Ioanna Ilia

Applying weight of evidence method and sensitivity analysis to produce a landslide susceptibility map

The main purpose of this study is to define the main variables that contribute to the occurrence of landslides in Kimi, Euboea, Greece, and to produce a landslide susceptibility map using the weight of evidence method. For the developed model, a sensitivity analysis is carried out in order to understand the model’s behavior when small changes are introduced in the weight value of the landslide-related variables. Landslide locations were identified from field surveys and interpretation of aerial photographs which resulted in the construction of an inventory map with 132 landslide events, while eight contributing variables were identified and exploited. All landslide-related variables were converted into a 5 × 5-m float-type raster file. These input-raster layers included a lithological unit layer, an elevation layer, a slope angle layer, a slope aspect layer, a distance from tectonic features layer, a distance from hydrographic network layer, a topographic wetness index layer, and a curvature layer. The validation of the developed model was achieved by using a subset of unprocessed landslide data, showing a satisfactory agreement between the expected and existing landslide susceptibility level, with the area under the predictive rate curve estimated to be 0.808. The area under the success rate curve was estimated to be 0.828 indicating a very high classification rate for existing landslide areas. According to the results of the sensitivity analysis, the lithological unit “yellowish gray to white marls” was the most sensitive as it had the highest change in the relative frequency of observed landslides. The overall outcomes of the performed analysis provide crucial knowledge in successful land use planning and management practice and also in risk reduction projects.

Case Event System for Landslide Susceptibility Analysis

Landslides are considered as a geological disaster that has an unfavourable effect on lives and properties, generating both direct and indirect economic and human losses every year. Compared to other geological disasters, landslides are considerably smaller in scale, more dispersed, but more disastrous in many cases. The presented methodology is based on a case–event system, which uses spatial analysis functions and artificial intelligence techniques, to evaluate potential instability problems concerning natural or artificial slopes. The methodology allows the user to examine new cases or areas of interest and compares them to previously recorded cases of instability problems that occur in the research area. The effectiveness of the methodology is evaluated in Kimi, Euboea, Greece, an area experienced substantial landslide events, where a well documented database of previous studies existed.

Mineralogy and technical properties of clayey diatomites from north and central Greece

Two bulk samples of clayey diatomite of Upper Miocene age originated from Western Macedonia, northern Greece and Thessaly central Greece were examined for their efficiency to be used as industrial absorbents. The samples were characterized using X-Ray Diffraction, Thermo-Gravimetric and Fourier Transform Spectroscopy, Scanning Electron Microscopy and ICP-MS analytical methods. The absorption capability of the clayey samples in oil and water were also examined. The mineralogy of both samples is predominated by the presence of clay minerals and amorphous silica. The clay minerals prevailed in the Klidi (KL) bulk sample, with muscovite being the dominant phase, and kaolinite and chlorite occurring in minor amounts. In the Drimos (DR) bulk sample, vermiculite was the predominant clay phase. Smectite was not found in either sample, whereas detrital quartz and feldspars were present in significant amounts. The amorphous silica phase (opal-A) occurs mainly with the form of disck-shaped diatom frustules. The chemistry of the samples is characterized by the predominance of silica, alumina, and iron, whereas all the other major and the trace elements are in low concentrations. Both clayey diatomite rocks exhibited sufficiently good oil and water absorption capacity, ranging between 70 to 79% in the clay-rich sample KL and 64 to 70% in the opal-A-rich sample DR. Comparing the properties of the rocks studied with other commercial absorbents, it is concluded that they may find applications as absorbents in industrial uses.

Application of fuzzy weight of evidence and data mining techniques in construction of flood susceptibility map of Poyang County, China

In China, floods are considered as the most frequent natural disaster responsible for severe economic losses and serious damages recorded in agriculture and urban infrastructure. Based on the international experience prevention of flood events may not be completely possible, however identifying susceptible and vulnerable areas through prediction models is considered as a more visible task with flood susceptibility mapping being an essential tool for flood mitigation strategies and disaster preparedness. In this context, the present study proposes a novel approach to construct a flood susceptibility map in the Poyang County, JiangXi Province, China by implementing fuzzy weight of evidence (fuzzy-WofE) and data mining methods. The novelty of the presented approach is the usage of fuzzy-WofE that had a twofold purpose. Firstly, to create an initial flood susceptibility map in order to identify non-flood areas and secondly to weight the importance of flood related variables which influence flooding. Logistic Regression (LR), Random Forest (RF) and Support Vector Machines (SVM) were implemented considering eleven flood related variables, namely: lithology, soil cover, elevation, slope angle, aspect, topographic wetness index, stream power index, sediment transport index, plan curvature, profile curvature and distance from river network. The efficiency of this new approach was evaluated using area under curve (AUC) which measured the prediction and success rates. According to the outcomes of the performed analysis, the fuzzy WofE-SVM model was the model with the highest predictive performance (AUC value, 0.9865) which also appeared to be statistical significant different from the other predictive models, fuzzy WofE-RF (AUC value, 0.9756) and fuzzy WofE-LR (AUC value, 0.9652). The proposed methodology and the produced flood susceptibility map could assist researchers and local governments in flood mitigation strategies.

A Geographical Information System (GIS) Based Probabilistic Certainty Factor Approach in Assessing Landslide Susceptibility: The Case Study of Kimi, Euboea, Greece

Landslides are referred to as unexpected and unpredictable movements usually on unstable surface layers making them one of the most frequent natural hazards with significant social—economic consequences and human losses. Understanding and dealing with landslide hazards requires geographically—referenced data that may vary in scale, resolution, reliability and come from a different set of sources. These raw spatial data needs to be organized and processed in order to support decision making and produce information for further study. Geographical Information Systems (GIS) is a set of tools and techniques that manage data and information by overlaying, quantifying, synthesizing them. The present paper considers the development and use of a GIS based Probabilistic Certainty Factor method to assess the geo—environmental parameters that influence the manifestation of landslide phenomena in order to produce a landslide susceptibility map, in the area of Kimi, Euboea, Greece. Certainty Factor method was implemented to evaluate the interaction between these parameters and the landslide occurrence, in order to highlight their contribution to landslide susceptibility.

Applying genetic algorithms to set the optimal combination of forest fire related variables and model forest fire susceptibility based on data mining models. The case of Dayu County, China

The main objective of the present study was to utilize Genetic Algorithms (GA) in order to obtain the optimal combination of forest fire related variables and apply data mining methods for constructing a forest fire susceptibility map. In the proposed approach, a Random Forest (RF) and a Support Vector Machine (SVM) was used to produce a forest fire susceptibility map for the Dayu County which is located in southwest of Jiangxi Province, China. For this purpose, historic forest fires and thirteen forest fire related variables were analyzed, namely: elevation, slope angle, aspect, curvature, land use, soil cover, heat load index, normalized difference vegetation index, mean annual temperature, mean annual wind speed, mean annual rainfall, distance to river network and distance to road network. The Natural Break and the Certainty Factor method were used to classify and weight the thirteen variables, while a multicollinearity analysis was performed to determine the correlation among the variables and decide about their usability. The optimal set of variables, determined by the GA limited the number of variables into eight excluding from the analysis, aspect, land use, heat load index, distance to river network and mean annual rainfall. The performance of the forest fire models was evaluated by using the area under the Receiver Operating Characteristic curve (ROC-AUC) based on the validation dataset. Overall, the RF models gave higher AUC values. Also the results showed that the proposed optimized models outperform the original models. Specifically, the optimized RF model gave the best results (0.8495), followed by the original RF (0.8169), while the optimized SVM gave lower values (0.7456) than the RF, however higher than the original SVM (0.7148) model. The study highlights the significance of feature selection techniques in forest fire susceptibility, whereas data mining methods could be considered as a valid approach for forest fire susceptibility modeling.

The Use of a Spatial Multi—Criteria Technique for Urban Suitability Assessment, Due to Extensive Mass Movements. The Case Study of Vitala Village, Kimi, Euboea, Greece

The present study illustrates a spatial multi—criteria analysis technique for Urban Suitability Assessment for the area of Vitala village, which is located in the municipality of Kimi, Euboea, Greece. The extend and severity of the landslide phenomena encountered in the research area, made clear that any mitigation measure will fail since the reactivation of the landslide phenomena is mostly certain. Therefore, the main objective of the present study was to investigate in detail the geo—morphological, geo—dynamic and hydro—geological conditions in Vitala village in order to classify the area into the following categories: (a) areas suitable for urban development (b) areas suitable with some restrictions (c) areas unsuitable for urban development and (d) restricted areas. The evaluation incorporates topography, surface and bedrock geology, groundwater conditions and tectonic features as the most critical parameters.

Land subsidence phenomena investigated by spatiotemporal analysis of groundwater resources, remote sensing techniques, and random forest method: the case of Western Thessaly, Greece

The main objective of the present study was to investigate land subsidence phenomena and the spatiotemporal pattern of groundwater resources in an area located in western Thessaly, Greece, by using remote sensing techniques and data mining methods. Specifically, the nonparametric Mann–Kendall test and the Sen’s slope estimator were used to estimate the trend concerning the groundwater table, whereas a set of Synthetic Aperture Radar images, processed with the Persistent Scatterer Interferometry technique, were used investigate the spatial and temporal patterns of ground deformation. Random forest (RF) method was utilized to predict the subsidence deformation rate based on three related variables, namely: thickness of loose deposits, the Sen’s slope value of groundwater-level trend, and the Compression Index of the formation covering the area of interest. The outcomes of the study suggest a strong correlation among the thickness of the loose deposits and the deformation rate and also show that a clear trend between the deformation rate and the fluctuation of the groundwater table exists. For the RF model and based on the validation dataset, the r square value was calculated to be 0.7503. In the present study, the potential deformation rate assuming different water pumping scenarios was also estimated. It was observed that with a mean decrease in the Sen’s slope value of groundwater-level trend of 20%, there would be a mean decrease of 9.01% in the deformation rate, while with a mean increase in the Sen’s slope value of groundwater-level trend of 20%, there would be a mean increase of 12.12% in the deformation rate. The ability of identifying surface deformations allows the local authorities and government agencies to take measures before the evolution of severe subsidence phenomena and to prepare for timely protection of the affected areas.

Developing a landslide susceptibility map based on remote sensing, fuzzy logic and expert knowledge of the Island of Lefkada, Greece

The main objective of the study was to develop a novel expert-based approach in order to construct a landslide susceptibility map for the Island of Lefkada, Greece. The developed methodology was separated into two actions. The first action involved the construction of a landslide inventory map and the second the exploitation of expert knowledge and the use of fuzzy logic to produce a landslide susceptibility map. Two types of movements were analyzed: rapid moving slides that involve rock falls and rock slides and slow to very slow moving slides. The landslide inventory map was constructed through an evaluation procedure that involved the use of a group of experts, who analyzed data acquired from remote sensing techniques supplemented by landslide records and fieldwork data. During the second action an expert-driven model was developed for identifying the tendency of landslide occurrences concerning both types of movements. A set of casual variables was selected, namely: lithological units, slope angle, slope orientation, distance from tectonic features, distance from hydrographic network and distance from road network. The performance and validation of the developed model were compared with models that are constructed on the bases of each expert’s judgment. The results proved that the most accurate and reliable outcomes are obtained from the aggregated values assigned by the group of experts and not from the individual values assigned by each expert. The area under the receiver operating characteristic curves for the models constructed by the expert’s group was 0.873 for prediction curves of rapid moving slides and 0.812 for prediction rate curves of slow to very slow moving slides, respectively. These values were much higher than those obtained by each expert. From the outcomes of the study it is clear that the produced landslide susceptibility maps could provide valuable information during landslide risk assessments at the Island of Lefkada.

Comparing the Performance of a Logistic Regression and a Random Forest Model in Landslide Susceptibility Assessments. the Case of Wuyaun Area, China

The main objectives of the study was to apply a Logistic Regression and a Random Forest model for the construction of a landslide susceptibility map in the Wuyuan area, China, and to compare their results by performing non-parametric and linear regression analysis. Thirteen landslide variables were analyzed, namely: lithology, soil, slope, aspect, altitude, topographic wetness index, stream power index, stream transport index, plan curvature, profile curvature, distance to roads, distance to rivers and distance to faults, while 255 sites classified as landslide and 255 sites classified as non-landslide were separated into a training dataset (70%) and a validation dataset (30%). The comparison and validation of the outcomes of each model were achieved using statistical evaluation measures, the receiving operating characteristic and the area under the success and prediction rate curves. The presence of linear correlation between the two models was estimated by performing a simple linear regression analysis. The most accurate model was Random Forest, which identified correctly 98.32% of the instances during the training phase, followed by Logistic Regression (87.43%). During the validation phase, the Random Forest achieved a classification accuracy of 85.52%, while Logistic Regression model achieved an accuracy of 80.92%. The area under the success and prediction rate curves for the Random Forest were calculated to be 0.9805 and 0.9324, respectively, while the Logistic Regression model showed as slightly lower predictive performance, 0.9372 and 0.8903 respectively. Finally, by performing a non-parametric analysis, the two models were found to be significantly different. Strong evidence of linear relationship between the two models exist, having a p-value less than 0.0001 at a 95% confidence level and an R2 value estimated to be 0.6993 indicating that 69.93% of the variability in the Logistic Regression model can be explained by variation in the Random Forest model.