1.

INTRODUCTION

Today we are witnessing of crises caused by natural disasters covering significant territories, and in some cases they even extend to the territories of neighboring countries, as was the case with the earthquakes of February 6, 2023 in the territories of Turkey and Syria. Most of these crises cause huge destructions, involve human, financial and material resources, continued with days. Daily map of the events of the earthquakes is shown on figure 1.

Figure 1.

Daily map of the M7.8 and M7.5 earthquakes in Türkiye and Syria [1]

The figure 1 shows the scale and huge territory that the two strong consecutive earthquakes struck. Thanks to the rapid development of technologies and to the open access satellite Earth observation data, it is possible to assess the destruction in urban areas on big scale. The approach which is tested in this survey is usage of change detection (CD) techniques using open access high resolution multispectral data from Sentinel SCI Hub repository.

2.

METHODOLOGY

2.1

Satellite data

For the disaster assessment [7, 9, 11, 13, 15, 17, 19] in the area of the Türkiye – Syria border Sentinel-2 images with close to zero percentage cloud coverage was selected. Two products covered plate T36SYF in accordance of European Space Agency (ESA) naming nomenclature from dates 2022.10.27 and 2023.03.01 was used for the survey presentation in the paper.

For reference map was used map of Antakya, Turkiye (Figure 2), showing number, location and statistics of destroyed buildings per building block determined using data from the Humanitarian Open Street Map Team (HOTOSM) (downloaded on 15 February, 11:12, GMT+1) [2].

Figure 2.

Map of Antakya, Turkiye (downloaded on 15 February, 11:12, GMT+1) [2]

2.2

Change detection techniques

The following pixel-based change detection (PBCD) techniques for urban area disaster assessment in high resolution satellite data was tested: image differencing (using Sentinel-2 bands 2 and 4 separately) [3], image rationing (using Sentinel-2 bands 2 and 4 separately) [3] and the disturbance index (DI) [4].

Mathematical expressions of the selected PBCD techniques are:

Using the above-mentioned data (Table 1 and Figure 3) and mathematical expressions, a study was conducted to assess the adequacy of the selected PBCD techniques for urban area disaster consequences assessment. The study was executed using MATLAB by developing the GUI, algorithms and corresponding source code for reading, image processing and mixing the satellite image (as a background) and the PBCD index mask showing the locations and areas with changes detected [6, 8, 10, 12, 14, 16, 18].

Figure 3.

Sentinel-2 bands defined by ESA

Table 1.

ESA Sentinel-2 products used.

3.

RESULTS

Figures 4 (from a to d) and Figure 5 shown the results of data processing. The pixels in which a change was detected depending of the PBCD technique are shown in red. It should be noted that the image used to assess the earthquakes consequences (Sentinel-2) is dated March 1, 2023, and the image used for reference (Figure 2) is dated February 15, 2023. It should be noted that during these almost two weeks clearing activities of the collapsed buildings were carried out.

Figure 4.

PBCD difference and rationing results for Antakya area

Figure 5.

PBCD Disturbance Index (DI) results for Antakya area

Table 2.

Summary of PBCD index images and corresponding mask thresholds

The images of Figures 4 and 5 show a high degree of correlation with the reference image (Figure 2). However, the PBCD indices shows more destruction areas than depicted in Figure 2, which may be due to clean-up activities carried out during the two-week time interval between data sets (Sentinel-2 and HOTOSM) or to an incomplete reporting of destructions in all parts of the city by the HOTOSM.

Table 3 shows the summary of the survey final results expressed in square kilometers and ratio to gain a more realistic assessment of the performance of the tested techniques. For this purpose, one pixel of the scene is assumed to be 10 by 10 meters or 100 square meters. Through this approach, the area of the scene and the area formed by the number of red pixels, which represents the assessment of the consequences of successive earthquakes in the urban area were calculated.

Table 3.

Summary of the survey final results

The PBCD difference and disturbance index techniques show similar results in estimating the percentage of disaster assessment in an urban area, which gives us the right to rate them as more adequate than the rationing technique.

4.

CONCLUSION

The conducted study of the possibilities of PBCD techniques for assessing the consequences of natural disasters, in particular earthquakes, shows that by using multispectral high resolution satellite data is realistic to visualize and calculate with sufficient accuracy the area of the affected urban areas.

Higher estimation accuracy and visualization precision can be achieved by using very high resolution multispectral satellite data. Such data are commercially available and have a resolution of 1.5 to 0.3 meters per pixel.

In addition, further research can be conducted to evaluate the adequacy of other change detection techniques or to develop new ones.