以衛載雷達影像分析進行崩塌潛勢及潛在量體評估

Study on the Assessment of Landslide Potential and Volume Based on Satellite Radar Data Analysis

林俊廷(Jun-Ting Lin)

指導教授 ： 王國隆

國立暨南國際大學/科技學院/土木工程學系/博士(2021年)

https://doi.org/10.6837/ncnu202100058

摘要

台灣的地理條件造就了高災害潛勢的環境，而坡地災害是台灣面臨的主要災害之一，由於坡地災害影響深遠又難以預測，如何調查及評估災害潛勢一直是倍受關注的課題，近年來因為遙測技術的進步，使災害評估的方式能更快速且精確。
合成孔徑雷達(SAR)有不受天候干擾、廣域監測、固定再訪周期等優勢，能紀錄包含強度、極化、相位等物理訊息，此外還能運用差分干涉合成孔徑雷達技術(DInSAR)能得到微小的地表變形資訊，在本研究中即以ALOS、ALOS-2的SAR影像建立崩塌的災害探查、潛勢範圍劃設、以及崩塌量體評估的三個階段研究。
災害探查方面，本研究使用SAR的強度與極化資訊，分析新生崩塌地的位置，運用後向散射歸一化指標(NDSI)與雷達植生差異化指標(RVID)的崩塌判釋結果，組合成NDSI&RVID的合成指標，莫拉克災害的案例達到Kappa=38.9%的分類精度，在1公頃以上的崩塌事件中有30~88%的判釋率，並建議崩塌判定公式NDSI&RVID<μ−cσ中的門檻值c介於1.4~1.7之間。
崩塌潛勢範圍的研究中，以GPS校正DInSAR的分析成果可大面積的萃取潛在的崩塌範圍，如被重覆判釋為潛在崩塌區，則會提高範圍內新生崩塌的機率，出現三次以上的崩塌發生率會達到50%以上，判釋的潛在崩塌區主要是舊崩塌地的延伸，顯示崩塌地擴大有地表變形的徵兆，同時也能發現潛在崩塌地的活躍程度。
滑動量體評估的研究中，在無地質資料的情況下，運用Okada模型反演算出崩塌滑動深度的近似解，並假設崩塌地為橢球體，估算潛在滑動量，最後建立了各種逆推參數的約束條件，在清境及廬山的案例中，分別計算出符合滑動事件的滑動面，與測傾管監測比對後，滑動面的誤差量約在1~13m之間，誤差率約1~18%之間，根據研究結果，此研究方法適用於平面型滑動，弧形滑動則須進行修正。

關鍵字

崩塌 ； 崩塌判釋 ； 崩塌潛勢 ； 崩塌量體 ； 合成孔徑雷達 ； 差分干涉技術

並列摘要

Taiwan's geographical conditions create a high-disaster-potential environment, and slope hazard is one of Taiwan's major disasters. Because of the significant influences and unpredictability of slope hazard, the disaster potential investigation and assessment have always been the great-concern topic. Due to the advancement of remote sensing technology, disaster assessment can now be faster and more accurate.
Synthetic Aperture Radar(SAR) has the advantages of weatherproof, wide-area monitoring, and fixed revisit period, which can record physical information including intensity, polarization, phase, etc. Besides, differential interferometric synthetic aperture radar (DInSAR) technology can obtain micro-topography deformation information. This study uses ALOS and ALOS-2 SAR images to establish the three-stage research, including landslide disaster detection, landslide potential range determination, and landslide volume assessment.
This study uses SAR intensity and polarization information to analyze the new landslide location in the stage of landslide disaster detection. After comparing the landslide interpretation characteristics between NDSI and RVID, the results composite the combination index of NDSI&RVID. In typhoon Morakot, the classification accuracy of Kappa=38.9% is obtained with the interpretation rate between 30% to 88%, which more than 1 hectare. The threshold value c can be recommended as 1.4 to 1.7 in the landslide determination formula NDSI&RVID<μ−cσ.
In the potential landslide range, the analysis results of DInSAR with GPS calibration can extract the potential landslide area in a wide range. If the area is repeatedly judged as a potential landslide, the probability of a new landslide will increase in the area. If the determined result is judged more than three times as a potential landslide, the landslide occurrence rate will reach more than 50%. The estimated potential landslides are mainly the extension of the past landslides, indicating that there are signs of surface deformation when the landslide area expands, and the potential landslide active level can also be found.
In the landslide volume assessment stage, with no geological data, this study uses the Okada model to calculate the approximate solution of the landslide depth inversely. The landslide area is assumed to be an ellipsoid to estimate the potential sliding volume and establish various inverse parameters. In Qingjing and Lushan, the results calculate the sliding surface that conforms to the landslide event. After comparing the clinometer data, the sliding surface's datum error is about 1m to 13m, and the error rate is about 1% to 18%. According to the study results, this method is suitable for planar sliding. On the other hand, this method used on arc sliding must be corrected.

並列關鍵字

landslide ； landslide detection ； landslide potential ； landslide volume ； SAR ； DInSAR

參考文獻

8.1 英文文獻

1. Agliardi, F., Crosta, G., and Zanchi, A., (2001), Structural constraints on deep-seated slope deformation kinematics. Engineering Geology 59, 83–102.

2. Aryal, A., Brooks, B. A., and Reid, M. E., (2015), Landslide subsurface slip geometry inferred from 3-D surface displacement fields. Geophys. Res. Lett., Vol. 42, pp.1411-1417.

3. Aulakh, N. S., Chhabra, J. K., Kamar, A., and Aggarwal, A. K., (2004), Development of a Fiber Optic based System to Monitor Landslide Activity, IETE Technical Review, Vol. 21 No. 1, pp.75-81.

4. Bähr, H., and Hanssen, R.F., (2012), Reliable Estimation of Orbit Errorsin Spaceborne SAR Interferometry, Journalof Geodesy, Vol. 76 No.12 , pp.1147-1164