Machine Learning Feature Selection Methods for Landslide Susceptibility Mapping

This paper explores the use of adaptive support vector machines, random forests and AdaBoost for landslide susceptibility mapping in three separated regions of Canton Vaud, Switzerland, based on a set of geological, hydrological and morphological features. The feature selection properties of the three algorithms are studied to analyze the relevance of features in controlling the spatial distribution of landslides. The elimination of irrelevant features gives simpler, lower dimensional models while keeping the classification performance high. An object-based sampling procedure is considered to reduce the spatial autocorrelation of data and to estimate more reliably generalization skills when applying the model to predict the occurrence of new unknown landslides. The accuracy of the models, the relevance of features and the quality of landslide susceptibility maps were found to be high in the regions characterized by shallow landslides and low in the ones with deep-seated landslides. Despite providing similar skill, random forests and AdaBoost were found to be more efficient in performing feature selection than adaptive support vector machines. The results of this study reveal the strengths of the classification algorithms, but evidence: (1) the need for relying on more than one method for the identification of relevant variables; (2) the weakness of the adaptive scaling algorithm when used with landslide data; and (3) the lack of additional features which characterize the spatial distribution of deep-seated landslides.     

Salt damage and microclimate in the Postumius Tomb, Roman Necropolis of Carmona, Spain

The necropolis of Carmona (Seville, Spain) is one of the most significant Roman burial sites in southern Spain used during the first and second centuries ad. Of its more than 600 tombs, the Postumius Tomb is one of the best examples of a tomb affected by severe salt damage. To define safe microclimatic conditions for its conservation, environmental parameters were recorded from June 2007 to April 2009, both inside and outside the tomb, and mineralogical, textural, petrophysical, and durability characterization studies of the host-rock were made. Experimental tests revealed a high susceptibility to salt deterioration of a host-rock (calcarenite) with low mechanical properties and a complex porous medium that favors salt weathering, water condensation, and capillary rise. The analysis of the weathered material showed the presence chiefly of gypsum (CaSO₄·2H₂O), thenardite (Na₂SO₄) and halite (NaCl) in the tomb of Postumius, with alteration that was more intensive in spring and autumn, and less so during summer months. Salt damage activity was calculated by quantifying the number of transitions of crystallization–dissolution of saline phases. The calculated seasonality for water condensation and salt damage is coeval. The host-rock alteration is in accord with the estimated salt decay, and was more intensive in spring and autumn and less so during summer. The seasonality of halite transitions is similar to that of the sodium sulfate system, which suggests that salt weathering is produced by the two types of salts. By combining different methodological approaches (pore structure, water condensation, salt and environmental conditions), it is possible to explain why salt crystallization occurs in a tomb with hygrometric conditions that are not suitable for this process to occur. These methodological approaches are also used to other rock-decaying processes, such as the development of microorganisms, clay swelling and calcite dissolution by NaCl- and CO₂-rich pore waters, and can be used to predict safe threshold microclimatic conditions that minimize all rock-decaying processes.     

Distribution of trace elements in the coastal sea sediments of Maslinica Bay, Croatia

Spatial distributions of trace elements in the coastal sea sediments and water of Maslinica Bay (Southern Adriatic), Croatia and possible changes in marine flora and foraminifera communities due to pollution were investigated. Macro, micro and trace elements’ distributions in five granulometric fractions were determined for each sediment sample. Bulk sediment samples were also subjected to leaching tests. Elemental concentrations in sediments, sediment extracts and seawater were measured by source excited energy dispersive X-ray fluorescence (EDXRF). Concentrations of the elements Cr, Cu, Zn, and Pb in bulk sediment samples taken in the Maslinica Bay were from 2.1 to over six times enriched when compared with the background level determined for coarse grained carbonate sediments. A low degree of trace elements leaching determined for bulk sediments pointed to strong bonding of trace elements to sediment mineral phases. The analyses of marine flora pointed to higher eutrophication, which disturbs the balance between communities and natural habitats.     

Tomographic inversion using ℓ1-norm regularization of wavelet coefficients

We propose the use of ℓ₁ regularization in a wavelet basis for the solution of linearized seismic tomography problems   A m = d   , allowing for the possibility of sharp discontinuities superimposed on a smoothly varying background. An iterative method is used to find a sparse solution m that contains no more fine-scale structure than is necessary to fit the data d to within its assigned errors.     

Geochemical characterisation, provenance, source and depositional environment of ‘Roches Argilo-Talqueuses’ (RAT) and Mines Subgroups sedimentary rocks in the Neoproterozoic Katangan Belt (Congo): Lithostratigraphic implications

The chemical characteristics of sedimentary rocks provide important clues to their provenance and depositional environments. Chemical analyses of 192 samples of Katangan sedimentary rocks from Kolwezi, Kambove–Kabolela and Luiswishi in the central African Copperbelt (Katanga, Congo) are used to constrain (1) the source and depositional environment of RAT and Mines Subgroup sedimentary rocks and (2) the geochemical relations between the rocks from these units and the debate on the lithostratigraphic position of the RAT Subgroup within the Katangan sedimentary succession. The geochemical data indicate that RAT, D. Strat., RSF and RSC are extremely poor in alkalis and very rich in MgO. SD are richer in alkalis, especially K₂O. Geochemical characteristics of RAT and Mines Subgroups sedimentary rocks indicate deposition under an evaporitic environment that evolved from oxidizing (Red RAT) to reducing (Grey RAT and Mines Subgroup) conditions. There is no chemical difference between RAT and fine-grained clastic rocks from the lower part of the Mines Subgroup. The geochemical data preclude the genetic model that RAT are syn-orogenic sedimentary rocks originating from Mines Group rocks by erosion and gravity-induced fragmentation in front of advancing nappes.     

Learning wind fields with multiple kernels

This paper presents multiple kernel learning (MKL) regression as an exploratory spatial data analysis and modelling tool. The MKL approach is introduced as an extension of support vector regression, where MKL uses dedicated kernels to divide a given task into sub-problems and to treat them separately in an effective way. It provides better interpretability to non-linear robust kernel regression at the cost of a more complex numerical optimization. In particular, we investigate the use of MKL as a tool that allows us to avoid using ad-hoc topographic indices as covariables in statistical models in complex terrains. Instead, MKL learns these relationships from the data in a non-parametric fashion. A study on data simulated from real terrain features confirms the ability of MKL to enhance the interpretability of data-driven models and to aid feature selection without degrading predictive performances. Here we examine the stability of the MKL algorithm with respect to the number of training data samples and to the presence of noise. The results of a real case study are also presented, where MKL is able to exploit a large set of terrain features computed at multiple spatial scales, when predicting mean wind speed in an Alpine region.     

Can a Simulated Flood Experience Improve Social Resilience to Disasters?

The increase of extreme meteorological phenomena, along with continuous population growth, has led to a rising number of flooding disasters. Therefore, there is an urgent need to develop better risk reduction strategies, among which is increased social resilience. Experiencing a disaster is recognized as a factor that positively influences overall community resilience, with particular effects on social resilience; it appears to be more influential than school education. It also has many negative effects, though. Previous studies underline that citizens do not distinguish between different types of experiences. Thus, we investigated whether a simulated experience of a flood can improve social resilience, without being hampered by negative repercussions. The study was executed in five municipalities in three Italian regions involved in the European project LIFE PRIMES, which planned simulation activities for each studied area. Data, collected through the administration of anonymous questionnaires before and after a flood drill, were processed by applying a multicriteria decision analysis tool (PROMETHEE). Results show that the drill significantly augmented perceived social resilience in the smaller studied communities but not in the larger ones, a fact that should be further investigated. Key Words: multicriteria decision analysis, simulated flood experience, social resilience to disasters.