The effect of vegetation index on the land surface temperature in South Badung Regency, Bali Province

The land surface temperature (LST) is a crucial component of the earth’s energy balance system. The temperature differences between the earth’s surface and the atmosphere are reflected in LST. Conversion of land, including vegetated land, may result in changes to LST. Using the vegetation index approach—NDVI and EVI—this study seeks to ascertain how variations in vegetation density impact LST. Using Landsat 7 ETM+ satellite imagery from 2003 and Landsat 8 OLI-TIRS from 2015 and 2020, this research combines remote sensing technologies and GIS to get vegetation density and LST values, which were then subjected to field verification and spatiotemporal analysis. According to the study’s findings, variations in vegetation density and soil surface temperature have an inverse or opposing relationship. The study’s findings suggest that variations in vegetation density and soil surface temperature have an opposing or inverse connection. In South Badung Regency, places with low vegetation density vary more in proximity to metropolitan areas, resulting in higher soil surface temperatures. These findings suggest that several additional factors, including population density and size, land use, urban planning, rainfall, and season, influence variations in land surface temperature in South Badung Regency.


Inxtroduction
One of the critical hydro-meteorological variables in the earth's regional and global energy balance is the soil surface's temperature [8][16] [38].Land surface temperatures can predict hydrological processes, track droughts, evaluate the danger of wildfires, and track climate change [35] [38].Changes in the amount of vegetation cover are one of the elements that affect land surface temperature [22] [24].A hot pole will form, and the land surface temperature will rise due to the non-vegetated area absorbing more solar heat and reflecting it [24].
The LST may be impacted by the reduction in vegetation cover and density caused by converting green open space to a built-up area [8].Using sunlight for photosynthesis, vegetation on the earth's surface can reduce air temperature.Because of its ability to tolerate sunlight above the canopy, it can resist lower temperatures below it [15].Vegetation's capacity to absorb solar heat can lower local surface temperature (LST) and the microclimate [27].The temperature difference between the ground surface and the atmosphere is known as the land surface temperature or LST.Each location has distinct variations in land surface temperature, influenced by terrain, soil type, humidity, vegetation cover, and climatic conditions [22].Numerous techniques, including the split-window algorithm [9] [13], mono-window algorithm [29] [38], and brightness temperature [4] [22], have been used in land surface temperature research.
Vegetation and LST have a high correlation [3][4] [23] [33].According to Indrawati et al. (2020) [15], Nega et al. (2019) [23], and Sukristiyanti & Marganingrum (2008) [33], the land surface temperature is lower in areas with low plant cover and density and higher in areas with tall vegetation cover and density.According to that argument, the vegetation present in a region affects the land surface temperature [23], and plants can influence the temperature of the land surface.The Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI) can be used to measure the actual density of vegetation [28].
One measure of vegetation development and coverage frequently used to characterize the spatiotemporal features of land use and vegetation cover is the NDVI [1].It is sensitive to the amount of chlorophyll in plants [14].NDVI is a vegetation indicator that displays the degree of greenness in the plant cover, whether dense or sparse [15].Developed from NDVI, EVI is a vegetation index that can remove atmospheric background noise [8][18] [21].Compared to NDVI, EVI is more sensitive to changes in canopy type, structure, leaf area index (LAI), and areas with high biomass [14][18] [21][37].Certain vegetation classes can be extracted using the vegetation index to correct the effects of uneven lighting in a given location.
In this study, data on vegetation density and land surface temperature in South Badung Regency are obtained using remote sensing technologies.Landsat 7 ETM+ and Landsat 8 OLI TIRS satellite pictures from 2003, 2015, and 2020 can be used to obtain these data, which are then subjected to spatiotemporal analysis.The objectives of this study are to determine the variations in land surface temperature and vegetation density that occurred in the South Badung Regency in the years 2003, 2015, and 2020, as well as to examine the impact of those variations on land surface temperature.Research on land surface temperature (LST) is carried out to understand better how differences in vegetation cover and density could affect LST and assist people in protecting the environment.

Study Area
This research was conducted in South Badung Regency, Bali Province, which consists of Kuta Subdistrict, North Kuta Subdistrict, and South Kuta Subdistrict, with an area of 155,99 km 2 or 15.599 ha.Geographically, South Badung Regency is between 08 o 14'20" to 08 o 50'48" S and 115 o 05'00" to 115 o 26'16" E.

Data and Processing
Vegetation density is the independent variable in this study, whereas LST is the dependent variable.Landsat 7 ETM+ and Landsat 8 OLI-TIRS imagery were used in this study (Table 1).Landsat satellite imagery was selected based on its accessibility, medium resolution, multispectral capabilities, and lack of sophisticated processing.Since NDVI and EVI are the most widely used vegetation indices to detect changes in vegetation, they were selected [8] [21].Low topographic characteristics are considered a good fit for research regions using NDVI and EVI.In the meantime, since it is thought to be more superficial to employ in this investigation, the brightness temperature computation for LST is utilized.

NDVI
NDVI is a general measure of vegetation index because it can quickly identify the vegetation area using multispectral remote sensing data [26].The near-infrared (NIR) and red (R) bands of satellite images can be used by the NDVI to predict the quantity, quality, and development of vegetation.On Landsat 7 ETM+ and Landsat 8 OLI, the Red and NIR bands are positioned in different bands; on Landsat 7 ETM+, the Red band is in band 3, and on Landsat 8 OLI, it is in band 4. In Landsat OLI, the NIR is in band 5, and the red band is in band 4. In this experiment, the NDVI was ascertained using equation (1) [3][28].
The values of NDVI pixels range from -1 to +1; the more significant the NDVI value in a region, the higher the vegetation density.Conversely, values nearer 0 or -1 indicate low vegetation density or a body of water [3][23].

EVI
When determining vegetation activity, EVI offers advantages in reducing soil sensitivity, nonphotosynthetic vegetation, and atmospheric influence capabilities.Compared to NDVI, EVI is more sensitive to changes in canopy density and regions with high biomass [18].The blue (B), red (R), and near-infrared (NIR) bands were used in the EVI computation according to equation (2) [28]. (2)

LST
By translating digital numbers into spectral emission for each satellite image, LST data was produced using TOA Brightness Temperature on Landsat 7 ETM+ and Landsat 8 OLI Brightness temperature values [15].The method used to convert DN to spectral radiance on Landsat 7 ETM+ and Landsat 8 OLI differs.Equation (3) for Landsat 7 uses ETM+ band 6, and equation (4) for Landsat 8 uses OLI band 10 or 11 [22][26] [30].
Our spectral radiance is L. Maximum spectral radiance is denoted by Lmax.The minimum spectral radiance is Lmin.The digital number (DN) is represented by Band 6.For QCALmax, the highest pixel value is 255, while for QCALmin, the minimum pixel value is 1.
Al is the additive radiance from DN, Ml is the multiplicative radiance from DN, and QCAL is the DN (Band 10 or Band 11).Equation ( 5) calculates the surface temperature, and the previously acquired value of spectrum radiation is used to convert the spectral radiance value to brightness temperature [8][19] [17].
QCAL is the DN (Band 10 or Band 11), Al is the additive radiance from DN, and Ml is the multiplicative radiance from DN.
The surface temperature is determined by equation ( 5), and the brightness temperature is obtained by converting the spectral radiance value to the previously obtained spectrum radiation value [8][19] [17].

Accuracy Assessment
A confusion matrix was employed in this study's satellite picture accuracy test.By comparing the category lines and the relationship between the reference data from the classification results produced by the system and the correct classification results, the confusion matrix or error matrix is calculated [5].The accuracy of the producer, user, overall, and kappa may be determined using the error matrix.In 2003, 2015, and 2020, the accuracy of the classification of satellite images was verified using accuracy evaluation tools.Less than 40% indicates poor accuracy, 40-55% fair agreement, 55-70% good agreement, 70-85% excellent agreement, and > 85% excellent agreement are the findings of the Kappa classification method [13].

Data Analysis
The primary data comes from verifying vegetation density using the vegetation indices from satellite image processing (NDVI and EVI).To verify the location, a pre-selected sample point is used for observation.The satellite image is then interpreted using the observation location at the sample point of vegetation density.In the meantime, a digital anemometer is used to assess the outdoor temperature while conducting direct measurements in the field shortly before taking satellite photographs.
Next, a straightforward linear regression analysis was performed on the data.Equation ( 6) calculates simple regression, a commonly used statistical method to ascertain the relationship between vegetation density based on vegetation index and LST [34] [36].The temperature of the land surface, or Y, is the dependent variable in the equation; an is a constant number; b is the regression coefficient's value; and X is the independent variable or vegetation density.

Results and Discussion
There are four categories of vegetation density based on the outcomes of processing the vegetation indices: dense, medium, sparse, and non-vegetated.The classification was modified based on earlier research tailored to the study area [6][7][28.]Despite having the same classification, the NDVI and EVI vegetation indices have different values in each categorization class.The algorithms' variations account for the differences in values within each category.Although their methodologies differ, the two vegetation indices have been standardized and can be used for index comparison.In the studied area, variations in vegetation density also result in variations in LST.

Vegetation Density Change
Land surface temperature and vegetation density results from 2003, 2015, and 2020 demonstrate annual variations.Changes in extent and distribution in each sub-district of the South Badung Regency annually show variations in land surface temperature and vegetation density.Additional investigation into the impact of alterations in vegetation density on land surface temperatures within the South Badung Regency can be conducted based on the outcomes of these modifications.Variations in land use from one land use to another may result in changes in vegetation density.
Variations in land use and land cover can result in changes in vegetation density, as determined by the NDVI index (Table 2).The non-vegetated land area, including water bodies and undeveloped terrain, grew from 217.63 hectares in 2003 to 485.42 ha in 2015.The site not covered by vegetation shrank to 338.71 hectares in 2020.In the meantime, from 2003 to 2020, the density of sparse vegetation-such as highways, shops, and settlements-grew.The area covered by sparse vegetation increased from 1.078,30 ha in 2003 to 2.246,41 ha in 2020.Between 2003 and 2020, the median vegetation densitywhich includes fields and rice fields-grew.The area covered by medium vegetation density increased from 3.172,67 ha in 2003 to 6.398,43 ha in 2020.
In the meantime, between 2003 and 2020, the density of thick vegetation-such as plantations, bushes, and urban forests-decreased.In 2020, the area with high vegetation density 2003 decreased to 6,057.53 ha from 11,130.70 ha, or from 71% to 39% of the total area.The research area saw cloud cover in 2015, which may have affected the NDVI value due to the white spectrum released.As a result, the NDVI value in the cloud would have been lower in 2015, approaching 0 or -1 [32].Because it does not exhibit reflections or actual emissions from objects acquired by satellite images, the area covered by clouds in this study is not included in the examined area [31].however, between 2015 and 2020, the distribution of the vegetation declined.A rise in the density of medium and sparse vegetation replaced it.Dense vegetation was only seen in the South Kuta Sub-district in 2015 and 2020.Meanwhile, the center of the medium vegetation density is found in the North Kuta Sub-district, while the sparse vegetation density is in the Kuta Sub-district.This explains why it is thought to be closer to the core of South Badung.
Meanwhile, the spatial data on differences in vegetation density based on EVI are shown in figure 1(b).The image shows the South Badung Regency's territory in 2003, 2015, and 2020.A light green tint denotes the medium vegetation density.The medium vegetation density declined in 2015 and was replaced by sparse vegetation density (represented by dense vegetation distributed in South Kuta Subdistrict and the yellow color spreading throughout Kuta and North Kuta Sub-districts).2020 saw a decline in the medium vegetation density and a rise in lush greenery, but the distribution was limited to the South Kuta Sub-district.Regarding the non-vegetated lands, the distribution in the western Kuta Sub-district is nearly identical yearly.Although the number of sites decreased from 2003 to 2020, it is still higher than dense vegetation areas, which continue to increase yearly.Based on figure 1, it is known that the non-vegetated area in South Badung Regency is the area that has a minor area of the total area.By paying attention to the NDVI and EVI indices, it is known that the areas not vegetated from 2003 to 2020 were distributed in almost the exact location as in the western part of Kuta District, where the land use is the airport.At sparse vegetation density, the two vegetation indices show that the distribution is highest in Kuta Sub-district, known as the city.The medium vegetation density was spread in North Kuta and South Kuta Sub-districts but dominates the North Kuta Sub-district, which is still dominated by rice fields.Meanwhile, for dense vegetation, the NDVI index is inversely proportional to the EVI, wherein the NDVI index for the area of lush vegetation decreases every year, but the EVI index increases every year.On the other hand, the South Kuta Sub-district, where there are still a lot of woods and no changes to land use, is home to dense vegetation areas with NDVI and EVI indices virtually precisely.The vegetation density will predominate over sparse vegetation density the closer it is to the core of the South Badung Regency or Kuta Sub-district, which is the city center with the domination of land use in the form of towns and places of activity.

LST Change
The area and distribution of the land surface temperature in South Badung Regency similarly during the same period.The changes in land surface temperature in 2003, 2015, and 2020 are shown in  There is little vegetation in Kuta Subdistrict since residential land use and activities predominate there.At 6.41 km^2, Kuta District has a higher population density than North and South Kuta Districts.The ground surface temperature rises due to the airport's location in Kuta District's western section.The airport area primarily comprises paved roads and buildings with little greenery.This finding is consistent with other research that indicated low LST values in vegetated areas and near water bodies and high LST values in built-up areas or open, undeveloped terrain [2][10] [11] [20].An inverse relationship exists between the land surface temperature in the Kuta Sub-district and the land surface temperature in the North and South Kuta Districts.The land surface temperatures in North Kuta and South Kuta Districts are lower than those of Kuta Districts, consistent with the higher vegetation density in these districts.Particularly in South Kuta District, where lush vegetation predominates and lowers the land surface temperature, the vegetation density in North and South Kuta Districts is characterized by medium and thick vegetation.Most land use in the North Kuta District is still rice fields, categorized as having a medium vegetation density.Similarly, plantations and bushes  Meanwhile, in 2003, the rain was 131 mm.High rainfall influences the temperature of the land surface [11].The increased rain makes the water reserves used by plants to carry out the evapotranspiration process more and more, so it can decrease the land surface temperature [11].

Accuracy Assessment
Based on the findings of vegetation density field observations, the confusion matrix computation was performed.Verifying the vegetation density at each sample point enabled field observations with predetermined 32 NDVI and 30 EVI sample points.The accuracy test yielded an overall accuracy calculation score of 91% and a Kappa Accuracy of 87% (coefficient 0.87), utilizing a confusion matrix based on the findings of satellite imagery categorization and field observation data of NDVI vegetation density.In the meantime, the calculation value for overall accuracy for the field observations of EVI vegetation density is 90%, and the value for Kappa accuracy is 86% (coefficient 0.86).The research of Phompila et al. (2015) [25], which has a kappa accuracy result of 86%, is comparable to the EVI accuracy test results.According to the accuracy test findings, every NDVI and EVI vegetation index has a kappa accuracy value of > 85%.According to Hua and Ping (2018) [13], this accuracy number implies that the classification of satellite imagery in this investigation demonstrates full compatibility between pictures.

The Influence of Vegetation Indices on LST
A comparison between the maps of vegetation density and land surface temperature from 2003, 2015, and 2020 was done as part of the analysis of vegetation density and land surface temperature.Based on NDVI and EVI measurements, regression analysis was used to ascertain the association between plant density and land surface temperature.The study suggests a yearly negative association between the ESG score and the two vegetation indexes.In other words, the LST value will decrease by the value of the vegetation index's regression coefficient for every 1% increase in the index's worth, and vice versa.The equation y = -2.9296x+ 27.639, which depicted the relationship between LST and NDVI in 2003, indicates that an increase in vegetation density of 1% will result in a corresponding rise in LST of -2.9296.According to the 2015 LST-NDVI equation, the LST will climb by -7.1782 for every 1% increase in the vegetation density level.According to the 2003 LST-EVI relationship, an increase of 1% in vegetation density will result in a corresponding rise of -3.6232 in the LST value, in addition to the other years.Based on 62 preset sample points, Table 5 shows the correlation value between plant density (NDVI and EVI) and land surface temperature in South Badung Regency, which varies annually.The correlation coefficient (R2) data show little link between the vegetation index and LST in the South Badung Regency.The coefficient of determination will demonstrate the degree to which vegetation density affects LST.R2 is closer to 1 than it is.The more significant the difference in influence between the independent variable (vegetation density as determined by the NDVI and EVI indices) and the dependent variable (land surface temperature) [13] [34].
Additionally, the R2 value shows little link in the research area between vegetation density and ESG.On the other hand, an inversely proportionate relationship between the land surface temperature and the vegetation indices utilized is suggested by a negative (-) association between the vegetation index and LST.According to this study, the ground surface temperature rises, and the variation in low vegetation density is more significant the closer one is to a metropolis.

Conclusions
In South Badung Regency, there were variations in the density of vegetation and the temperature of the soil surface in 2003, 2015, and 2020.Each regression coefficient value has a negative sign (-), showing an inverse or opposite association between vegetation density and land surface temperature.Changes in plant density have an impact on land surface temperature.According to the regression value, the land surface temperature will change by the value of the vegetation density regression coefficient for every 1% variation in the value of vegetation density.The changes in low vegetation regions are broader the closer one gets to South Badung Regency's metropolitan districts, and as a result, the temperature at the ground surface rises.The R2 value, which is far from 1 each year based on the two vegetation indices, indicates that the impact of changes in vegetation density on ESG in the study area is not substantial.Variations in vegetation density do not fully explain differences in ESG within the research area.However, other factors, including season, land use, rainfall, urban structure, population size and activity, and land usage might also have an impact.

Figure 1 (
Figure1(a)  shows the spatial data of vegetation density variations according to NDVI.The graphic indicates a known annual decrease in the thick vegetation density, indicated by the dark green tint.Rich vegetation, denoted by the hue dark green, covered nearly the whole of South Badung Regency in 2003; however, between 2015 and 2020, the distribution of the vegetation declined.A rise in the density of medium and sparse vegetation replaced it.Dense vegetation was only seen in the South Kuta Sub-district in 2015 and 2020.Meanwhile, the center of the medium vegetation density is found in the North Kuta Sub-district, while the sparse vegetation density is in the Kuta Sub-district.This explains why it is thought to be closer to the core of South Badung.

Figure 1 .
Vegetation density based on a) NDVI and b) EVI in 2003, 2015, and 2020 Between 2003 and 2020, the non-vegetated area fell from 229,30 ha to 144,41 ha.The percentage of the site that has not changed, which is only 1% of the entire scope of South Badung Regency, indicates that this decline in non-vegetated regions did not become too large.The density of sparse vegetation declined in 2020 after significantly increasing in 2015.Although South Badung Regency's sparse vegetation density and EVI are consistent year-over-year, the area is marginally smaller in 2020 than in 2003 and 2015.Between 2003 and 2020, the medium vegetation density dropped from 7.047,21 ha to 5.648,23 ha.From 3.429,25 acres, dense vegetation has increased.Based on the EVI, the medium vegetation density is a vegetation density with a larger area each year than other vegetation densities.

Figure 2
Figure2provides spatial data on how the distribution of ground surface temperature in South Badung Regency changed between 2003 and 2020.Kuta District, seen by the deeper tint, experienced a more notable shift in soil surface temperature between 2003 and 2020 compared to other sub-districts in South Badung Regency.From 2003 to 2020, Kuta Subdistrict's vegetation density was comparatively low.There is little vegetation in Kuta Subdistrict since residential land use and activities predominate there.At 6.41 km^2, Kuta District has a higher population density than North and South Kuta Districts.The ground surface temperature rises due to the airport's location in Kuta District's western section.The airport area primarily comprises paved roads and buildings with little greenery.This finding is consistent with other research that indicated low LST values in vegetated areas and near water bodies and high LST values in built-up areas or open, undeveloped terrain[2][10][11][20].
.1088/1755-1315/1291/1/012024 8 with dense vegetation and areas with a medium level of vegetation dominate the land use in South Kuta District.Due to this land usage, the ground surface temperature in the Kuta subdistrict is higher than the ground surface temperature in the North and South Kuta subdistricts.

Figure 2 .
Figure 2. Land Surface Temperature Change of South Badung Regency in 2003 -2020 The average land surface temperature for 2003, 2015, and 2020 also shows changes in the land surface temperature in South Badung Regency.South Badung Regency's average land surface temperature was 25,52°C in 2003; it rose to 26,22°C in 2015.The regions and proportion of areas with ground surface temperatures between 24 and 28 degrees Celsius were more significant in 2015 than in 2003.In contrast, South Badung Regency's average land surface temperature dropped to 23,30°C in 2020, with temperatures between 21°C and 24°C predominating.The South Badung Regency experienced a lower land surface temperature in 2020 than prior years.That can happen because of the higher rainfall in 2020 compared to the earlier years, which is 213,8 mm from Ngurah Rai Station and 371,5 mm from Sanglah Station.In 2015, the rain in the South Badung Regency was 170,3 mm and 249,8 mm.

Table 1 .
Data Acquired

Table 2 .
Area and Percentage of Vegetation Density of South Badung Regency in 2003, 2015, and 2020 based on NDVI

Table 3 .
Area and Percentage of Vegetation Density of South Badung Regency in 2003, 2015, and 2020 based on EVI

Table 4 .
Between 2003 and 2015, the percentage of land surface areas with temperatures below 21°C dropped dramatically, from 1,298.65 ha (8.33%) to 50.13 ha (0.33%).The South Badung Regency's <21°C classification grew to 362.05 ha, or 2.32% of its total area, between 2015 and 2020.The Type 21-24°C area dropped from 1,464 ha in 2003 to 780.08 ha in 2015.2020 saw considerable growth in the land surface temperature classification of 21-24°C, reaching 10,976.1 hectares, or 70.36% of the total area of South Badung Regency.Areas with class 24-28°C land surface temperatures rose from 10,808.11ha in 2003 to 11,617.08 ha in 2015.Nevertheless, by 2020, it had drastically dropped to 4,255.26 ha or 27.28% of the area.Then, in 2003, 2,028.07ha were included in the classification of ground surface temperature > 28°C; by 2015, that area had grown to 2,7379.70 ha.Out of all the classes, the >28°C category in 2020 has the smallest area-just 5.98 ha, or around 0.04% of South Badung Regency's total area.

Table 4 .
Land Surface Temperature Change of South Badung Regency in 2003 -2020

Table 5 .
Results of vegetation indices and LST regression