Fakultät für Geowissenschaften - Digitale Hochschulschriften der LMU

Risk and vulnerability assessment for natural hazards is of high interest. Various methods focusing on building vulnerability assessment have been developed ranging from simple approaches to sophisticated ones depending on the objectives of the study, the availability of data and technology. In-situ assessment methods have been widely used to measure building vulnerability to various types of hazards while remote sensing methods, specifically developed for assessing building vulnerability to tsunami hazard, are still very limited. The combination of remote sensing approaches with in-situ methods offers unique opportunities to overcome limitations of in-situ assessments.

The main objective of this research is to develop remote sensing techniques in assessing building vulnerability to tsunami hazard as one of the key elements of risk assessment. The research work has been performed in the framework of the GITEWS (German-Indonesian Tsunami Early Warning System) project. This research contributes to two major components of tsunami risk assessment: (1) the provision of infrastructure vulnerability information as an important element in the exposure assessment; (2) tsunami evacuation modelling which is a critical element for assessing immediate response and capability to evacuate as part of the coping capacity analysis.

The newly developed methodology is based on the combination of in-situ measurements and remote sensing techniques in a so-called “bottom-up remote sensing approach”. Within this approach, basic information was acquired by in-situ data collection (bottom level), which was then used as input for further analysis in the remote sensing approach (upper level). The results of this research show that a combined in-situ measurement and remote sensing approach can be successfully employed to assess and classify buildings into 4 classes based on their level of vulnerability to tsunami hazard with an accuracy of more than 80 percent. Statistical analysis successfully revealed key spatial parameters which were regarded to link parameters between in-situ and remote sensing approach such as size, height, shape, regularity, orientation, and accessibility. The key spatial parameters and their specified threshold values were implemented in a decision tree algorithm for developing a remote sensing rule-set of building vulnerability classification. A big number of buildings in the study area (Cilacap city, Indonesia) were successfully classified into the building vulnerability classes. The categorization ranges from high to low vulnerable buildings (A to C) and includes also a category of buildings which are potentially suitable for tsunami vertical evacuation (VE).

A multi-criteria analysis was developed that incorporates three main components for vulnerability assessment: stability, tsunami resistance and accessibility. All the defined components were configured in a decision tree algorithm by applying weighting, scoring and threshold definition based on the building sample data. Stability components consist of structure parameters, which are closely related to the building stability against earthquake energy. Building stability needs to be analyzed because most of tsunami events in Indonesia are preceded by major earthquakes. Stability components analysis was applied in the first step of the newly developed decision tree algorithm to evaluate the building stability when earthquake strikes. Buildings with total scores below the defined threshold of stability were classified as the most vulnerable class A. Such the buildings have a high probability of being damaged after earthquake events. The remaining buildings with total scores above the defined threshold of stability were further analyzed using tsunami components and accessibility components to classify them into the vulnerability classes B, C and VE respectively.

This research is based on very high spatial resolution satellite images (QuickBird) and object-based image analysis.

The research focuses on providing reliable spatial information in support of tsunami risk and vulnerability assessment within the framework of the German-Indonesian Tsunami Early Warning System (GITEWS) project. It contributes to three major components of the project: (1) the provision of spatial information on surface roughness as an important parameter for tsunami inundation modeling and hazard assessment; (2) the modeling of population distribution, which is an essential factor in tsunami vulnerability assessment and local disaster management activities; and (3) the settlement detection and classification from remote sensing radar imagery to support the population distribution research.

Regarding the surface roughness determination, research analyses on surface roughness classes and their coefficients have been conducted. This included the development of remote sensing classification techniques to derive surface roughness classes, and integration of the thus derived spatial information on surface roughness conditions to tsunami inundation modeling. This research determined 12 classes of surface roughness and their respective coefficients based on analyses of published values.

The developed method for surface roughness classification of remote sensing data considered density and neighborhood conditions, and resulted in more than 90% accuracy. The classification method consists of two steps: main land use classification and density and neighborhood analysis. First, the main land uses were defined and a classification was performed applying decision tree modeling. Texture parameters played an important role in increasing the classification accuracy. The density and neighborhood analysis further substantiated the classification result towards identifying surface roughness classes. Different classes such as residential areas and trees were combined to new surface roughness classes, as “residential areas with trees”. The density and neighborhood analysis led to an appropriate representation of real surface roughness conditions. This was used as an important input for tsunami inundation modeling.

By using Tohoku University’s Analysis Model for Investigation Near-field Tsunami Number 3 (TUNAMI N3), the spatially distributed surface roughness information was integrated in tsunami inundation modeling and compared to the modeling results applying a uniform surface roughness condition. An uncertainty analysis of tsunami inundation modeling based on the variation of surface roughness coefficients in the Cilacap study area was also undertaken. It was demonstrated that the inundation modeling results applying uniform and spatially distributed surface roughness resulted in high differences of inundation lengths, especially in areas far from the coastline. This result showed the important role of surface roughness conditions in resisting tsunami flow, which must be considered in tsunami inundation modeling.

With respect to the second research focus, the population distribution, a concept of population distribution modeling was developed. Within the modeling process, weighting factor determination, multi-scale disaggregation and a comparative study to other methods were conducted. The basis of the developed method was a combination of census and land use data, which led to an improved spatial resolution and accuracy of the population distribution. Socio-economic data were used to derive weighting factors to distributing people to land use classes. Moreover, in case of missing input data, an approach was developed that allows for the determination of generalized weighting factors. The approach to use specific weightings, where possible and generalized ones, where necessary, led to a flexible methodology with respect to the achievable accuracy and availability of data. A comparative study was performed by comparing this new model with previously developed population distribution models. The newly developed model showed a higher accuracy.

The

This thesis is an outcome of the project “Flood and damage assessment using very high resolution SAR data” (SAR-HQ), which is embedded in the interdisciplinary oriented RIMAX (Risk Management of Extreme Flood Events) programme, funded by the Federal Ministry of Education and Research (BMBF). It comprises the results of three scientific papers on automatic near real-time flood detection in high resolution X-band synthetic aperture radar (SAR) satellite data for operational rapid mapping activities in terms of disaster and crisis-management support.

Flood situations seem to become more frequent and destructive in many regions of the world. A rising awareness of the availability of satellite based cartographic information has led to an increase in requests to corresponding mapping services to support civil-protection and relief organizations with disaster-related mapping and analysis activities. Due to the rising number of satellite systems with high revisit frequencies, a strengthened pool of SAR data is available during operational flood mapping activities. This offers the possibility to observe the whole extent of even large-scale flood events and their spatio-temporal evolution, but also calls for computationally efficient and automatic flood detection methods, which should drastically reduce the user input required by an active image interpreter.

This thesis provides solutions for the near real-time derivation of detailed flood parameters such as flood extent, flood-related backscatter changes as well as flood classification probabilities from the new generation of high resolution X-band SAR satellite imagery in a completely unsupervised way. These data are, in comparison to images from conventional medium-resolution SAR sensors, characterized by an increased intra-class and decreased inter-class variability due to the reduced mixed pixel phenomenon. This problem is addressed by utilizing multi-contextual models on irregular hierarchical graphs, which consider that semantic image information is less represented in single pixels but in homogeneous image objects and their mutual relation. A hybrid Markov random field (MRF) model is developed, which integrates scale-dependent as well as spatio-temporal contextual information into the classification process by combining hierarchical causal Markov image modeling on automatically generated irregular hierarchical graphs with noncausal Markov modeling related to planar MRFs. This model is initialized in an unsupervised manner by an automatic tile-based thresholding approach, which solves the flood detection problem in large-size SAR data with small a priori class probabilities by statistical parameterization of local bi-modal class-conditional density functions in a time efficient manner.

Experiments performed on TerraSAR-X StripMap data of Southwest England and ScanSAR data of north-eastern Namibia during large-scale flooding show the effectiveness of the proposed methods in terms of classification accuracy, computational performance, and transferability. It is further demonstrated that hierarchical causal Markov models such as hierarchical maximum a posteriori (HMAP) and hierarchical marginal posterior mode (HMPM) estimation can be effectively used for modeling the inter-spatial context of X-band SAR data in terms of flood and change detection purposes. Although the HMPM estimator is computationally more demanding than the HMAP estimator, it is found to be more suitable in terms of classification accuracy. Further, it offers the possibility to compute marginal posterior entropy-based confidence maps, which are used for the generation of flood possibility maps that express that the uncertainty in labeling of each image element. The supplementary integration of intra-spatial and, optionally, temporal contextual information into the Markov model results in a reduction of classification errors. It is observed that the application of the hybrid multi-contextual Markov model on irregula

Although remote sensing has been widely used in geological investigations, the lithological classification of the area interested, based on medium-spatial and spectral resolution satellite data, is often not successful because of the complicated geological situation and other factors like inadequate methodology applied and wrong geological models.

The study area of the present thesis is located in southwest of the Prieska sub-basin, Transvaal Supergroup, South Africa. This area includes mainly Neoarchean and Proterozoic sedimentary rocks partly uncomfortably covered by uppermost Paleozoic and lower Mesozoic rocks and Tertiary to recent soils and sands. The Precambrian rocks include various formations of volcanic and intrusive rocks, quartzites, shales, platform carbonates and Banded Iron Formations (BIF). The younger rocks and soils include dikes and shales, glacial sedimentary rocks, coarser siliciclastic rocks, calcretes, aeolian and fluvial sands, etc. Prospect activity for mineral deposits necessitates the detailed geological map (1:100000) of the area.

In this research, a new rule-based classification system (RBS) was put forward, integrating spectral characteristics, textural features and ancillary data, such as general geological map (1:250000) and elevation data, in order to improve the lithological classification accuracy and the subsequent mapping accuracy in the study area. The proposed technique was mainly based on Landsat TM data and ASTER data with medium resolution. As ancillary data sets, topographic maps and general geological map were also available. Software like ERDAS©, Matlab©, and ArcGIS© supported the procedures of classification and mapping.

The newly developed classification technique was performed by three steps. Firstly, the geographic and atmospheric correction was performed on the original TM and ASTER data, following the principal component analysis (PCA) and band ratioing, to enhance the images and to obtain data sets like principal components (PCs) and ratio bands. Traditional maximum-likelihood supervised classification (MLC) was performed individually on enhanced multispectral image and principal components image (PCs-image). For TM data, the classification accuracy based on PCs-image was higher than that based on multispectral image. For ASTER data, the classification accuracy of PCs- image was close to but lower, than that of multispectral image. As one of the encountered Banded Iron Formations, the Griquatown Banded Iron Formation (G-BIF) was recognized well in TM-principal components image (PCs-image).

In the second step, textural features of different lithological types based on TM data were analyzed. Grey level co-occurrence matrix (GLCM) based textural features were computed individually from band 5 and the first principal component (PC1) of TM data. Geostatistics-based textural features were computed individually from the 6 TM multispectral bands and 3 principal components (PC1, PC2 and PC3). These two kinds of textural features were individually stacked as extra layers together with the original 6 multispectral bands and the 6 principal components to form several new data sets. Ratio bands were also individually stacked as extra layers with 6 multispectral bands and 6 principal components, to form other new data sets. In the same way new data sets were formed based on ASTER data. Then, all of the new data sets were individually classified using a maximum likelihood supervised classification (MLC), to produce several classified thematic images. The classification accuracy based on the new data sets are higher than that solely based on the spectral characteristics of original TM and ASTER data. It should be noticed that for one specific rock type, the class value in all classified images should correspond to its identified (ID) value in digital geological map.

The third step was to perform the rule-based system (RBS) classification. In the first part of the RBS, two classified imag