Natural Hazards. Gülşen Taşkın Kaya, PhD. gulsen.taskin at gmail.com. ITU Earthquake Engineering & Disaster Management June, 23-27, 2014.

Transcription

1 Natural Hazards Gülşen Taşkın Kaya, PhD. gulsen.taskin at gmail.com ITU Earthquake Engineering & Disaster Management June, 23-27, 2014.

2 Afet Yönetimi

3 Problemin Tanımı

4 Applications of RS Disaster Mitigation Preparedness (Warning) Earthquakes Volcanic Eruptions Landslides Floods Storm surge Hurricanes Fires Drought Mapping geological lineaments and faults, the study of tectonic setting of an area and Neo tectonic studies, land use Topography and land use maps, identification of potentially dangerous volcanoes especially in remote areas, mapping volcanic landforms and deposits Topographic and land use maps, mapping the factors related to the occurrence of landslides, landslide inventory Land use maps, flood plain maps; numerical weather prediction models to assess hydrological and hydro geological risk, mapping of historical floods, detailed geo-morphological terrain mapping Land use and land cover maps Tracking Mapping vegetation type and stress Soil moisture, vegetation type Geodynamic measurements ofstrain accumulation Detection/measurement of gaseous emissions, detecting and monitoring volcanic eruptions, measurement of heat increase Rainfall, slope stability Local rainfall measurements, regional rainfall Sea state; ocean surface wind velocities Synoptic weather forecasts Monitoring fires, monitoring fires smoke Long ranged climate models Application of remote sensing in disaster management Recovery / Response Locating stricken areas, damage map Mapping lava flows, ash falls and lahars, damage map Mapping slide area Flood damage map Mapping extent of damage Mapping extent of damage Detecting burned areas Monitoring vegetative biomass

5 Remote sensing methodology for damage assessment

6 Image based damage assessment system

7 Types of collapsed building

8 Table 3 Classification of damage to masonry and reinforced buildings (Taken from EMS98, Grunthal, 1998). Masonry buildings Reinforced buildings Classification of damages Grade 1: Negligible to slight damage (no structural damage, slight non-structural damage) Grade 2: Moderate damage (slight structural damage, moderate non-structural damage) Grade 3: Substantial to heavy damage (moderate structural damage, heavy non-structural damage) Grade 4: Very heavy damage (heavy structural damage, very heavy non-structural damage) Grade 5: Destruction (very heavy structural damage)

9 Damage Scales The European Macroseismic Scale 1998 (EMS98) for masonry and reinforced building.

10 DEPREM HASAR TESPİTİ YAKLAŞIMLARI

11 Deprem hasar belirleme

12 Çoklu zamansal yaklaşım

13 Görüntü ön işleme 1. Radyometik düzeltme Piksel değerlerinin düzeltilmesi 2. Atmosferik düzeltme difference at spectral responses. 3. Geometrik düzeltme (görüntü kayıtlama) Distorsiyonlu görüntü koordinatlar ile harita koordinatlar aras ndaki iliflki. Harita (x,y) u=f (x,y) Görüntü (u,v) v=g (x,y) 4. Geometrik bozulma different angle of view.

14 Görüntü kayıtlama

15 Tekli zamansal yaklaşım

16 Sınıflandırma tipleri Denetimsiz Ön bilgi gerektirmez. Sınıflar önceden bellidir. Denetimli Ön bilgi gerektirir. Eğitim veri seti hazır olmalıdır.

17 Eğitim verilerinin seçimi Yerleflim ve orman kategorilerine ait e itim alanlar.

18 Doku Bilgisi Piksel Tabanlı Sınıflandırma Doku tabanlı sınıflandırma (contextual) Dimensionally reduction Feature extraction Feature selection Gri değer tekrarlama matrisi (Gray Level Co-occurrence Matrix-GLCM) 2. Derece Haralick öznitelikleri

19 Hasar tespit doğruluğu Yüksek çözünürlüklü veriler Arazi ölçmeleri Görsel doğrulama Düşük çözünürlüklü veriler Arazi ölçmeleri Öncesi görüntü Hata matrisleri ve istatistiksel analizler Doğruluk (%) Kappa değeri

20 Accuracy Assessment Assigned classes Ground truth classes Water Vegetation Urban Water Vegetation Urban Damage Pixels TEST DATA Building Pixels D D D B B B B LABELS D D B B B D D D B CONFUSION MATRIX D B Total 3 4 Producer's Acc. 2/3 66% 2/4 50% Total 4 3 User's Acc. 2/4 =50% 2/3 =66% Ground Truth Data Thematic Map Damage Class Producer's Accuray : 66% of damage pixels are correctly classified as damage by the classifier User's Accuracy: 50% of pixels classified as damage in thematic map is actually damage in the ground.

21 LITERATURE

22 Damage Map obtain with Multi-temporal Approach: Haiti Case Taşkın Kaya, G., and et al., Change detection in very high resolution imagery based on dynamic time warping: An implementation for Haiti earthquake damage assessment, Analysis of Multi-temporal Remote Sensing Images (Multi-Temp), th International Workshop on the, Pre-earthquake image Post-earthquake image Damage Map Change map

23 Damage Map obtain with Multi-temporal Approach: Adapazarı case Taşkın Kaya, G., Support vector selection and adaptation and its application to earthquake damage assessment, PhD. Thesis, Computational Science & Engineering, ITU, Pre-earthquake image Post-earthquake image

24 Damage Map obtain with Multi-temporal Approach: Adapazarı case Post earthquake image The damage map Urban Damage Vegetation

25 Post-earthquake damage assessment Ta kın Kaya, G., and et al., Damage Assessment of 2010 Haiti Earthquake with Post-Earthquake Satellite Image by Support Vector Selection and Adaptation, Photogrammetric Engineering & Remote Sensing, 77(10), Satellite image after the event Haiti 1 km Damage map Corlx Des Bossaies Region-1 Nazon Saint Antoine Earthquake epicenter Mome A Tuff Port-au-Prince National Palae Region to be classified Region-2 Turgeau Haiti Port-au prince earthquake damage assessment Saint-Gerard 85 to 100% damage 60 to 85% damage 20 to 60% damage 0 to %20 damage

26 Junho, Yeom and et al., Building Damage Assessment using High Resolution Satellite Imagery for Port-au-Prince, Haiti, Disaster Advances, 5, 2012 Before After Damage map

27 Acqua, Fabio Dell and et al., Earthquake Damages Rapid Mapping by Satellite Remote Sensing Data : L Aquila April 6th, 2009 Event, Haiti, IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING,4(4), It shows that optical data are more suitable to distinguish between damage and nondamage classes, while SAR textures features allow to better distinguishing different classes of damages at block scale such as low and heavy damage. Damage map Ground truth

28 Gang and et al., Damage information derived from multi-sensor data of the Wenchuan Earthquake of May 2008, International Journal of Remote Sensing, 31(13), 2010.

29 Ratjhe E. M and et al., Damage Patterns from Satellite Images of the 2003 Bam, Iran, Earthquake, Earthquake Spectra, 21(S1), S ,2005.

30 Chiroiu L., Damage Assessment of the 2003 Bam, Iran, Earthquake Using Ikonos Imagery, Earthquake Spectra, 21(S1), S , 2005.

31 Yamazaki F., Visual Damage Interpretation of Buildings in Bam City Using QuickBird Images Following the 2003 Bam, Iran, Earthquake, Earthquake Spectra, 21(S1), S , 2005.

32 Rezaeian M., Assessment Of Earthquake Damages By Image-based Techniques, PhD. Thesis, ETH Zurih, 2010.

33

34 L. Dong, J. Shan / ISPRS Journal of Photogrammetry and Remote Sensing 84 (2013) Types of data applied for uilding damage detection Studies those using both pre- and post-earthquake data Studies those using only post-earthquake data Using ancillary data Optical and LiDAR data Optical and SAR data LIDAR SAR Optical Image Year Fig. 1. Selected representative studies since 1998 in terms of data used.