In Detail: Earthquake Urban Damage Detection Using Sentinel-1 Data

This UN-SPIDER Recommended Practice emphasizes the use of Sentinel-1 SAR data for building and infrastructure classification. The data before and after a disaster can be utilized in a simple change detection methodology to quickly and easily highlight areas of major destruction. This method can be best applied in earthquakes and other disasters where there is no influx of water or debris from outside the region of interest.

On 12 November 2017 a 7.3 magnitude earthquake struck close to the town of Sarpol-e-Zahab in the Kermanshah province of Iran. The earthquake ranked as the strongest of 2017 and among the deadliest for the decade, killing over 600 people. Fatalities were spread across the border province and destruction of local infrastructure, healthcare centers and emergency services made response harder. This Recommended Practice will outline the process for obtaining and processing SAR data before and after the 2017 earthquake for the purpose of creating actionable and helpful maps for disaster managers to make informed decisions.

Data requirements:

  • This recommended practice highlights the steps necessary to classify urban areas from Level-1, Ground Range Detected (GRD) SAR data in Interferometric Wide Swath (IW) mode. Change detection can be used for other SAR data but would need to undergo more preprocessing steps that are not covered in this Recommended Practice.
  • The Sentinel data downloaded for this Recommended Practice have a descending orbit collected in relative orbit number 6. Data collected in an ascending orbit may also be used, so long as both before and after data have the same orbit direction.
  • Free registration is required to download Sentinel Data

 

Software requirements:

  • The preprocessing of Sentinel SAR data requires SNAP software, which is freely available (http://step.esa.int/main/download/snap-download/) for Windows 32 and 64-bit versions and Mac OS X; the current version is 6.0.0.
  • The processing and post-processing of the data in this Recommended Practice use QGIS which is available for free as well (https://www.qgis.org/es/site/) for Windows 32 and 64-bit versions and Mac OS X; the current version is 3.6.3.
  • OpenStreetMaps (OSM) street map data was downloaded for free (http://download.geofabrik.de/) for use in the post-processing portion of this Recommended Practice.

This recommended practice can be used with any SAR images of urban areas. For this practice, the Copernicus Programme’s Sentinel-1 satellite data is used. The selected data has a polarization of VV VH, in preprocessing only VV will be singled out, which has the clearest results for differentiation of urban and infrastructure areas from other areas. (Deepthi et al., 2018) It is necessary for the before and after images to have the same orbit direction (ascending or descending), it is beneficial if they have the same or close orbit tracks (relative orbit numbers).

This methodology can be applied to urban areas with large building destruction. Errors can be introduced as outside debris enters the area or flooding occurs, thus it is not recommended for hurricanes, tornadoes or tsunamis. Errors can also be introduced through precipitation change or vegetation growth, thus knowing the urban-agricultural layout of the region of interest and knowing about major snowfall or rain in the time range is important to avoid errors.

Strengths:

This method is recommended for any area with large amounts of urban damage. The backscatter response from SAR increases as there is a higher spatial interaction between ground and buildings, allowing for the “double-bounce” effect to increase scattering strengths. In this way, this recommended practiced is easy to apply for urban areas. SAR does not depend on the position of the sun or any other regional or temporal factors. SAR sensors freely penetrate clouds and the data collected is weather independent.

Finally, the open Earth Observation (EO) data as part of the COPERNICUS constellation of freely available globally and allows the recurrent acquisition of SAR images over major urban areas, facilitating the use of this method.

 

Limitations:

Small or single structure destruction is difficult to isolate from errors. Areas with flooding or large amounts of debris may find too many errors for this method to produce useful results. Mountains or sharp natural rock features in the area of interest may appear as errors in this process but can be edited out easily as they often form far away and outside of urban areas.

Deepthi, R. & Ravindranath, Sudha & Kasaragod, Ganesha Raj. (2018). Extraction of Urban Footprint of Bengaluru City Using Microwave Remote Sensing. ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. XLII-5. 735-740. 10.5194/isprs-archives-XLII-5-735-2018.

Miranda, N. & Meadows, R.J. (2015). Radiometric Calibration of S-1 Level-1 Products Generated by the S-1 IPF. European Space Agency. Issue 1. Revision 0. https://sentinel.esa.int/documents/247904/685163/S1-Radiometric-Calibration-V1.0.pdf.

Weiß, Thomas. (2018) Sar-pre-processing Documentation. https://buildmedia.readthedocs.org/media/pdf/multiply-sar-pre-processing/get_to_version_0.4/multiply-sar-pre-processing.pdf.

Yommy, Aiyeola Sikiru & Liu Rongke & Wu, And Shaung. (2015) SAR Image Despeckling Using Refined Lee Filter. 7th International Conference on Intelligent Human-Machine Systems and Cubrnetics. Hangzhou. Pp. 260-265. https://ieeexplore.ieee.org/document/7334965/authors.

Yousefi, Kosar & Pirani, Davood & Sahebi, Ali. (2018). Lessons Learned from the 2017 Kermanshah Earthquake Response. Iranian Red Crescent Medical Journal. In Press. 10.5812/ircmj.87109.