Journal of Civil Engineering Science & Technology

Vol. 02/No. 01/ April 2021

DETECTION OF FLOOD IMPACTED AREAS IN USING SENTINEL-1 IMAGERY

Bagas Aryaseta1*

1 Department of Civil Engineering, Engineering Faculty, University of Pembangunan Nasional “Veteran” Jawa Timur, *Corresponding Author: [email protected]

Abstract Article Info Flash Floods in East Nusa Tenggara occured on April 4th, 2021. These Flash Floods Received 25 April 2021 are scattered from East , Regency, , Malacca Revised 27 April 2021 Regency, , City, , and . Accepted 29 April 2021 The cause of these Flash Floods is the high intensity of rain caused by the tropical . Mapping of flood locations plays an important role in prevention and mitigation efforts. In this study, InSAR data processing was carried out from the Sentinel 1A satellite to find flood-affected locations in East Nusa Tenggara. 32 images of Sentinel-1 were processed before and 31 images after the Flash Floods incident. The method used is the classification method using cloud computing, Google Earth Engine. The results show that the flood-affected areas can be detected based on a lower pixel value (indicating a very small signal backscatter value), then compared to the conditions before the flood. The four sample points identified, namely points A, B, C, and D each have pixel values of -8.58, -9.99, -12.43, and -9.29 for the VV polarized image, respectively. For VH polarized image is -17.35, -17.96, -17.84, and -14.22, respectively. Keywords: flash flood, east nusa tenggara, sentinel-1, insar, google earth engine

INTRODUCTION technology to map and monitor floods includes coherent techniques [5] [6] [7] [8], Normalized Difference Ratio On Sunday, April 4th, 2021, there was a Flash Floods (NDR) technique [9], naked eye identification by relying on in East Nusa Tenggara. Flash floods in East Nusa Tenggara a Grayscale visualization which is based on the backscatter are scattered from , , signal pixel coefficient [10], Quasi-Persistent Scatterers (Q- Alor Regency, Malacca Regency, Sabu Raijua Regency, PS) [11] and so forth. Kupang City, Kupang Regency, and Ende Regency. The In this study, InSAR data processing obtained from the cause of this Flash Floods is the high intensity of rain caused Sentinel 1A satellite was carried out to obtain flood-affected by the tropical cyclone Seroja. According to the National locations in East Nusa Tenggara using a classification Disaster Management Agency (BNPB), this Flash Floods is technique based on the backscatter signal pixel coefficient. the worst natural disaster in the last 10 years in East Nusa The results are expected to become a reference for Tenggara [1] [2]. Meanwhile, according to Regional policymakers as well as a preliminary study for other Disaster Management Agency (BPBD) East Nusa Tenggara researchers. as of April 28th, 2021, it has been recorded that the resulting impact is 182 people who have died in 10 districts and cities RESEARCH METHOD in East Nusa Tenggara [3]. Meanwhile, 47 missing people A. Data are still in search. In addition, a total of 1992 houses and 87 The research area covers the entire province of public facilities were affected [4]. East Nusa Tenggara. The data used are data before and Mapping of flood locations plays an important role in after the occurrence of Flash Floods in East Nusa prevention and mitigation efforts. In the last few decades, Tenggara on April 4th, 2021. Sentinel-1 images that are remote sensing technology has been developed very processed are 31 images before and 32 images after the rapidly. Not only the spatial resolution is increased, but the Flash Floods incident. The image used is IW mode with temporal resolution has also been improved. That way, VV and VH polarization. The entire image has efforts to monitor the earth's surface can be done better. ascending orbit with a resolution of 10 meters. Several studies have shown that the use of InSAR

37 CI-TECH Journal/Bagas Aryaseta1*/Vol. 02/No. 01/2021

B. Methodology appropriate enough, the next step is to produce a mosaic. The method used is the backscatter signal pixel After that, the Sentinel-1 data mosaic process was coefficient classification method using cloud carried out. The mosaic process is a process of computing, Google Earth Engine. The processing flow combining several Sentinel-1 data before the flood and diagram can be seen in Figure 1. The first process is after the flood for each channel (VV and VH). After this done by defining the Region of Interest (ROI) by process, the entire InSAR imagery coverage in the study drawing a line manually through the Google Earth area will be seen (Figure 3). Engine website. The ROI area covers the entire province of East Nusa Tenggara (figure 2). After that, the Sentinel-1 data filtering process is carried out. The first filter is used to select the Sentinel-1 product type (mode, polarization, orbit, and resolution). The second filter is a filter based on date. In this study, 32 images were filtered before Flash Floods started on February 12th, 2021 to March 12th, 2021. For the images after Flash Floods incident, there were 31 images from April 5th, 2021 to April 30th, 2021.

Start

Figure 2. Region of Interest

Defining Region of Interest

Filtering Sentinel-1

data

Filtered Sentinel-1 data Figure 3. InSAR coverage

The next process is to display RGB colors for Mosaicing easier analysis. This process is done by adding images before the incident (R), after the incident (G), and before Combination of filtered the incident (B). The results can be seen in figures 4 and Sentinel-1 data 5. Reddish color (pink) indicates inundated (flood) areas. Then applying a speckle filter of 50 to provide a smooth effect on the image to reduce the noise effect Showing RGB (salt and paper effect) to facilitate classification. However, this process will reduce the image resolution. The figure 6 and 7 are the comparison between before RGB image and after applying speckle filter.

Speckle filter

The final result (classification map)

Figure 1. Processing workflow

Then, data identification was carried out to see the image quality and identifying the flood manually. If it is Figure 4. RGB image of Island

38

CI-TECH Journal/Bagas Aryaseta1*/Vol. 02/No. 01/2021

Figure 5. RGB image of Flores Island

Figure 9. Points C dan D

Coordinate Pixel Before After Before After Figure 6. Before applying speckle filter Point Longitude Latitude Flood Flood Flood Flood VV VV VH VH

A 120.2041 -9.7734 -2.79 -8.58 -11.36 -17.35 B 120.1379 -9.8211 1.84 -9.99 -4.78 -17.96 C 121.2465 -8.4963 0.23 -12.43 -7.41 -17.84 D 121.2673 -8.5564 6 -9.29 -1.79 -14.22

The final step is to carry out the final classification by applying a threshold. The threshold selected is 1.25. Figure 7. After applying speckle filter This means that values above 1.25 represent flooded areas (areas in blue). The final result is a final map of The next step is to calculate the difference between the classification in Figures 10 and 11. There are several before and after the flood. Areas in white indicate areas errors in classifying the flooded area as shown in Figure that are flooded. In this process, an inspection is carried 11 on the right side of the figure. This can be caused by out by looking at and comparing the pixel values before poor image quality. and after the incident. The results of this difference can be seen in Table 1. For the sample, 4 points were taken. Points A and B are located in the Lalatan and Laau areas (Sumba Island). Points C and D are located in Makitoenggoeng and Pugubengo (Flores Island). According to BNPB data, the four points were confirmed to have been affected by Flash Floods.

Figure 10. Final Classifications of Points A and B

Figure 8. Points A dan B

39

CI-TECH Journal/Bagas Aryaseta1*/Vol. 02/No. 01/2021

https://nasional.kompas.com/read/2021/04/07/16405 661/bnpb-banjir-bandang-ntt-terbesar-dalam-10- tahun-terakhir-di-provinsi-itu.

[3] Supriatin, "Total Korban Jiwa akibat Siklon Tropis di NTT Jadi 182 Orang," 29 April 2021. [Online]. Available: https://www.merdeka.com/peristiwa/total- korban-jiwa-akibat-siklon-tropis-di-ntt-jadi-182- orang.html.

[4] Liputan6, "Banjir Bandang di NTT: 1.992 Rumah Rusak, 87 Fasilitas Umum Terdampak," 8 April 2021. [Online]. Available: https://www.liputan6.com/news/read/4526473/banjir- Figure 11. Final Classifications of Points C and D bandang-di-ntt-1992-rumah-rusak-87-fasilitas- umum-terdampak.

CONCLUSION [5] C. Chaabani, M. Chini, R. Abdelfattah, R. Hostache, and K. Chokmani, "Flood Mapping in a Complex From this research, it can be concluded that InSAR Environment Using Bistatic TanDEM-X/TerraSAR- satellite technology can be used for various things, one of X InSAR Coherence," remote sensing, 2018. which is to map flood-affected areas. This research uses a classification technique based on the backscatter signal [6] F. Canisius, B. Brisco, K. Murnaghan, M. V. D. pixel coefficient. The results show that the area after being Kooij and E. Keizer, "SAR Backscatter and InSAR affected by the flood has a lower pixel value than before it Coherence for Monitoring Wetland Extent, Flood was impacted. From the four points, namely A, B, C, and D, Pulse and Vegetation: A Study of the Amazon the pixel values obtained after Flash Floods event for VV Lowland," remote sensing, 2019. were -8.58, -9.99, -12.43, and -9.29, respectively. Whereas the pixel values for VH were -17.35, -17.96, -17.84, and - [7] M. Chini, R. Pelich, L. Pulvirenti, N. Pierdicca, R. 14.22, respectively. In the future, processing Sentinel-1 Hostache and P. Matgen, "Sentinel-1 InSAR imagery could be also done by using Descending mode as a Coherence to Detect Floodwater in Urban Areas: comparison and a more comprehensive result. Houston and Hurricane Harvey as A Test Case," remote sensing, 2019.

[8] A. Refice, D. Capolongo, G. Pasquariello, A. ACKNOWLEDGEMENT D’Addabbo, F. Bovenga, R. Nutricato, F. P. I would like to express my special thanks of gratitude Lovergine and L. Pietranera, "SAR and InSAR for to the Department of Civil Engineering UPN “Veteran” Flood Monitoring: Examples With COSMO-SkyMed Jawa Timur for providing me with all the facility that was Data," IEEE JOURNAL OF SELECTED TOPICS IN required. APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, 2014. REFERENCES [9] C. Alexandre, R. Johary, T. Catry, P. Mouquet, C. Révillion, S. Rakotondraompiana and G. Pennober, [1] S. A. Siswanto, "Banjir Bandang NTT Awal April "A Sentinel-1 Based Processing Chain for Detection 2021 Terparah Dalam 10 Tahun TerakhirBanjir of Cyclonic Flood Impacts," remote sensing, 2020. Bandang NTT Awal April 2021 Terparah Dalam 10 Tahun Terakhir," 05 April 2021. [Online]. Available: [10 Jefriza, H. Lateh, I. M. Yusoff, I. A. Abir, S. https://www.suara.com/news/2021/04/05/150529/ban ] Syahreza, P. Razi and M. Rusdi, "Application of jir-bandang-ntt-awal-april-2021-terparah-dalam-10- Interferometric SAR using Sentinel-1A for Flood tahun-terakhir. Monitoring in South of Sulawesi, Indonesia," The 5th International Conference of Indonesian Society for [2] R. N. Chaterine, "BNPB: Banjir Bandang NTT Remote Sensing, 2019. Terbesar dalam 10 Tahun Terakhir di Provinsi Itu," 07 April 2021. [Online]. Available: 40

CI-TECH Journal/Bagas Aryaseta1*/Vol. 02/No. 01/2021

[11 Jefriza, I. M. Yusoff, I. A. Abir, S. Syahreza, M. ] Rusdi, P. Razi and H. Lateh, "The applications of InSAR technique for natural hazard detection in smart society," The 9th International Conference on Theoretical and Applied Physics (ICTAP), 2020.

41