Title Metodologija izrade karata klizišta korištenjem digitalnoga modela terena visoke rezolucije u podsljemenskoj zoni Grada Zagreba : doktorski rad Title (english) Methodology for landslide mapping using high resolution digital elevation model in the Podsljeme area (City of Zagreb) : doctoral dissertation Author Sanja Bernat Gazibara Mentor Snježana Mihalić Arbanas (mentor)Committee member Martin Krkač (predsjednik povjerenstva)Committee member Marko Komac Committee member Anja Vrbaški (član povjerenstva)Granter University of Zagreb Defense date and country 2019-05-10, Croatia Scientific / art field, TECHNICAL SCIENCES Universal decimal classificationUDC ) 55 - Geology. Meteorology. Hydrology Abstract U okviru disertacije prikazani su rezultati doktorskog istraživanja čija svrha je bila uspostavljanje metodologije za izradu karata klizišta primjenom tehnologije laserskog skeniranja (LiDAR). Osnovna prednost LiDAR-skih podataka u odnosu na ostale metode daljinskih istraživanja je izrada digitalnog modela terena (DMT) bez vegetacije i vrlo visoke rezolucije (<2 m) koji omogućava identifikaciju i kartiranje malih i plitkih klizišta na područjima pod gustom vegetacijom, kakvo je područje podsljemenske zone u Gradu Zagrebu. Za područje (ukupne površine 180 km2) provedeno je lasersko skeniranje iz zraka u prosincu 2013. godine, nakon događaja masovnog pokretanja klizišta uslijed niza ekstremnih oborinskih događaja, čime je omogućena izrada DMT-a s reprezentativnim brojem (re)aktiviranih klizišta. Pojedinačni rezultati doktorskog istraživanja provedenog na pilot području površine 21 km2, postignuti u okviru razvoja metodologije za izradu karata klizišta korištenjem LiDAR tehnologije su: digitalni modeli terena bez vegetacije rezolucije 0,3x0,3 m i 1x1 m optimalni za vizualno i automatizirano kartiranje klizišta; detaljan i potpun inventar klizišta izrađen vizualnom interpretacijom morfometrijskih karata izvedenih iz LiDAR DMT-a; definirane granične vrijednosti morfometrijskih karata izvedenih iz LiDAR DMT-a koje ukazuju na nepravilnu morfologiju klizišta te omogućavaju automatizirano kartiranje klizišta; pikselno orijentiran model za automatizirano kartiranje klizišta na temelju morfometrijskih parametara izvedenih iz LiDAR DMT-a; tri karte podložnosti na klizanje dobivene analizom preduvjeta klizanja i karata inventara klizišta izrađenih na temelju vizualne interpretacije i automatiziranog kartiranja LiDAR DMT-a; i doprinos poznavanju prostorne raspodjele klizišta te preduvjeta koji utječu na pojavu klizišta u podsljemenskoj zoni Grada Zagreba. Najvažnije nove spoznaje o značajkama klizišta na istraživanom području se odnose na pouzdane podatke o karakterističnoj veličini klizišta (u rasponu od 43 do 8.064 m2, najučestalija površina klizišta 400 m2), broju i gustoći klizišta (702 kartirana klizišta, prosječne gustoće 33,3 klizišta/km2) i prostornoj raspodjeli klizišta. S obzirom na visoku točnost izrađenih karata klizišta, zaključeno je da metodologija razvijena za automatizirano kartiranje klizišta na temelju morfometrijskih parametara omogućava izradu pouzdanih prognostičkih karata hazarda klizanja, na terenima istih ili sličnih inženjerskogeoloških uvjeta.
Abstract (english) In the framework of the doctoral thesis, methodology for landslide mapping using LiDAR (Light Detection and Ranging) technology was established, encompassing inventory mapping and derivation of prognostic susceptibility maps. Airborne laser scanning (ALS) allows generation of bare-earth digital terrain model (DTM) optimal for identification and mapping of small and shallow landslides under dense vegetation. Acquisition of the LiDAR data for the Podsljeme area (180 km2) in the City of Zagreb was performed in December 2013, following the extreme precipitation period in winter 2012/2013 that caused multiple occurrences of regional landslides (MORLE). The first result of the doctoral research is high-resolution DTM optimal for mapping of targeted landslides, generated from a point cloud based on suitable interpolation method and spatial resolution (0,3x0,3 m and 1x1 m). A series of topographic derivative datasets were generated from the LiDAR DTM for landslide mapping that was performed for the pilot area of 21 km2, which represents 12 % of the hilly Podsljeme area in the Zagreb City. Landslide inventory maps for the pilot area were prepared using two different methods: visual identification of landslides on LiDAR DTM derivatives and automated mapping based on morphometric characteristics of landslide areas compared to non-landslide areas. The second phase of research involved visual interpretation of three LiDAR DTM derivative maps (hillshade, slope and contours) with a spatial resolution of 0,3 x 0,3 m which resulted with inventory map containing 702 landslide polygons in the pilot area (21 km2). Seventy-five percent of the landslides in inventory have an area between 150 and 2000 m2 while the area of the smallest identified landslide is 43 m2. Around 10 % of visually identified landslides were additionally checked and evaluated in the field, while the whole inventory map for the pilot area was compared with conventionally prepared historical landslide inventory maps. A comparison of landslide inventories and the frequency–area distribution of visually interpreted landslides proves that LiDAR DTM enables preparation of detailed and complete landslide inventory maps (after MORLE) in the Podsljeme area. In the third phase of research, the quantitative geomorphological analysis on visually interpreted landslides was performed to define the topographic signature of landslide morphology using statistical parameters. In soil-covered areas, such as the Podsljeme area, landslides are generally characterized by a different morphometric values, i.e., higher surface roughness compared to the surrounding landslide-free areas. Based on defined morphological parameters which indicate hummocky landslide morphology in the test area (10.5 km2), the pixel-based model for landslide mapping was developed. Verification showed that landslides and unstable slopes can be predicted with an accuracy of 70 % and it can be concluded that developed model allows recognition and characterization of morphologic properties of forested small and shallow landslides on soil-covered hillslopes. In the last phase of the study, visually and automatically prepared landslide inventories were used to perform the landslide susceptibility maps. Landslide causal factors analysis and susceptibility assessment was done using bivariate statistical method and comparison of three landslide susceptibility maps were performed due to the area under the receiver operating characteristic (ROC) curves. The final landslide susceptibility map, using automatically prepared landslide inventory as input, resulted with high prediction rate (AUC=83 %), due to validation with visually interpreted landslides. It can be concluded that the model developed for automated landslide mapping is a useful tool for fast and cost-efficient production of reliable landslide susceptibility maps. Therefore, the methodology established for landslide mapping using LiDAR technology has practical implication in landslide risk reduction.
Keywords
klizište
LiDAR
vizualna interpretacija
digitalni model terena
model za automatizirano kartiranje klizišta
inventar klizišta
karta podložnosti na klizanje
podsljemenska zona Grada Zagreba
Keywords (english)
landslides
LiDAR
visual landslide mapping based on LiDAR DTM
automated landslide mapping based on LiDAR DEM
landslide inventory maps
landslide susceptibility mapping
Podsljeme area
City of Zagreb
Croatia
Language croatian URN:NBN urn:nbn:hr:169:398183 Study programme Title: Applied Geosciences, Mining and Petroleum Engineering Type of resource Text Extent XVI, 257 str. ; 30 cm File origin Born digital Access conditions Open access Terms of use Created on 2019-06-10 10:45:45
