Damage Proxy Map (DPM) of the 2016 Gyeongju and 2017 Pohang Earthquakes Using Sentinel-1 Imagery

Abstract

The ML 5.8 earthquake shocked Gyeongju, Korea, at 11:32:55 UTC on September 12, 2016. One year later, on the afternoon of November 15, 2017, the ML 5.4 earthquake occurred in Pohang, South Korea. The earthquakes injured many residents, damaged buildings, and affected the economy of Gyeongju and Pohang. The damage proxy maps (DPMs) were generated from Sentinel-1 synthetic aperture radar (SAR) imagery by comparing pre- and co-events interferometric coherences to identify anomalous changes that indicate damaged by the earthquakes. DPMs manage to detect coherence loss in residential and commercial areas in both Gyeongju and Pohang earthquakes. We found that our results show a good correlation with the Korea Meteorological Administration (KMA) report with Modified Mercalli Intensity (MMI) scale values of more than VII (seven). The color scale of Sentinel-1 DPMs indicates an increasingly significant change in the area covered by the pixel, delineating collapsed walls and roofs from the official report. The resulting maps can be used to assess the distribution of seismic damage after the Gyeongju and Pohang earthquakes and can also be used as inventory data of damaged buildings to map seismic vulnerability using machine learning in Gyeongju or Pohang.

1. Introduction

An M_L5.8 earthquake shocked Gyeongju at 11:32:55 UTC on September 12, 2016. The epicenter was 8.7 km from the south of the city and 15 km beneath the surface (Kim et al., 2016). One year later on November 15, 2017, an ML5.4 earthquake occurred in Pohang, South Korea at 05:29:31 UTC, causing widespread damage in and around the city (Kang et al., 2019). The Pohang earthquake was just 40 km from the Gyeongju earthquake. The depth of the mainshock was approximately 7.0 km, and several aftershocks were reported around the area (Kim et al., 2018). Both earthquakes were the largest to occur in the Korean Peninsula since earthquake monitoring was initiated by the Korea Meteorological Administration (KMA) in 1978 and were among the most destructive events to occur in Korea during the past century (Grigoli et al., 2018; Woo et al., 2019).

Gyeongju earthquake resulted in 5368 damaged properties with approximately \($\)9.5 M (USD), 111 victims, 23 injured people (Han et al., 2020). While Pohang earthquake injured 135 residents, displaced more than 1, 700 people into emergency housing, and caused more than \($\)75 M (USD) indirect damage to over 57, 000 structures and over \($\)300 M (USD) of total economic impact, as estimated by the Bank of Korea (Lee, 2019). More than 100 heritage buildings and monuments sustained damage from the earthquakes (Sun, 2019). In Gyeongju, there is the Gyeongju Historic Area that registered as a UNESCO world cultural Heritage in November 2000, an area that embodies the time-honored history and culture of Gyeongju, the ancient capital of the Silla Kingdom (57 BC-AD 935). Some damages were found in this area, such as Dabotap Pagoda (dislocated banister), Cheomseongdae Observatory (shifted and tilted), and Gyeongju Gyochon Traditional Village (cracked walls) (Doo, 2016). This was the most damaging earthquake to strike the Korean Peninsula for centuries.

Several studies have been carried out toward Gyeongju and Pohang earthquakes, such as groundwater system responses assessment to the Gyeongju earthquake (Kim et al., 2019), investigating related factor associated with the Gyeongju earthquake (Jin et al., 2020), detecting liquefaction phenomena using Sentinel-2 and Landsat-8 after Pohang earthquake (Baik et al., 2019), investigating whether the Pohang earthquake was induced events (Kim et al., 2018), and modeling static slip using interferometric synthetic aperture radar (InSAR) (Song and Lee, 2019). A government-commissioned research team was formed and after a year-long study of the Pohang earthquake, the team concluded in 2019 that the earthquake was triggered by enhanced geothermal stimulation (EGS) project (Lee, 2019). As the largest quake ever recorded, the Gyeongju earthquake did not cause surface deformation around the epicenter (Park et al., 2018). On the other hand, the Pohang earthquake resulted in surface deformation in the eastern of the epicenter about ±5 cm. This was the first time that surface deformation could be detected by synthetic aperture radar (SAR) imagery in Korea. Considering the uniqueness of both earthquakes, we choose Gyeongju and Pohang earthquakes as our study case.

Damage proxy map (DPM) method using SAR satellites data has been shown to be useful for damage mapping following an earthquake and other natural disaster events, including the 2015 MW7.8 Gorkha, Nepal earthquake using COSMO-SkyMed and ALOS- 2 satellites (Yun et al., 2015), the 2019 typhoon Hagibis, Japan using Sentinel-1 satellites (Tay et al., 2020), the 2019 MW7.1 Ridgecrest Earthquake in California using Sentinel-1 satellites (Hough et al., 2020), and the 2014 eruption of Kelud volcano (Indonesia) using COSMO SkyMED satellites (Biass et al., 2021). DPM method is part of an ongoing collaborative effort between the Jet Propulsion Laboratory (JPL) and the California Institute of Technology, called the Advanced Rapid Imaging and Analysis (ARIA) project.

This study aims to produce DPMs derived from European Space Agency (ESA) Sentinel-1 synthetic aperture radar (SAR) imagery to explore the distribution of local damage after the Gyeongju and Pohang earthquakes. DPMs were generated from subtracting interferometric coherence of pre- and co-events to identify anomalous changes in the ground-surface properties (Hough et al., 2020). The resulted maps are expected to be useful as a basis for the risk assessment, prioritizing evacuation, and mitigation efforts.

2. Study Area

The study areas were Gyeongju and Pohang, Gyeonsangbuk-do, South Korea (Fig. 1). Gyeongju and Pohang are side-by-side cities and are located on the south-eastern edge of the Korean Peninsula. Gyeongju has a population of 254, 853, an area of 1, 324.82 km², and consists of 23 administrative districts (Gyeongju City Hall, 2021). Pohang has a population of 519, 216, an area of 1, 127 km², and consists of 29 administrative districts (Pohang City Hall, 2021).

Fig. 1. Gyeongju and Pohang, South Korea. The red and purple rectangles are the coverage of Sentinel-1 during descending observation. The blue and green lines indicate the study areas.

Gyeongju earthquake was accompanied by 600 aftershocks, including ML5.1 foreshock that occurred near the mainshock at 10:44:32 UTC and the largest aftershock (ML4.5) occurred at 11:33:58 UTC on September 19, 2017 (Jin et al., 2020). A series of aftershocks of the Pohang earthquake was also observed to have occurred with magnitude 3.0 or more until May 31, 2018. A magnitude 4.3 occurred near the epicenter, 2 hours after the mainshock, and a magnitude 4.6 earthquake occurred 4 km to the southwest at around 20:03:04 UTC on February 11, 2018 (Kim et al., 2018). Several faults are distributed within the study areas, including Dongrae, Moryang, Miryang, Ulsan, Wangsan, and Yangsan (Ellsworth et al., 2019). Therefore, the probability of earthquake occurrence in Gyeongju and Pohang is considered relatively high.

3. Method

DPMs generated from the comparison of pre- and co-event SAR images can help identify damage caused by earthquakes using remote sensing imagery (Yun et al., 2015). The method relies on the reduction in the coherence of the radar echoes between satellite-based SAR images taken before and after the earthquake to identify anomalous changes in ground surface properties. Coherence measures the change in radar backscatter from the ground, a proxy for the ground- surface properties changes. A low coherence implies a large change to the ground surface that reflected the SAR radiation (Zebker and Villasenor, 1992). Changes can be caused by damage to the ground itself or damage to structures.

For the Gyeongju earthquake, we acquired two Copernicus C-band Sentinel-1 single look complex (SLC) SAR images (Table 1), taken before the event (August 11 and 23, 2016) and one after the event (September 16, 2016). Pohang is located in between two frames (470 and 475) path 61 of the Sentinel-1 imagery (Fig. 1); therefore, we obtained six Sentinel-1 SLC images for the Pohang earthquake, with four scenes before the event on October 23, 2017, and November 4, 2017, and two scenes after the event on November 16, 2017. Furthermore, we need to merge the scenes before processing. The images were co- registered, the August 23, 2016 and November 4, 2017 scenes as reference. This co-registration process has been done at sub-pixel accuracy to match one another scene perfectly. The co-registered images were cropped to generate images that only contain the study area.

Table 1. Parameters of Sentinel-1 SAR data used for damage mapping

We used the complex pixel value, c of the pre-processed SLCs for the change detection analysis, where damage is inferred from loss of coherence or decorrelation between SAR images (Yun et al., 2015). We computed the pre- and co-event interferometric coherences, γ(Equation 1), from pair of SLCs before the event and another pair spanning across the event, respectively (Tay et al., 2020).

\(\gamma=\frac{\left|\right|}{\sqrt{\left\langle c_{1} c_{2}^{*}>\right.}}, 0 \leq \gamma \leq 1\)\)       (1)

where c₁and c₂ are complex pixel values of two co- registered SAR images and * denotes the complex conjugate. The resulting coherence is ranging from 0 (incoherent) to 1 (coherent). The coherence is equal to 1 if the observation is identical in the two images because of the stable object like buildings in the scene. The pre-event coherence represented change unrelated to the event and was assumed to be the background value. Then, we obtained a coherence difference (COD) by subtracting γ_pre-event from γ_co-event. Therefore, the process could generate COD ranging from -1 to 1. A negative COD (or coherence gain) usually indicates surface changes occurring between the pre-event scenes and is associated with changes not related to the event. Coherence gain could happen in agriculture area when the fields full of crops. Then harvested during the period time of the pre-event interferometric pair, the area has low coherence. After harvesting, leaving an empty field, the area has a greater coherence in the co-event interferometric pair (γ_co-event> γ_pre-event), so that negative COD is obtained. A positive COD (or coherence loss) indicates surface changes between the co-event scenes spanning the events, such as major damage to a building significantly increase the interferometric phase variance causing decrease in coherence. Hence, the loss of coherence is most effective for detecting damage in built-up areas caused by earthquake. However, COD is generally less effective and less reliable in vegetation areas where coherence change may be random. Therefore, this study focused on DPM in urban, built-up areas as they can be detected easily in SAR imagery. The greater loss in coherence generally correlates with greater severity of the change, such as fully collapse building causing more significant coherence loss than partial collapse (Tay et al., 2020).

The threshold for significant coherence loss can be chosen by comparing observed coherence changes with reported damage and areas in which it is known that no damage occurred. Yun et al. (2015) compared DPM with National Geospatial-Intelligence Agency (NGA) analysis and the United Nations Operational Satellite Application Programme (UNOSAT) damage assessment map. Tay et al. (2020) used high-resolution aerial imagery from the Geospatial Information Authority of Japan (GSI). Here, coherence loss thresholds for DPMs were chosen by considering the damaged area from the Korea Meteorological Administration (KMA) report of the Gyeongju and Pohang earthquake (Korea Meteorological Administration, 2018).

4. Results and discussion

1) Damage Proxy Map (DPM)

The DPMs were generated from the Sentinel-1 dataset and geocoded to the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) 1 arcsecond. Fig. 2 reveals the map view of Sentinel-1 DPMs overlaid with images from the 4-bandPlanetScope true-color scene taken on November 2, 2016, for the Gyeongju earthquake and November 16, 2017, for the Pohang earthquake. The assessment technique is most sensitive to the destruction of the built environment. Pixels are set to be relatively transparent that corresponding to areas where decorrelation did not significantly change during the time spanning the earthquake suggesting little to no destruction. Increased opacity of the radar image pixels reflects increasing ground and building change or potential damage. Color range from yellow to red indicates an increasingly significant coherence change in the area covered by the pixel. Each pixel in the DPMs was registered to the SRTM DEM and had a corresponding dimension of about 30 m.

Fig. 2. Damage proxy maps (DPMs) overlaid over PlanetScope images for (a) the 2016 Gyeongju in (b) Gwangmyeongdong and (c) Seonggeon-dong and for (d) the 2017 Pohang earthquake in (e) Heunghae-eup and (f) Duho-dong. Yellow to red pixels indicates increasingly more significant potential damage.

The DPMs show distribution of coherence loss areas afterthe earthquakes throughout Gyeongju and Pohang. According to the land cover map derived from the Korea Institute of Geoscience and Mineral Resources (KIGAM), Fig. 2(a) reveals DPM after the Gyeongju earthquake over residential areas, commercial areas, agriculture, and vegetation areas. Fig. 2(b) and Fig. 2(c) show the widespread COD in Gwangmyeong-dong and Seonggeon-dong, respectively. These areas consist of residential and commercial areas. Fig. 2(d) shows DPM that indicatesCOD over residential, commercial, industrial, agriculture, and vegetation areas affected by the Pohang earthquake. Some of the spatially large and strong coherence loss in Pohang correspond to commercial and residential areas located in Heunghaeeup (Fig. 2(e)), while Fig. 2(f) located in Duho-dong district, consists of residential areas.

2) Comparison of DPM with official report

We found a good correlation between the DPMs and the released map from the KMAreport about Gyeongju and Pohang earthquakes from our qualitative validation. The map was derived from field surveys and damage surveys data from local governments. The map shows the distribution of Gyeongju and Pohang earthquakes magnitude using the Modified Mercalli Intensity (MMI) scale (Korea Meteorological Administration, 2018). Fig. 3(a)show DPM of the Pohang earthquakes indicated by red dots and Fig 3(b) show DPM of the Gyeongju earthquake indicated by blue dots. The DPMs overlaid on the released damage map from KMA. The results were similar to the KMAreport that these areas suffered MMI Ⅶ to Ⅷ. KMA defines MMI Ⅶ to Ⅷ as major structural parts such as pillars, walls, and roofs, even in well-designed and built buildings. Fig. 3 also shows some areas that suffered MMI V to VI; however, the DPMs did not show any COD in these areas. The scales defined that the damages are inside buildings, such as minor crack walls and damage caused by dropping objects or tiles; therefore, SAR cannot detect a significant coherence change.

Fig. 3. DPMs of the (a) Pohang and (b) Gyeongju earthquakes indicated by blue and red dots, respectively, overlaid on the released damage map from KMA of the Gyeongju and Pohang earthquakes using the MMI scale. Modified from (Korea Meteorological Administration, 2018).

KMA reported that damages from the earthquakes are mainly distributed from MMI Ⅴ to Ⅷ. The damage from the Gyeongju earthquake impacted urbanization/dry areas and agriculture areas. Of the total 4,316 earthquake damage, 68% occurred in residential areas, followed by 7% in commercial areas, 7% in public facilities, 5% in rice fields, and 3% in fields. The Pohang earthquake damage occurred in urbanization/dry areas, agriculture areas, forest areas, and bare land. Among them, the damage to residential areas was the largest. Of the total 20,074 earthquake damage in Pohang, 66% occurred in residential areas, followed by 12% in other areas, 9% in coniferous forests, 8% in commercial areas, and 1% in rice fields. In the case of MMI Ⅷ, both earthquakes mainly damage residential areas. However, 25% of the Gyeongju earthquake damaged area was rice filed (Korea Meteorological Administration, 2018).

Here, the total damaged areas from DPMs were calculated by multiplying the total number of pixels by the pixel cellsize, which Gyeongju earthquake yielded an area of 1.09 km² and the Pohang earthquake yielded an area of 1,32 km2. KMA reported a total of 4.316 earthquake damage, 69 damaged buildings by the Gyeongju earthquake, and 20,074 earthquake damage with 504 damaged buildings by the Pohang earthquake (MMI Ⅶ and Ⅷ). The Pohang earthquake caused more damage than the Gyeongju earthquake, one of which was due to the depth of the epicenter. The shallower the epicenter of an earthquake, the more damage it causes. The epicenter depth of the Pohang earthquake was 7 km, while the epicenter depth of the Gyeongju earthquake was 15 km. Also, the surface deformation caused by the Pohang earthquake affected the occurrence of damage in Pohang.

3) Seismic characteristic analysis

Comparing the seismic waveform and spectrum of both earthquakes, high-frequency energy was relatively high in the Gyeongju earthquake, and low-frequency energy was large in the Pohang earthquake (Korea MeteorologicalAdministration, 2018). Generally, low-frequency energy impacts high-rise buildings, and high-frequency energy can damage low-rise buildings. Therefore, the Gyeongju earthquake caused damage to roofs and fences, and cracked walls on the first and second floor was common. The Pohang earthquake, which has low-frequency energy, damage high-rise buildings such as school and apartments.

Fig. 4 shows significant decorrelation associated with damaged buildings after the Gyeongju and Pohang earthquake. Fig. 4(a) shows DPM in Gwangmyeongdong corresponding to collapsed (Fig. 4(b)) walls and (Fig. 4(c)) roof of a house after the Gyeongju earthquake (Ministry of Public Safety and Security, 2017). Fig. 4(d) shows DPM over Handong Global University associated with (Fig. 4(e)) damaged walls by the Pohang earthquake (Lee, 2017). Based on the MMI scale by KMA, these were categorized as MMI Ⅶ or Ⅷ.

Fig. 4. DPM in (a) Gwangmyeong-dong corresponding to collapsed (b) roof and (c) walls after Gyeongju earthquake and (d) DPM corresponding to damaged (e) walls after Pohang earthquake in Handong Global University.

With frequent revisits for six days and freely accessible of the Copernicus Sentinel-1 satellites, the Sentinel-1 dataset can provide fast and reliable damage maps after a disaster using the DPM method to assist mitigation strategies to reduce casualties. However, given the SAR pixel size, the DPM from Sentinel-1 does not, as expected, have sufficient resolution to identify all limited damage to individual structures and may not reveal, for example, some MMI Ⅷ in a high density of buildings such as in Heunghae-up district could not be detected. As these earthquakes occurred in Korea, the Korean government could use the high resolution KOMPSAT-5 SARsatellite (2.2 m) operated by Korea Aerospace Research Institute (KARI) to make damaged area maps utilizing the DPM method. Nevertheless, our results managed to reveal damaged buildings after the Gyeongju and Pohang earthquakes in residential, commercial, and industrial areas. The resulting map can be used to assess the distribution of seismic damage and can be used as inventory data of damaged buildings to map a seismic vulnerability using machine learning in Gyeongju or Pohang.

5. Conclusion

The Damage Proxy Maps were produced from ESA Sentinel-1 SAR data for the 2016 Gyeongju earthquake and the 2017 Pohang earthquake. DPMs were generated by comparing pre- and co-events interferometric coherences to identify anomalous changes that indicate damaged by the earthquakes. DPMs manage to reveal coherence loss in residential and commercial areas in both Gyeongju and Pohang earthquakes. We found that our results show a good correlation with the KMA report with MMI values of more than Ⅶ. The color scale of Sentinel-1 DPMs indicates an increasingly significant change in the area covered by the pixel, delineating collapsed walls and roofs from the official report. With the frequent six-day revisit and freely accessible of the Copernicus Sentinel1 satellites, the Sentinel-1 dataset could provide a fast and reliable damaged map after a disaster occurred using the DPM method to assist mitigation strategies and reduce the number of victims. The results can also be used as inventory data of damaged buildings to map seismic vulnerability using machine learning in Gyeongju or Pohang.