Abstract (english) | The main motivation to research the landslide susceptibility assessment for application in land use planning arises from the national landslide risk assessment (BERNAT GAZIBARA et al., 2019) which recognised landslides as a second natural risk in Croatia (CNPDRR, 2019). Furthermore, the preliminary regional landslide susceptibility analysis showed that approx. 20 % of the Republic of Croatia area is potentially prone to sliding. Therefore, landslide susceptibility assessment for national, county and local levels was carried out in the frame of two scientific projects Methodology development for landslide susceptibility assessment for land-use planning based on LiDAR technology (LandSlidePlan, HRZZ IP-2019-04-9900) (BERNAT GAZIBARA et al., 2022) and project Applied landslide research for the development of risk mitigation and prevention measures (PRI-MJER, KK.05.1.1.02.0020). The national landslide susceptibility map at a small scale is created to give a general overview of critical areas for an entire country, and its purpose is to inform policymakers and the general public (MIHALIĆ ARBANAS et al., 2022). County-level landslide susceptibility assessment on a medium scale synthesizes available data and identifies wider areas with landslide problems and can be used to define areas for more detailed research on a local level. The third level is the local-scale landslide mapping and zonation that includes specific areas of local administrative units (municipality or city) or complex critical areas. The results were landslide susceptibility maps for seven study areas: (i) the Republic of Croatia; (ii) City of Zagreb, Karlovac County and Primorje–Gorski Kotar County; and (iii) the study areas in the Zagreb City (BERNAT GAZIBARA et al., 2023), Hrvatsko Zagorje (SINČIĆ et al., 2022a), Karlovac City (SINČIĆ et al., 2022b) and Istria. Methodology development for landslide susceptibility assessment on national and county scales was carried out using a heuristic approach, i.e. Fuzzy Logic method, and available topographical and geological data. Given that the validation of the final landslide susceptibility map is mandatory, and systematic landslide inventories at the national or county level do not exist, we used the landslide database conducted by the University of Zagreb, Faculty of Mining, Geology and Petroleum Engineering. The database consists of 2,186 landslides with the exact location of the occurrence. All landslide susceptibility maps showed high accuracy and were classified into three susceptibility zones, considering The Area Under the Receiver Operating Characteristic Curve (AUCROC). Methodology development for landslide susceptibility assessment on a local scale was carried out using different mapping units and statistical methods (e.g. Information Value method, Weights of Evidence method, Logistic Regression and Discriminant Analysis, and machine learning methods, including Support Vector Machine, Artificial Neural Network and Random Forest). Moreover, landslide susceptibility models were computed using different scenarios of high-resolution input data, i.e. geometrical types of LiDAR-based inventory and variations of causal factors. Finally, all landslide susceptibility models were evaluated based on model fitting performance, model prediction performance, and model uncertainty. The purpose of comparing landslide susceptibility models is to define the most suitable maps for application in spatial planning at national, regional, and local levels. The research was based on innovative technologies, limitations related to the availability of spatial data in Croatia (limited amount of geological data), and urgent needs for efficient solutions applicable in the Croatian spatial planning system in line with European and global requirements related to sustainable development, human and environmental protection. |