Procjena podložnosti na klizanje primjenom različitih statističkih metoda (Gornja Dubrava, Grad Zagreb)

master's thesis

• Title: Procjena podložnosti na klizanje primjenom različitih statističkih metoda (Gornja Dubrava, Grad Zagreb) : diplomski rad
• Title (english): Assessment of landslide susceptibility using different statistical methods (Gornja Dubrava, City of Zagreb) : master's thesis
• Author: Ivan Melkić
• Mentor: Sanja Bernat Gazibara (mentor)
• Mentor: Marko Sinčić (pomoć pri izradi, neposredni voditelj)
• Committee member: Sanja Bernat Gazibara (predsjednik povjerenstva)
• Committee member: Martin Krkač (član povjerenstva)
• Committee member: Snježana Mihalić Arbanas (član povjerenstva)
• Granter: University of Zagreb; Faculty of Mining, Geology and Petroleum Engineering; (Department of Geology and Geological Engineering); Zagreb
• Defense date and country: 2024-09-30, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Mining, Petroleum and Geology Engineering; Geological Engineering

Abstract

Tema diplomskog rada je analiza utjecaja različitih statističkih metoda na rezultate procjene podložnosti na klizanje na pilot području površine 19,3 km2 u Donjoj Dubravi u Zagrebu. Metode korištene za procjenu podložnosti na klizanje su bile Weight of Evidence, neuralna mreža, slučajne šume i logistička regresija. Ulazni podaci korišteni za procjenu podložnosti na klizanje su inventar klizišta dobiven interpretacijom LiDAR DMR-a, te geomorfološki, geološki, hidrološki i antropogeni preduvjeti klizanja, njih ukupno devet. Temeljem kvalitativnih i kvantitativnih usporedbi dobivenih karata podložnosti na klizanje utvrđeno je da postoje razlike u pojedinim metodama analize podložnosti na klizanje. Bivarijantna statistička metoda, Weight of Evidence, i metoda logističke regresije pokazale su veću objektivnost i podjednaku točnost modela na podacima za trening i validaciju modela, a metode neuralne mreže i slučajnih šuma pokazale su pretreniranost modela, sa znatno nižim vrijednostima prilikom validacije modela. Može se zaključiti da je izbor najoptimalnije statističke metode za analizu podložnosti na klizanje i dalje otvoreno pitanje, ali sama pouzdanost modela svakako najviše ovisi o kvaliteti i prostornoj točnosti ulaznih podataka.

Abstract (english)

The topic of this thesis is the analysis of the influence of different statistical methods on the results of landslide susceptibility assessment in a pilot area of 19.3 km² in Donja Dubrava, Zagreb. The methods used for landslide susceptibility assessment were Weight of Evidence, neural networks, random forests, and logistic regression. The input data for assessing landslide susceptibility included a landslide inventory derived from the interpretation of LiDAR DMR, as well as geomorphological, geological, hydrological, and anthropogenic landslide preconditions, nine in total. Based on qualitative and quantitative comparisons of the generated landslide susceptibility maps, it was determined that differences exist between the various methods of landslide susceptibility analysis. The bivariate statistical method, Weight of Evidence, and logistic regression method demonstrated greater objectivity and similar accuracy in training and validation datasets, while the neural network and random forest methods exhibited model overfitting, with significantly lower values of model validation. It can be concluded that the selection of the most optimal statistical method for landslide susceptibility analysis remains an open question, but the reliability of the model primarily depends on the quality and spatial accuracy of the input data.

• Keywords: karta podložnosti na klizanje; klizište; statističke metode; strojno učenje; GIS
• Keywords (english): landslide susceptibility map; landslide; statistical methods; machine learning; GIS
• Language: croatian
• URN:NBN: urn:nbn:hr:169:682843
• Study programme: Title: Geological engineering; specializations in: Hydrogeology and engineering geology, Environmental geology; Course: Environmental geology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka geološkog inženjerstva (magistar/magistra inženjer/inženjerka geološkog inženjerstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2025-04-22 08:42:34