Identification of lava presence on Anak Krakatau volcano using normalized hotspot index before a flank collapse in 2018

On December 22, 2018, a catastrophic tsunami struck Anyer Beach in Banten, Indonesia, resulting in numerous casualties and extensive damage to the region. A flank collapse in the southwest sector of Anak Krakatau primarily triggered this tsunami. Intense eruptions and subsequent lava discharges precipitated this collapse. Indonesia’s Vulcanology Geological Hazard Mitigation Center has reported heightened activity at this location since June 2018. Comprehensive field studies, bolstered by remote sensing data, have been pivotal in enhancing our understanding of the behavior of Anak Krakatau over recent years. This research strives to pinpoint lava’s presence in Anak Krakatau during its eruptions. The methodologies employed include the Normalized Hotspot Index (NHI) transformation index combined with the SWIR-1 and SWIR-2 channels of the Landsat 8 OLI imagery. The goal is to elucidate the intense lava flow patterns associated with the flank collapse. Our findings indicate that the NHI can detect elevated lava flow activity in Anak Krakatau. Within the 13 datasets from 2018, lava flow was discerned on seven occasions between July 29 and December 20, predominantly directed towards the southwest sector of the volcano.


Introduction
Anak Krakatau volcano, or "The Son of Krakatau" in the Indonesian language, is located in the Sunda Strait, administratively in the Province of South Lampung, Indonesia [1].Anak Krakatau complex consists of 4 islands: Sertung Island, Rakata Island, Panjang Island, and Anak Krakatau volcanic island.Anak Krakatau volcano was built on the caldera of Krakatau volcano that erupted in 1883 and was one of the deadliest volcanic events in history [2] [3].The paroxysmal eruption of Krakatau led to the collapse of Krakatau's edifice, triggered 20-meter tsunami waves, and caused approximately 36.417 casualties [4].
Anak Krakatau volcano emerged to the surface on December 29, 1927, from 180 meters underwater [1] and has erupted many times.On December 22, 2018, a massive eruption generated the flank collapse of Anak Krakatau volcano, led to a destructive tsunami that hit the coastal areas along the shoreline of Sunda Strait, resulting in 437 casualties, 31.943injuries, 10 still missing, and over 16.000 people displaced [5][6] [7].The eruption activity was high since June 2018 and continued until December 2018.The impact of the flank collapse on the volcano edifice can be seen in Figure 2, where most of the southwest flank disappeared (shown in a red circle).Also, the vegetation on the east flank had gone, and the shoreline changes were detected, increasing eastward and mainly decreasing westward.The Shoreline Changes ± morphology of Anak Krakatau volcano changed where the elevation before the flank collapse was 310 meters and after the collapse left to approximately 150 meters [8].
Remote sensing has ability to monitor phenomenon on the earth surface [9][10] [11], and off course to monitor lava occurances.Several studies have been conducted to monitor the activity of this volcano but few of them discuss the lava flow activity [12].Moreover, the flank collapse was preceded by the high intensity of the lava release.It is crucial to identify the lava activity to investigate the correlation between lava release and the potential of volcano flank collapse.Furthermore, it provides information to mitigate the primary as well as secondary disasters in the upcoming future.Most of the cascading hazard is underestimated [13] therefore the effectiveness of disaster management is highlighted because mass evacuation of volcanic eruption and its related-secondary hazard may lead to the chaotic events and enhance the risk to the community [14].
The problem that usually appears when it comes to the direct monitoring of volcanoes is the difficulty of reaching the area due to its potential hazard [15].On the other hand, remote sensing data can do continuous monitoring over Anak Krakatau volcano.Remote sensing technology has been widely used in volcano monitoring, such as combining SAR (synthetic aperture radar) and optical remote sensing data [12].Optical imagery can be applied to monitor volcanic activity, such as thermal anomalies and the temperature of pyroclastic flow [16][17].
Landsat-8 satellite carries the Operational Land Imager (OLI) and Thermal Infrared Sensors (TIRS) instruments [18].The OLI instrument measures the visible, near-infrared, and short-wave infrared.The specification of the satellite is listed in Table 1.Many studies have proposed the use of short-wavelength infrared on volcanic monitoring activity.The short-wavelength infrared channel has high sensitivity to hot objects, which can increase the SWIR radiance [19].The SWIR channel records the hot emitting object even at nighttime [20] and can measure the temperature above 150°C [21].Unlike the middle infrared (MIR) and thermal infrared (TIRS), which can detect the more excellent surface of the volcanic object, SWIR is specifically used to detect scorching surfaces such as incandescent lava.With that advantage, the SWIR channel can focus on monitoring the lava activity, such as the lava presence during a volcanic eruption.The performance of the SWIR channel can be increased by using index transformation as the Normalized Hotspot Index (NHI), which it uses to detect the presence of lava, combining the specific advantage of the SWIR channel in detecting the hot object on volcanic activity.The previous study shows that NHI has been tested using Sentinel-2, ASTER, Landsat-8, Himawari-8, and GOES-R, and shows the results that NHI can be used to detect the hot target, specifically lava presence [22][23] [24][25] [26].
This study aims to identify the presence of lava using the NHI index.Since the flank collapse was begun by high-intensity eruption followed by lava release activity, it is important to discover the linked relation between the two hazardous events.

Data Acquisition and Imagery Pre-processing
Landsat-8 OLI is downloaded from USGS (https://earthexplorer.usgs.gov/).The datasets range from January to December 2018 (Table 2).Only SWIR-1 (band 6) and SWIR-2 (band 7) channels were used in the processing to obtain the lava flow activity.The datasets were sorted manually by cloud coverage in the study area.The datasets containing clouds or plumes covering the crater area or the volcano edifice will not be used in the data processing.Subsequently, radiometric correction is needed in index transformation processing.The radiometric correction converts digital numbers from the raw data into top-of-atmosphere (ToA) spectral radiance.The formula for radiance correction is shown in Equation 1.
Lλ = MLQCal + AL (1) In Equation 1, Lλ is ToA spectral radiance (Watts/(m 2 srad µm)), ML is the radiance multi-band (can be found in metadata), AL is the radiance add band (can be found in metadata), and Qcal is the digital number of the band (the dataset).

Normalized Hotspot Index Processing
The hot object is emitted in the SWIR channel of the satellite sensor.Behalf of that, the SWIR channel in the satellite can be used to identify the thermal anomaly, specifically in volcano monitoring.Normalized Hotspot Index (NHI) is a multi-channel algorithm widely used to detect the thermal anomaly of volcano activity [26].Two equations express the formula of NHI (Equation 2& 3).
NHI values can vary from negative to positive values where the positive value (NHI<0) indicates the existence of lava and shows a moderate thermal anomaly [22][24], while the negative value (NHI<0) means the non-lava object.

Results and Discussion
The activity of lava release is increased, as shown in Figure 3.The positive value of the NHI index shows the lava presence of Anak Krakatau's volcano activity.The Volcanology Geological Hazard Mitigation Center (VGHMC) Indonesia observation reports showed that the eruption was detected on June 25, 2018.The Landsat-8 OLI imagery on June 11 did not show the lava presence activity.The activity of lava presence was continuously higher until December 20, but the NHI value on October 17 had dropped to 0.2 but still classified as lava presence.For some datasets, lava presence can not be detected because of the existence of clouds or plumes that cover the study area.
The NHI value is highly affected by the clouds or plumes, leading to the underestimation of volcanic thermal anomaly [22] [23], which is the lava presence in this case.In the 2018 monitoring, only 13 datasets are able to be used because clouds or plumes cover the rest of the datasets and cannot be used to identify the presence of lava.The visualization of lava presence in Anak Krakatau's volcano eruption can be seen in the following figure (Figure 4).The first lava presence in 2018 monitoring activity detected on July 29, 2018, mainly exists in the vent area of the volcano.The lava was continuously detected in the vent area on August 14, 2018.The following analyzed datasets imagery on August 30 and September 15 can be seen flowing to the volcano's Southeast flank.The biggest lava flow of all datasets was detected on October 1 when the lava flowed to the Southwest flank and reached the shore.The next lava flow occurred on October 17 in the vent area, but no lava was detected in November.The differences of lava form in the eruption period are mainly restrained by its viscosity include its temperature, chemical composition, gas content, and crystallinity.Eruption rate and ground slope are also predisposed the lava flow and how far it can travel [27].The lava was detected on December 20 on the vent.The eruption activity was high until the flank collapse occurred on December 22.
The eruptive phase of Anak Krakatau in June 2018 was recorded as the most intense since the monitoring began in 2000 [28] The previous study had stated the potential of flank collapse due to the high intensity of Anak Krakatau activity [29] [30].The lack of structural stability is one of the main factors of unstable edifice leading to flank collapse [8].The instability of the edifice can also be triggered by endogenetics, such as the lava release from the magma chamber or the addition of material to the surface that can lead to the oversteepening of the slope [31].
The previous event of Krakatau volcano's catastrophic eruption, which was caldera collapse formation, also shows similar behavior.Intensive eruption and the lava release were causing erosion on the diatreme, where it was dilated and left a bowl-indentation on the earth's surface [32].A similar theory was also proposed by Bemmelen in 1949, indicating that the intensive eruption urged the gas contained in magma out, leading to the dilated diatreme causing the magma chamber's emptiness [33].A high amount of eruption and lava released to the surface caused the instability of the magma roof and induced the collapse.Destabilization for years or even months can keep massive power and trigger unpredictable disasters like flank collapse.Based on past events, lava presence activity indicates its connection to the caldera formation's collapse and Anak Krakatau volcano's flank collapse.Therefore, it is crucial to keep monitoring the Anak Krakatau volcano to maintain the updated information and also increase a comprehensive study on the behavior of the Anak Krakatau volcano shown in lava presence activity.The high sensitivity of the NHI index adds the advantage of a wide range of time for monitoring activity.This means reducing the impact of the possible event to recur with an unknown probability of destructive potential.

2 Figure 1 .
Figure 1.(a) The complex of Anak Krakatau located in Sunda Strait between the island of Java and Sumatera [1], (b) The complex of Anak Krakatau consists of 4 islands and Anak Krakatau volcano is the center of this complex.

Figure 2 .
Figure 2. (a) Anak Krakatau edifice in 2018 before the flank collapse, (b) Anak Krakatau edifice after the flank collapse (red circle shows the southwest flank) shown in Planetscope imagery, (c) The shoreline changes before and after the flank collapse

Figure 4 .
Figure 4.The presence of lava flow detected by NHI using Landsat-8 OLI imagery

Table 2 .
List of datasets used in the processing