Aktualności

Seminarium Naukowe IGiG - Dr Sylwester Kamieniarz

14-03-2023

IGG ma przyjemność zaprosić Państwa na seminarium naukowe, które odbędzie się w formie wirtualnej na platformie Zoom. Prezentację zatytułowaną "Geostatistical methods and artificial neural networks for landslide hazard prediction – the example of Kraków city (southern Poland)" przedstawi dr Sylwester Kamieniarz z Państwowego Insystytutu Geologicznego.





Seminarium rozpocznie się w czwartek 16 marca 2023 r. o godzinie 09:00 AM (CEST).


Dołącz do spotkania Zoom: ZOOM LINK
Identyfikator spotkania: 811 427 0260
Kod dostępu: igig


Geostatistical methods and artificial neural networks for landslide hazard prediction – the example of Kraków city (southern Poland)
Dr Sylwester Kamieniarz, Państwowy Instytut Geologiczny


Abstrakt:
The aim of the study was to quantify a landslide hazard (H; understood as the probability of a landslide occurrence within a specified period of time and a given place) in an urbanized area with a complex geological conditions.
Landslides that were active in the years 1969 – 2019 were identified with the analysis of the numerical differential terrain models and available archival data (landslide registers, orthophotomaps, geological and engineering documentations, landslide monitoring reports, publications). Based on the observed correlation between the landslides distance and their activity, the technique of areal binomial kriging available in Geostatistical Wizard module (ArcGIS) was implemented to create a time probability model (TPM) of the landslides occurrence in the next fifty years. Due to the differentiation of landslide types in the study area, a multi-layer perceptron was used to determine the landslide susceptibility model (LSM). The learning process was performed in the r.landslide module using 8 thematic layers: slopes, slopes exposure, absolute height, relative height, convergence index, surface lithology, sub-Quaternary lithology, distance from tectonic discontinuities. The one-at-a-time sensitivity analysis of the thematic layers was also carried out. The landslide hazard map of Kraków was obtained by multiplying TPM and LSM. About 11% of Kraków area is covered by areas where the probability of a landslide occurrence in the years 2020 – 2070 exceeds the value of 0,2. The greatest hazard (H > 0,6) occurs in the southern part of the city. In the case of area located in the central part of the city, the hazard values seem to be slightly overestimated by the applied kriging algorithm. The sensitivity analysis showed that among the thematic layers used for modelling the slopes, convergence index, distance from tectonic discontinuities and the sub-Quaternary lithology have the greatest impact on the landslide occurrence in Kraków area.
An artificial neural networks are useful in determining the landslide susceptibility in areas with complex geological structure and topography, with various types of landslides. Geostatistical methods may lead to local overestimation of the time probability model values in the areas of scattered landslides, constituting single occurrences unrelated to the natural predispositions of a given area to the formation of mass movements.