Computationally-feasible uncertainty quantification in model-based landslide risk assessment

Introduction: Increasing complexity and capacity of computational physics-based landslide run-out modelling yielded highly efficient model-based decision support tools, e.g. susceptibility or maps, geohazard risk assessments. A reliable, robust reproducible development such tools requires a thorough quantification uncertainties, which are present in every step workflow from input data, as topography release zone, to framework used, numerical error. Methodology: Well-established methods reliability analysis Point Estimate Method (PEM) Monte Carlo Simulations (MCS) can be used investigate the uncertainty model outputs. While PEM less resources, it does not capture all details uncertain output. MCS tackles this problem, but creates bottleneck. comparative study is presented herein by conducting multiple forward simulations for synthetic real-world test case, construct Gaussian process emulators surrogate facilitate high-throughput tasks. Results: It was demonstrated that provide similar expectancies, while variance skewness differ, terms post-processed scalar outputs, impact area point-wise flow height. Spatial distribution height clearly affected choice method quantification. Discussion: If only expectancies assessed then one work with computationally-cheap PEM, yet has when higher order moments needed. In case machine learning techniques, emulation, strategies tackle further suggested computational-feasibility assessment significantly improved using modelling. should also noted gain compute time emulation critically depends on effort needed produce training dataset simulations.