The Impact of Environmental, Social, And Economic Factors Due to Anak Krakatau Volcano Tsunami Using Remote Sensing Technology

Tsunamis often occur in the territory of Indonesia because it is widely traversed by encounters between earth plates under the sea. Thus, if an earthquake occurs, it has the potential to generate a tsunami. The Indonesian region is also located in the ring of the fire area, and is the country with the highest number of active volcanoes with more than 127 events, that can encourage volcanic earthquakes. In addition, when volcanoes erupt offshore, the potential for a tsunami will be even greater because the expulsion of material can generate landslides toward the sea, causing a tsunami. The tsunami that happened in the eruption of Anak Krakatau on December 22, 2018, in the Sunda Strait, resulted in a tsunami in the Lampung and Banten areas. This study tries to compare environmental, social, and economic factors before and after the tsunami in the coastal areas of Lampung and Banten. The integration of remote sensing data will be carried out to determine environmental, social, and economic factors such as Land Cover, NDWI, NDVI, Night Light, and NO2 data. Random Forest Machine Learning will also be involved in shaping land cover models. Random Forest is often used because it has high accuracy. The results show a significant change in land cover in coastal areas and the impact of environmental, social, and economic factors. This study is expected to support the mitigation process in the event of a tsunami in the future. In addition, it is expected to be used by policymakers in planning development after the tsunami disaster.


Introduction
On 22 December 2018, Indonesia experienced a tsunami in the Sunda Strait area, which is the strait between Lampung and Banten Provinces (BNPB, 2018).Tsunamis can be caused by shallow earthquakes under the seabed, both tectonic and volcanic earthquakes, which can displace water volumes (Yeh et al., 2022).However, the cause of the tsunami in the Sunda Strait in 2018 was not caused by these two factors.An avalanche of volcanic material caused this tsunami from the eruption of Mount Anak Krakatau (ESDM, 2018).The height of the water waves from the 2018 South Sunda tsunami varied widely, from heights of 6.63 m (Syamsidik et al., 2019) to 12.58 m (Takabatake et al., 2019) in Pandeglang.
The tsunami in the Sunda Strait caused many fatalities.Around 437 people died, and hundreds of others were injured.One of the reasons for the high fatalities is the lack of early warning to the public (Syamsidik et al., 2020).Not only that, but this tsunami also caused a lot of damage to buildings (B.Syamsidik et al., 2020).Previous research conducted by Virtriana et al., (2023) showed that two years after the tsunami occurred in Pandeglang Regency (Banten), there are still traces of damage to buildings due to the tsunami, namely around 10% of buildings have not been repaired, and 19% of buildings are still badly damaged.It was also found that buildings in coastal areas and riverbanks near the estuary suffered more severe damage compared to other buildings (Virtriana et al., 2023).
Based on the previous explanation, a tsunami can harm the surrounding environment, both natural and man-made.In general, tsunamis can cause flooding, building damage, scour and soil instability, strong winds, and fire damage due to oil spills that occur as a result of tsunamis (Yeh et al., 2022).Several previous studies have discussed the impact of the tsunami on an area before and after the tsunami, especially using remote sensing technology and machine learning.The first is research by Sublime & Kalinicheva,( 2019) which conducted damage mapping after the tsunami using Deep Learning Remote Sensing with the Tohoku Tsunami case study in 2011.The second is research by Adriano et al., (2019) which integrates various remote sensing data with fusion techniques and machine learning to observe the damage after the earthquake and tsunami in Palu, Sulawesi, in 2018.The third is a study by Virtriana et al., (2023) which tries to observe the impact of the tsunami and focuses on damage to buildings before and after the tsunami using Random Forest with various types of predictors with the 2018 tsunami case study in the Sunda Strait.
Similarly to previous research, this study will also try to observe the tsunami's impact by integrating remote sensing technology and machine learning.However, this research will focus on two major aspects, namely environmental and socioeconomic characteristics before and after the tsunami.Environmental factors will consider land cover before and after the tsunami.Land cover will be processed through the random forest machine learning method.In addition, the environmental impact will also consider the vegetation index and air index.On the other hand, socioeconomic factors will consider air pollution and night light which are used as indicators of economic activities and activities of a community.This research aims to be used as a basis for tsunami disaster management in the future, especially in Indonesia.

Study Area
The study area of this research is the provinces of Banten and Lampung.The two regions are separated by the Sunda Strait, which is where Mount Anak Krakatau is located.According to BNPB data (2018), there are 14 sub-districts that cover coastal areas affected by the tsunami, where the fourteen sub-districts are located in Serang, Pandeglang and South Lampung Regencies.Therefore, this study will focus only on these three regencies that were affected by the tsunami.A more detailed look of the study areas can be seen in Figure 1.

Data
Data used in this study can be seen in Table 1.In general, this study uses five data consisting of administrative data, tsunami-effected region, Landsat-8, VIIRS Nightime, and Sentinel-5P NO2.The data landsat 8 that used in this study combine some images around 2018 until 2019 to identify the situation before tsunami and after tsunami.

Methods
The general methods in this study can be seen in Figure 2. The method can be divided into two sections.The first stage is the environmental factor data processing and the second stage is the socioeconomic data processing.

Environmental Factors
This study looks at changes in land cover that occur due to the tsunami.The changes will be analyzed using Landsat 8 satellite imagery.First, the land cover will be analyzed before the tsunami event using the random forest machine learning classification method.Furthermore, after obtaining changes in land cover before the tsunami, the land cover will be analyzed with the same classification method.After that, changes will be obtained as a result of the tsunami.The land cover classes used are built-up areas, water bodies, vacant lots, and vegetation.In the classification method, 100 points will be taken for each class, where splitting will be performed with 80% of the data used as training data and 20% as testing data.
In this research, the analysis for environmental factors is also represented by The Normalize Difference Vegetation Index (NDVI) and the Normalized Difference Water Index (NDWI).NDVI is an excellent tool for measuring the health and vigor of crop plants and detecting subtle changes in plant health, whereas NDWI is used to highlight open water features in a satellite image, allowing a water body to stand out against the soil and vegetation.With these indexes included in the environmental analysis for differences between after and before the tsunami, it is possible to predict how severely the tsunami affected vegetation and analyse the runoff of water to the mainland so that it changes other land covers into water bodies.Equations 1 and 2 show the NDVI (Weier and Herring, 2018) and NDWI (McFeeters, 1996) algorithms: In Landsat 8 imagery, the red reflectance values are obtained from Band 4 (0.630 -0.680 µm), and the near-infrared (NIR) reflectance values are obtained from Band 5 (0.845 -0.885 µm).The green reflectance values are obtained from Band 3 (0.525 -0.600 µm).

Socioeconomic Factors
Socioeconomic data will consider night light data (CSM, 2019) and NO2 air pollution data (ESA, 2018).Night light data is monthly averaged emission data using nighttime data from the Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) with units of nanoWatts/cm2/sr (CSM, 2019).While the NO2 data used is air pollution data in the troposphere (ESA, 2018).Night light and NO2 data can indicate human activity and indicate an urban area (Sakti et al., 2023).Sources of NO2 gas are more dominant in urban areas which can be caused by road activities (motorized vehicles), industry, mining, power plants, and land/forest fires (Sakti et al., 2023).Data processing will compare socioeconomic data at two different temporal points, before and after the tsunami.

Land Cover Change
This study has determined the changes in land cover due to the tsunami.Figure 3 shows land cover changes before and after the tsunami.Changes from built-up areas and vegetation to bare land often occur due to the tsunami.Some locations found a change of land cover from bare land to water and vice versa.The change of bare land to water could be caused by erosion as an effect of the tsunami.The tsunami's sedimentation can cause the phenomenon of land cover change from water to land.Figure 5 depicts the statistical comparison of the average NDVI value in the 14 tsunami-affected subdistricts.Except for the Labuan, the graph shows that the average NDVI value dropped in all subdistricts.With a lowered value of 0.076, Sumur and Katibung had the largest difference.The subdistricts with substantial decreases in NDVI are also found in three South Lampung Regency subdistricts, namely Rajabasa, Kalianda, and Sidomulyo, with values of 0.071, 0.073, and 0.072, respectively.This suggests that the tsunami impacted the destruction of vegetation in the area.However, in Labuan, the NDVI value appeared to have grown following the tsunami.This is because, first, the Labuan region is smaller in comparison to other sub-districts, while the NDVI value before the tsunami was still somewhat higher than Pagelaran, despite Pagelaran having a larger area.The second possibility is that the tsunami only harmed a limited area of vegetation near the coast, hence the increase in average NDVI occurred in other green areas in Labuan that previously had a relatively high value.This study also makes use of Normalized Difference Water Index (NDWI) observations to assess the environmental impact of the tsunami.Figure 6 depicts the results of the NDWI observations, which show that the index range between before and after the tsunami has increased.The maximum average NDWI score in South Lampung, Serang, and Pandeglang Regencies before the tsunami was 0.689.However, it increased to 0.735 after the tsunami.Furthermore, based on the affected districts, the NDWI value before the tsunami was as follows: South Lampung = 0.321, Serang = 0.376, and Pandeglang = 0.282.Meanwhile, the average NDWI values in Serang, South Lampung, and Pandeglang Regencies after the tsunami were 0.351, 0.301, and 0.319, respectively.According to these results, only Serang had a decrease in the water content index in vegetation, while the other two had a rise.This can be thought to be due to the rainfall factor, which varies by location and thus affects the water content.The average NDWI value was analysed at the regional administration level in more detail; in this case, for the 14 districts affected by the tsunami, it can be seen in Figure 7 that 9 districts experienced an increase in the average water content index.Meanwhile, the other 5 districts experienced a decrease.These districts are Anyar District in Serang Regencies, as well as Rajabasa, Kalianda, Sidomulyo, and Katibung Districts in South Lampung Regencies.Then, all affected districts in Pandeglang experienced an average increase in NDWI after the tsunami.

Air Pollution
The results of air pollution observations before and after the tsunami can be seen in Figure 6.Based on the results, it can be interpreted that the tsunami affected NO2 patterns before and after the incident.Before the tsunami, the highest average value in the study area was 148.706 μmol/m 2 .Meanwhile, before the tsunami, the highest average value decreased to 76.761 μmol/m2.This indicates that there is a decrease in human activity, especially in motorized vehicles.
This can also be caused because the affected coastal area is also a strategic area for tourism.So that when after the tsunami occurred, the region experienced a temporary stop in its tourism sector and experienced a decrease in air pollution gases.If we look at the air pollution before the tsunami based on the affected districts, South Lampung (Figure 8A), Serang (Figure 8B were 16 (1 -51) μmol/m2, 31 (10 -77) μmol/m2, and 15 (0 -39) µmol/m2.NO2 gas in the Pandeglang area experienced an increase, which could be caused by human activities in areas far from the coast.Thus, this increase was likely not an impact of the tsunami.Based on Figure 9, seven sub-districts experienced a decrease in NO2 gas after the tsunami occurred.The districts are Anyar, Cinangka, Rajabasa, Kalianda, Labuan and Pagelaran.On the other hand, some sub-districts experienced an increase in NO2 gas during the post-tsunami period, namely Sidomulyo, Katibung, Carita, Cigeulis, Cimanggu, Panimbang, Patia, and Sumur sub-districts.Either human transportation activities caused this increase in NO2 gas after the tsunami, or there is a bias from areas far from the coast that do not have a path from the tsunami.

Night light
Night light indicates human activity at night, which can be marked by light at that location.The night light can also identify the state of an area's economy.If a site has a high night light, the area has a high economic level, such as in a metropolitan city.Conversely, if an area has a low night light, then the area experiences a low financial status.
Based on observations of night light (Figure 7), before the tsunami, the average value of night light in the three districts was 149.775 nanowatts/cm 2 /sr and decreased after the tsunami to 88.905 nanowatts/cm 2 /sr.When viewed from the affected districts, South Lampung (Figure 10A), Serang District (Figure 10B), and Pandeglang (Figure 10C), respectively, have average values of light gas at night before the tsunami of 1.16 (0.08 -34.31) μmol/ m 2 , 2.21 (0.14 -149.77)μmol/m 2 , and 0.43 (0.01 -10.50) μmol/m 2 .Meanwhile, after the tsunami, the average values of night light in Serang, South Lampung, and Pandeglang Regencies were 1.06 (0 -26.89) μmol/m2, 3.15 (0 -88.91) μmol/m 2 , and 0.45 (0 -25.51) µmol/m 2 .Similar to NO2 gas, the night light in the Pandeglang area tends to increase compared to the other two cities, which had decreased values after the tsunami.This can be caused by an increase in night light in areas quite far from the coast (see Figure 10C), so it can be concluded that the rise in night light was not caused by a tsunami.

Land Cover Validation
Land cover classification before and after the tsunami event is calculated as the accuracy value of the land cover model.Table 2 and Table 3 are confusion matrices for validation of land cover models made using the random forest method before and after the tsunami.The model that has been made appears to be sufficiently accurate, with an overall accuracy value, at the time before the tsunami reaching 91.089% with a kappa value of 88.083%.The land cover model after the tsunami has an overall accuracy value of 91.176% with a kappa value of 88.144%.However, in modeling land cover with Landsat 8, which has a resolution of 30 meters, there are some shortcomings because it can only detect the affected area, not specific to the object of damage.In addition, the area around the Sunda Strait has quite a lot of cloud cover on each imagery, so it is necessary to carry out cloud masking techniques.The cloud masking technique followed by filling empty images with other images is one solution to eliminate the cloud effect at the time of classification.Although in some locations, a thin layer of cloud was still not successfully removed during cloud masking.

Conclussion
This study identifies changes after the tsunami by considering environmental factors based on land cover conditions, NDVI, NDWI, and socioeconomic factors represented by night light and NO2.The results of identifying land cover changes using a random forest classification show changes in coastal areas affected by the tsunami.Some locations show a change from vegetation to open land and built-up areas to open land.The overall accuracy value of the land cover model before and after the tsunami was compared with a reference of 91%.Based on the results of NDVI calculations, it can be concluded that there is a decrease in the value of NDVI in the affected area, where the highest average value before the tsunami was 0.458 to 0.386 after the tsunami.One of the causes of the decline in the value of NDVI is the change in vegetation land cover to vacant land.The value of NDWI in the affected areas has increased on average, directly proportional to the change in land cover to the waters due to the tsunami (land erosion).Based on socioeconomic factors, air pollution decreases, indicating a decrease in human activity, especially motor vehicles, due to the tsunami.The same is shown by the night light in the affected areas, where the index after the tsunami is smaller than before the tsunami, which shows a decrease in human activity. 3

Figure
Figure 1.Study Area

Figure 3 .
Figure 3. Land Cover Change3.1.2.NDVIThe impact observation of the tsunami disaster on the environment is analysed through the calculation of the vegetation index (NDVI).The index calculation results are shown in Figure4, where (a) is the appearance of the vegetation index before the tsunami and (b) after the tsunami.It can be seen in the figure that there has been a decrease in NDVI, from the highest average value of 0.458 before the tsunami to 0.386 after the tsunami.The NDVI values before the tsunami, when viewed based on the affected districts, were as follows: South Lampung = 0.158, Serang = 0.107, and Pandeglang Regencies = 0.171.Meanwhile, after the tsunami, the average NDVI index values in Serang, South Lampung, and Pandeglang Regencies were 0.097, 0.081, and 0.131, respectively.

Figure 5 .
Figure 5.Comparison of NDVI before and after the tsunami in the affected sub-districts

Figure 8 .
Figure 8. NO2 air pollution before (a) and after (b) the tsunami.

Figure 9 .
Figure 9.Comparison of NO2 Air Pollution before and after the tsunami in the affected sub-districts

Figure 10 .
Figure 10.Night light before (a) and after (b) the tsunami.

Figure 11
Figure11states the average value of night light in each sub-district in the affected area.Of the fourteen sub-districts observed, nine sub-districts experienced a decrease in the value of night light from before to after the tsunami.The districts are Cinangka, Rajabasa, Katibung, Carita, Cigeulis, Cimanggu, Labuan, Panimbang and Sumur.However, five sub-districts have not experienced a reduction in night light, including Anyar, Kalianda, Sidomulyo, Pangelaran, and Patia, which can be caused by an increase in night light far from the coast.Hence, it is not an impact of the tsunami.

Figure 11 .
Figure 11.Comparison of Night Light before and after the tsunami in the affected sub-districts

Table 2 .
Validation Matrix Land Cover Classification Before Tsunami

Table 3 .
Validation Matrix Land Cover Classification After Tsunami