsat data were converted to brightness temperature [12].In the calculation, the spectral radiance values in thermal band 6 were calculated in Landsat 5, while thermal band 10 was used from Landsat 8.The obtained temperature in Kelvin was then changed to a Celsius degree.Next, UHI was calculated using the following formula [13], where  refers to the mean LST alpha is the standard deviation.

Relationship between vegetation area and LST
Table 1 presents the average LST values corresponding to the vegetation total area in three years of the observation period.Based on Table 3  Table 2 shows the validation of classification results using supervised processing against conditions in the field in a particular year.Conditions in the field refer to the existing Semarang City base map from Google Earth data.The results show that from 30 sample points an accuracy of 90% was obtained in 2008 and an accuracy of 86% in 2013 and 2019.Therefore, the supervised classification has high accuracy regarding conditions in the field.
The relationship between the influence of green open space on UHI in Semarang City was analyzed using a simple regression calculation.The results of the analysis show that the R Square value is 0.9835.This shows that the magnitude of the influence of green open space (X) on UHI (Y) in Semarang City is 98.35%, while the remaining 1.65% is influenced by other variables outside the research.The significance value obtained was 0.019 (<0.5), which means that there is a significant influence of green openness on UHI in the city of Semarang.
LST values in 2008, 2013, and 2019 were presented in Figure 3 in Semarang.Figure 3 (a) shows that the average value of LST in 2008 was 29.24 0 C, with a minimum temperature found to be 20.18 0 C and a maximum temperature of 37.32 0 C.In Figure 3 (b), it was observed that the average value of LST in 2013 was higher (30.88 0 C), with a minimum temperature of 21.29 0 C and a maximum temperature of 37.83 0 C.In 2019, the average LST value was 31.31 0C, with a minimum temperature of 19.96 0 C and a maximum temperature of 39.84 0 C (Figure 3c).The most prominent changes occurred in the area of Semarang City in 2019.It can be seen that the dominant changes occurred in the western, northern, and eastern regions of Semarang City, which are the centers of government and economy.

Spatial and Temporal Pattern of Urban Heat Island
The distribution of UHI in Semarang for 2008, 2013, and 2019 are presented in Figure 3 a-c, respectively.Based on Figure 4 (a), it can be seen that UHI in 2008 was not too wide, only in a few areas such as Central Semarang, North Semarang, West Semarang, and a small part of the Banyumanik area.The non-UHI classification is mostly in the GunungPati area.The UHI classification < 2°C had a fairly large area spread within several areas such as Pedurungan, Candisari, Genuk and Gayamsari subdistricts.UHI classifications of 2-4°C and >4-6°C tended to be located in the central part of Semarang City and were located in only a few regions.

Figure 1 .
Figure 1.Administrative map of Semarang City

Figure 2 .
Figure 2. Flowchart of research , the distribution of vegetation area in Semarang City in 2019 has shrunk when compared to the size of vegetation land in Semarang City in 2008.In 2008, the total vegetation area was 20.347 Ha or equivalent to 54.43% of the Semarang City area, while in 2013, the vegetated area was only 0.344 Ha or equivalent to 27.67% of the Semarang City Area.Further, the vegetation area became only 7.241 Ha or equivalent to 19.37% of the Semarang City Area in 2019.The results show that there was a 13.106 Ha reduction of the vegetated area within a 10-year period, with an average annual decrease of1.191Ha.This means that the vegetated area in Semarang City in 2019 only covered a proportion of < 30% (19.37%) which is less than the required by the government.In summary, based on the results of the study, it is seen that the larger distribution of high LST in Semarang was correlated to the lower vegetation areas.

Figure 4 (
b) shows the UHI in 2013.It was revealed that the distribution of UHI was quite extensive and spread in several areas such as Central Semarang, North Semarang, West Semarang, Ngaliyan, and Candisari sub-districts.The non-UHI classifications were mostly in the Gunung Pati, Ngaliyan, Tugu, and Tembalang districts.Region of 0-2°C UHI had a fairly large area ranging from Pedurungan, Candisari, Genuk, Gayamsari, and Ngaliyan to Mijen sub-districts.UHI classifications of 2-4°C and 4-6°C were observed in the central part covering a large area of Semarang.UHI pattern in 2019 is shown in Figure 4 (c).According to the figure, the distribution of UHI was very wide and spread in several areas of Semarang in comparison to the two previous years.There was an increase in UHI, especially in the central area of Semarang City.Yet, some areas were still in the non-UHI classifications which were mostly in the GunungPati area, Tugu and Genuk sub-districts.The 0-2°C UHI classification had a fairly large spread within several areas such as Pedurungan, Candisari, Genuk, Gayamsari, Ngaliyan and Mijen sub-districts.UHI classifications of 2-4°C and 4-6°C were located in the central part of Semarang City.
The first generation of Landsat was launched in 1972 and it is still in active operation now with Landsat 9 in orbit.Besides the 1314 (2024) 012086 global coverage and extended period of observation, Landsat has thermal sensors which makes it suitable for surface temperature measurement and UHI monitoring.

Table 1 .
Comparison results of mean LST with vegetation land of Semarang City indicators.

Table 2 .
Results of the confusion matrix between the classified map and the base map.