Methodology

Data collection for landslide susceptibility assessment encompass: (A) collection of landslide data (preparation of landslide inventories); and (B) collection of landslide causal factors (preparation of landslide factor maps). Collection of landslide data and morphological landslide factor data will be performed from remote sensing data, i.e., airborne LiDAR data.

Airborne laser scanning will be carried out during leaf off period (December-March) with at an average density of 5 points/m² and an average spacing between the closest point pairs of 0.5 m. Resulting point clouds will serve for interpolation of bare-earth DTMs. LiDAR DTM will be derived in several different resolutions using several different interpolations. In each pilot area, preliminary visual interpretation of DTMs will be performed to create a set of landslides for quantitative geomorphological analysis, i.e., analysis for determining an optimal LiDAR DTM for visual landslide inventory mapping. The quality of DTM for visual landslide identification and mapping will be determined based on topographic signature of landslide morphology using statistical parameters. Collection of landslide data in the form of landslide inventory map include visual interpretation of the most optimal LiDAR DTM derivatives, delineation of landslide contours in GIS followed by field checking of small subset of landslides. Based on optimal DTM for interpreting landslide morphology, five topographic derivative datasets will be derived: hillshade maps, slope maps, contour lines, curvature and surface roughness maps. Landslide inventory mapping will be carried out at detail scale (1:100–1:500) to ensure a correct delineation of the landslide boundaries.

Collection of landslide causal factors will be carried out using GIS analyses, including morphological, geological, hydrological and antropogenic factors. The selection of landslide controlling factors and processes of its derivation is strongly related to the following up analyses of landslide susceptibility. The combinations of factors that have triggered landslides in the past are evaluated statistically. Final selection of landslide-conditioning factors will be decided based on availability of original data (date, scale and resolution of maps) and its relative importance to the occurrence of mass movements in particular pilot area.

Data analyses using statistical methods is planed by application of three data-driven approaches: (i) bivariate statistical methods (ii) multivariate statistical methods and (iii) machine learning methods. Selected statistical methods represent most commonly used approaches for landslide susceptibility assessment and shows best prediction modelling rates. Furthermore, susceptibility analysis will be carried out for six different mapping units. Susceptibility analyses will be performed using a landslide training group (50% of the inventory randomly selected), while independent validation will be carried out with the other 50% of the landslide inventory (landslide test group). Furthermore, landslide susceptibility models will be computed using the landslide training group represented as visually mapped landslide polygons, using the point features at the headscarps of mapped landslides, and other using the point features corresponding to the centroid of landslides. Results of data analyses will be series of more than 200 landslide susceptibility maps per pilot area.

The key indirect goal is to test possibilities and limitations of applications of three statistical techniques for landslide susceptibility assessment on particular data set representing specific environmental/engineering geological conditions related to particular landslide type(s)  as well as to determine resolution of input and resulting maps, cartographic units for analysis and representation of results and form of input dependent variable (landslide polygons vs. points representing landslides).

Verification of results will be be evaluated using visually mapped landslide inventory, i.e. landslide training group (50% of the inventory randomly selected) and landslide test group (other 50% of the landslide inventory, using ROC curves.

The key indirect goal is define the best method for landslide susceptibility assessment by means of its resolution, input parameters of independent variables (landslide factor maps) and input parameters in form of landslide inventory map for specific environmental/engineering geological conditions and landslide type.

Classification of resulting susceptibility maps is a crucial step for land use spatial planning and management which influences possibilities of practical use of maps and quality of information that are depicted by map. An essential issue is the determination of susceptibility thresholds, i.e., classification of derived landslide susceptibility maps into a limited number of susceptibility classes. Adopting one classification system or another will not only affect the map’s readability and final appearance, but most importantly, it may affect the decision-making tasks required for effective land management. The proposed research will compare and evaluate the reliability of the most commonly used classification methods.

According to clearly defined purpose of end-product maps which will be derived by the project, criteria and potential uses will be defined by practitioners and experts from the domain of land use planning. Methodology for deriving expert opinion will include questionnaires, transfer of experience from EU countries (Italy and Slovenia) and one round-table with representatives of local and regional government/administration under  surveillance of Croatian Ministry of Construction and Physical Planning and. The key indirect goal is to develop model of cartographic information about landslides for sustainable management of risks through the system of land use planning at the level of municipalities, cities and counties.

Writing guidelines about practical application of the developed methodology in the system of spatial planning in Croatia is based on similar guidelines for landslide susceptibility, hazard and risk zoning for land use planning. LandSlidePlan project guidelines will provide: (i) definitions and terminology; (ii) description of the types and levels of landslide zoning; (iii) suggested scales for zoning maps taking into account the needs and objectives of land use planners and regulators and the purpose of the zoning; (iv) guidance on the information required for different levels of zoning taking account the various environments; (v) guidance on the reliability, validity and limitations of the methods, (vi) advice on the required qualifications of the persons carrying out landslide susceptibility zoning.

Besides scientific activities, the methodology of the research includes organization of events with active participation of stakeholders from national, regional and local government to get information from real systems of physical planning at all levels in Croatia.