Landslide susceptibility modeling using bivariate statistical analysis methods for Podsljeme area of the City of Zagreb
| Autor: | Halapir, Ivan |
|---|---|
| Přispěvatelé: | Mihalić Arbanas, Snježana, Bernat Gazibara, Sanja, Sinčić, Marko, Arbanas, Željko, Krkač, Martin |
| Jazyk: | chorvatština |
| Rok vydání: | 2022 |
| Předmět: | |
| Popis: | Cilj ovog diplomskog rada bila je izrada modela podložnosti na klizanje za dio podsljemenske zone Grada Zagreba korištenjem bivarijantnih statističkih metoda analize podložnosti na klizanje, točnije metode Weight of Evidence i metode informacijske vrijednosti. Izrađeno je ukupno 18 modela podložnosti na klizanje prema predviđenim scenarijima. Scenariji se razlikuju s obzirom na prostornu rezoluciju korištenih faktorskih karata te prostornu rezoluciju i način geometrijskog prikaza klizišta iz inventara. Faktorske karte korištene za izradu modela podložnosti na klizanje bile su karte nagiba terena, karte raščlanjenosti terena, karte stratigrafskih jedinica, karte udaljenosti od geoloških granica, karte udaljenosti od drenažne mreže i karte namjene zemljišta prema tipu pokrova, prostornih rezolucija 1 m, 2 m i 5 m. U izradi modela podložnosti na klizanje korišteni su inventari klizišta prostornih rezolucija 1 m, 2 m i 5 m, pri čemu su klizišta bila prikazana poligonima, centroidima ili točkama na glavnim pukotinama. Nakon što su izrađeni svi modeli podložnosti na klizanje pristupilo se određivanju stupnja točnosti i predikcije pojedinih modela pomoću analize ROC krivuljom. Usporedbom rezultata analize ROC krivuljom utvrđeno je da svi modeli podložnosti na klizanje daju vrlo dobre rezultate, ali model podložnosti na klizanje prema scenariju S12 daje kombinaciju najviših stupnjeva točnosti (92%) i predikcije (92%) područja istraživanja. Navedeni model dobiven je primjenom metode informacijske vrijednosti i korištenjem faktorskih karta prostorne rezolucije 1 m te inventara klizišta iste prostorne rezolucije u kojem su klizišta prikazana točkama na glavnim pukotinama. The objective of this thesis was to create landslide susceptibility models for the Podsljeme area of the City of Zagreb using bivariate statistical analysis methods, specifically the Weight of Evidence method and Information Value method. A total of 18 landslide susceptibility models were created according to the predetermined scenarios. Each scenario is defined by the spatial resolution of the used factor maps and the spatial resolution and type of geometric representation of the landslides from the inventory map. The factor maps used to develop the landslide susceptibility models were in three different resolutions (1 m, 2 m and 5 m) as follows: slope maps, terrain breakdown, stratigraphic units, distance from geological contacts, distance from the drainage network and land use according to cover type. Landslide inventories with spatial resolutions of 1 m, 2 m and 5 m were used in the creation of the landslide susceptibility models, in which the landslides were represented either by polygons, centroids or points on the main cracks. After all the landslide susceptibility models were created, the levels of accuracy and prediction of individual models were determined using the ROC curve analysis. By comparing the results of the ROC curve analysis, it was determined that all models of landslide susceptibility give very good results, but the model of landslide susceptibility according to scenario S12 gives a combination of the highest degrees of accuracy (92%) and prediction (92%) of the research area. The mentioned model was obtained by applying the Information Value method and using factor maps with a spatial resolution of 1 m and landslide inventory of the same spatial resolution, in which landslides are represented by points on the main cracks. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|