11:00 AM - 1:00 PM
[STT39-P01] Flood area Extraction around the Tokigawa-river at the Typhoon No. 19 event in 2019 using Senitinel SAR
Keywords:synthetic aperture radar, flooding, backscattering coefficient
Typhoon No. 19 in the first year of Reiwa caused recorded heavy rains in Shizuoka prefecture, Kanto region, Koshinetsu region, Tohoku region, etc, and caused the largest damage ever. In recent years, the probability of heavy rainfall increased due to global warming, and many heavy rainfall disasters occurred. It is predicted that heavy rainfall and short-term heavy rainfall will increase in frequency and precipitation in the future, raising concerns about the occurrence of large-scale floods. On the other hand, flood damage investigation requires a large number of personnel and great danger. Therefore, the purpose of this research is to apply the two merits of synthetic aperture radar (SAR) to all-weather and wide-area and area observation for disaster response. Here, we develop the flood-area detection algorithm and verify the accuracy.
① From the SAR image acquired at the time of the typhoon No. 19, the backscattering coefficient (σ0) of the inundated area and that of the non-inundated area, are measured. Next, create a classification algorithm based on the threshold method, where the threshold(σth) was calculated statistically as their intersection.
② Using the above classification algorithm and Sentinel data acquired simultaneously with ground truthing in 2021 as the non-flooding case, the accuracy of water area detection was measured. The appropriateness of the method of this threshold method and the appropriateness of the threshold itself are verified.
③ Finally, the above classification algorithm was applied to the time of Typhoon No. 19 and examine the accuracy.The target area and data are as follows: ② For verification of the classification algorithm, four SAR images taken by the Sentinel-1 satellite taken from October to December 2021 and GPS by field observation that was almost synchronized with the arrival of SAR. The Normalized Difference Water Index (NDWI) created from the log data and the observation data of the Sentinel-2 optical satellite was used. In ③, the SAR of Sentinel-1 was taken on November 12, 2019, covering the areas of Kawagoe City, Sakado City, Higashimatsuyama City, and Kawajima Town, Hiki District, where flood damage occurred due to the collapse of the Arakawa River Oppe River and the Toki River. Images were used.
As a result, the classification of land and water area has a threshold value of -19.54 dB and the classification accuracy is 83.36% (verification using NDWI), and the classification of water area extraction only has a threshold value of -16.37 dB and an accuracy of 98.92% (using NDWI). Verification) was recorded. Next, when this algorithm was applied to the SAR image of Typhoon No. 19 in the first year of Reiwa, the classification accuracy was 60.27% at the threshold of -19.54 dB and the same accuracy of 74.02% was recorded at the threshold of -16.37 dB. It is considered that this difference in accuracy is due to the time difference of about 34 hours between the survey date and time of the inundation estimation stage map of the Geographical Survey Institute, which was set as the true value, and the observation date and time of SAR.
It was found that the water surface by using the threshold method was effectively detected with high accuracy. In the future, we would like to work on the detection of inundation in urban areas and the improvement of the accuracy of the two classifications of land area and water area, which will lead to the improvement of the detection accuracy of inundation area due to flood damage.
① From the SAR image acquired at the time of the typhoon No. 19, the backscattering coefficient (σ0) of the inundated area and that of the non-inundated area, are measured. Next, create a classification algorithm based on the threshold method, where the threshold(σth) was calculated statistically as their intersection.
② Using the above classification algorithm and Sentinel data acquired simultaneously with ground truthing in 2021 as the non-flooding case, the accuracy of water area detection was measured. The appropriateness of the method of this threshold method and the appropriateness of the threshold itself are verified.
③ Finally, the above classification algorithm was applied to the time of Typhoon No. 19 and examine the accuracy.The target area and data are as follows: ② For verification of the classification algorithm, four SAR images taken by the Sentinel-1 satellite taken from October to December 2021 and GPS by field observation that was almost synchronized with the arrival of SAR. The Normalized Difference Water Index (NDWI) created from the log data and the observation data of the Sentinel-2 optical satellite was used. In ③, the SAR of Sentinel-1 was taken on November 12, 2019, covering the areas of Kawagoe City, Sakado City, Higashimatsuyama City, and Kawajima Town, Hiki District, where flood damage occurred due to the collapse of the Arakawa River Oppe River and the Toki River. Images were used.
As a result, the classification of land and water area has a threshold value of -19.54 dB and the classification accuracy is 83.36% (verification using NDWI), and the classification of water area extraction only has a threshold value of -16.37 dB and an accuracy of 98.92% (using NDWI). Verification) was recorded. Next, when this algorithm was applied to the SAR image of Typhoon No. 19 in the first year of Reiwa, the classification accuracy was 60.27% at the threshold of -19.54 dB and the same accuracy of 74.02% was recorded at the threshold of -16.37 dB. It is considered that this difference in accuracy is due to the time difference of about 34 hours between the survey date and time of the inundation estimation stage map of the Geographical Survey Institute, which was set as the true value, and the observation date and time of SAR.
It was found that the water surface by using the threshold method was effectively detected with high accuracy. In the future, we would like to work on the detection of inundation in urban areas and the improvement of the accuracy of the two classifications of land area and water area, which will lead to the improvement of the detection accuracy of inundation area due to flood damage.