Land Cover Changes Based on Cellular Automata for Land Surface Temperature in Semarang Regency

Land cover changes based on cellular automata for surface temperature in Semarang Regency has increased significantly due to the continuous rise in its population. Therefore, this study aims to identify, analyze and predict multitemporal land cover changes and surface temperature distribution in 2028. Data on the land cover map were obtained from Landsat 7 and 8 based on supervised classification, while Land Surface Temperature (LST) was calculated from its thermal bands. The collected data were analyzed for accuracy through observation, while Cellular Automata Markov Chain was used to predict the associated changes in 2028. The result showed that there are 4 land cover maps with 5-year intervals from 2003 to 2018 at an accuracy of more than 85%. Furthermore, the existing land covers were dominated by forest with decreasing trend, while the built-up area continuously increased. The existing Land surface temperature range from 20.6°C to 36.6°C, at an average of 28.2°C and a yearly increase of 0.07°C. The temperature changes are positively correlated with the occurrence of land conversion. Land cover predictions for 2028 show similar forest dominance, with a 23,4% built-up area at a surface temperature of 28.9°C.


Introduction
According to the United Nations (2018), population increase is a global problem experienced in every country, with 55% of humans presently living in urban or regional areas, likely increasing by 68% in 2050. These changes tend to affect both local and global climate components, such as the land surface temperature (LST). For example, in Nigeria, there was an increase of 19,166.13 ha in urban built areas from 2002 to 2013, with a rise in LST by 6 °C (Igun & Williams., 2018).
Fahrudin Hanafi et al. / Geosfera Indonesia 6 (3), 2021, 301-318 Study carried out on the land cover change in Semarang regency is common for land suitability, flood (Susanti et al., 2012), landslide, sedimentation (Apriliyana, 2015), carbon stock, and spatial planning review (Pangi et al., 2017). These studies were specifically related to land and averaged surface temperatures (Kalinda & Bandi, 2018). Therefore, this study aims to model land cover changes based on raster data using cellular automata related to its surface temperatures in the future. In addition, it also intends to (1) analyze the surface temperature distribution and land cover changes of Semarang Regency in 2003Regency in , 2008Regency in , 2013Regency in , and 2018 evaluate the relationship between land cover changes and LST, and (3) investigate the distribution of land cover for the following 10 years.

Methods
This field survey was conducted in Semarang Regency, Central Java Province, from April to June 2019. The area was considered due to the record time of the imagery data input.
Furthermore, simulation data input only requires 2 land cover imageries, the initial and step year. However, for the sake of detailed information, this study used those acquired in 2003 (initial), 2008 (step 1), 2013 (Step 2) and 2018 (Step 3), which was compared using population growth and space needed, such as the assumption based on consistent population per built area on initial, and each step. The satellite image data used are (1)  Accuracy is determined using a confusion matrix involving the consideration of omission and commission. Overall, it indicates the probability that a pixel belongs to a certain class and its representation in the field (Lillesand et al., 2004).
Land cover prediction in 2028 was made with Selva's version of Idrisi software in accordance with the Markov Chain method based on Cellular Automata. Meanwhile, Markov Chain is used to analyze 2 land cover data realized in different years, namely past (2008) (1982) and Amiri et al. (2009). It is also used to determine an accurate brightness temperature of 8-14 чm wave (Artis & Carnahan, 1982), while this study utilized bands 6 (10.4 to 12.5 чm) and 10 (10.60 to11.19 чm).
The analysis technique is used to determine the effect of each land cover type on LST changes by comparing the t-count with the t-table. The t-test is carried out using a simple linear regression equation while the t-table size is calculated with the formula t (a/2, n-2) = t (0.05/2, 116-2) = t (0.025, 114) = 1.98099, based on the following criteria: (1) assuming the significance value is <0.05 and t-count> t-table, it means that there is an impact, and (2) assuming the significance value is > 0.05 and t-count <t-table, meaning there is no impact.

Land Cover Changes
Image classification performed with the supervised method produces land cover maps for 2003, 2008, 2013 and 2018. Confusion matrix analysis on Table 1 shows an accuracy of 91.38%, 92.24%, 88.79% and 86.21 %. According to Fariz (2016), misclassification is not only influenced by differences in recorded time as well as the misidentification of objects from residential land to the forest because tall trees cover their appearance. The accuracy test results obtained annually exceed 85%, therefore, the data is feasible and needs to be used for further analysis (Susanti et al., 2012). The results show that the dominated land covers are forest and agricultural areas because the Semarang Regency is near mountains and fertile, as shown in Figure 2. Table 2 shows the increased number of forests recorded in 2013, possible because "sengon" and teak plantations starts with land clearing and an approximately 5 years harvest period (Nuroniah & Putri, 2013) and 15 to 20 years for teak (Pudjiono, 2014). This result also corresponds with the Semarang Regency data by BPS (2017)  Land cover simulation processed with Markov Chain on Cellular automata with population growth as a control. Furthermore, increase in population and built areas are linearly correlated to each step year to average growth. Therefore, the acquired data serves as an area of increased control when cellular automata are executed. The cell movement is constrained by slope, existent built area, road and river networks. Land cover changes are influenced by several attributes, from macro (policy) to micro (regional complex) factors ( Figure 2). In the regional complex, Semarang Regency develops and changes the influence of roads, slopes, demographics, and land availability.  Land cover simulation results in 2028 show an increase in built areas, a decrease in forest (6.19%), shrubs (7.90%), agricultural (

Imagery Surface Temperature Vs Survey
The field survey led to the fluctuation of LST data every hour from 09:00 to 15:00 for different land covers, as shown in Figure 5. This is used to ascertain the response of each land cover type to solar radiation. Besides, each has a different response rate, although they generally tend to reach the peak temperature at 13:00. It is clear that the built area has the highest LST, reaching 3.6 ° C, and decreases gradually after 13:00. Apart from land cover information, satellite imagery is also used to extract LST. This study produced 4 LST According to the R-value, the relationship between LST imagery and field survey is positively correlated (Fu & Weng, 2016). Figure 6 shows the image temperature is high, as well as the field, however, the coefficient is different due to dissimilarities in weather conditions. The results of the LST obtained using satellite imagery shows that the maximum and minimum temperatures reached were 37.6 ° C, and 20.6 ° C, respectively. The average LST and standard deviation values are 28.2 ° C and 3.39 ° C. Interestingly, the LST temperature from 2003 to 2018 tended to increase, while a decrease was detected in the data acquired in 2008. The decline in land surface temperature is influenced by low solar radiation of approximately 425.55 watts/m 2 , accumulated over the years. This is caused by the position of the earth's aphelion to the sun, which usually occurs in June. The Australian Cold Munson phenomenon also affects the deviation, which blows cold and dry air from the Indonesian coast. It also causes upwelling followed by a decrease of not less than 3° C in the South Java Sea, besides this usually occurs from July to August (Sukresno et al., 2018).
The high-temperature area is regularly experienced in regional centers, as shown in Figure 6. This shows that LST is concentrated due to the urbanization influence caused by land cover changes (Peng Fu, 2016). However, it is different from the phenomenon that occurred in 2008 due to the concentration of newly built areas in Bancak District which has varying temperatures, depending on the various types (Pal & Ziaul, 2017).

Surface Temperature Vs Land Cover Changes
To compare continuous data from LST and land cover changes (ordinal), this study collected whole coverage and point samples using grid size as the location. The LST was generated as positive or negative changes compared to the land cover, which was converted to non-changes data. The results of correlation analysis (Figure 7) carried out on the land cover and LST from 2003 to 2018 was used to determine the strong relationship between the coefficient of determination (R 2 : 0.430) and the correlation coefficient (R: 0.s655) (Sarwono, 2006). Meanwhile, from the t-test on each land cover, it is evident that agricultural areas, forests, and shrubs have a dominant influence (Table 4). Land cover is associated with temperature changes, it is evident that the dominant one is the conversion of forests to agricultural areas with a temperature (averaged area) increase of 0.31 ° C/ha, as shown in Figure 8 and Table 5. This led to an increase in the number of farmers from 144,369 (2013) to 167,044 (2018) in Semarang Regency (Yuliati, 2018). In addition, this condition was also experienced at the national level from 2000 to 2005, where plantations and agricultural areas were the dominant land cover types that replaced 44% of forests. The increase in average surface temperature of regional LST at the study site was raised to relatively 1.06 ° C with an average increase per decade of 0.7 ° C. Lin & Yu (2005) classified the LST increase in average surface temperature as being greater than the value obtained in Beijing and approximately within a range of 0.25 to 0.31 ° C. Based on this approach, it was estimated that the average LST in 2028 is expected to reach 28,9 ° C.  Based on the map, the western and southern parts of the study location comprised mostly of andosols associated with brown and dark red Mediterranean lithosol, regosol, and brown latosol. Furthermore, soil type, texture and color affects the LST distribution. One of the main ingredients of andosol soil, which is usually in the form of sand,is its ability to absorb high heat as well as to withstand low soil moisture within the range of 4 to 70% (Bavel et al., 1972). Meanwhile, hydromorphous alluvial soil types dominate the center, with clay as the parent material. It is also able to properly absorb a heat capacity of 0.279 calg-1 °C-1, although it is lower compared to sand (Lal & Shukla, 2004).
Vegetation is also considered to influence LST variation due to height, canopy character, and location of growth tendency. In accordance with the BPS data, cloves, rubber, coffee, sengon, teak, and mostly rice fields and shrubs are cultivated in the eastern part of Semarang Regency. Meanwhile, pine trees, fruits, bamboo, rubber, rice fields, and shrubs are cultivated in this regency at dense temperature fluctuations with high humidity and wind speed of 0.6 m/s.Conversely, the cone canopy has a speed of 0.4 m/s and a higher temperature (Fadlurrahman, 2018).

Conclusion
The dominant land cover in Semarang Regency from 2003 to 2018 was a forest with the tendency to be converted into agricultural areas. Land cover predictions for 2028 also show a similar change pattern. Furthermore, these changes had a positive relationship with LST and a significant effect with a correlation coefficient (R) of 0.655 (strong). Forest, agricultural areas, open fields, shrubs and built-up areas affect 51.1%, 56.6%, 33.1%, 43.3%, and 28.5% on LST. Land cover simulation shows that the built area is bound to increase by 18.56%, indicating dense growth in Tengaran, East Ungaran, Bank, and Kaliwungu Districts.
Based on calculations, this caused an average increase of 28.9 ° C at a rate of 0.07 ° C/year.

Conflict of interest
The authors declare that there is no conflict of interest with any financial, personal, other people or organizations related to the material in this study.