Exploring implications of input parameter uncertainties in glacial lake outburst flood (GLOF) modelling results using the modelling code r.avaflow

Abstract

Modelling complex mass flow processes, such as glacial lake outburst floods (GLOFs), for hazard and risk assessments requires extensive data and computational resources. Researchers often rely on low-resolution, open-access datasets and parameters derived from plausibility due to the difficulty involved in conducting direct measurements. This results in considerable uncertainties in forward modelling, potentially limiting the accuracy and reliability of predictions. To determine the sensitivity of the model outputs stemming from input parameter uncertainties in the forward modelling, we selected 9 parameters relevant to GLOF modelling and performed a total of 84 simulations, each representing a unique GLOF scenario in the physically based r.avaflow model. Our results indicate that mass-movement-triggered moraine-dammed GLOF modelling outputs are notably sensitive to five parameters, which are, in order of importance: (1) volume of mass movement entering the lake, (2) DEM datasets, (3) origin of mass movement, (4) entrainment coefficient, and (5) basal friction angle. The GLOF output parameter resulting from the volume of mass movement entering the lake has the greatest coefficient of variation (CV) (47 %), while the internal friction angle had the lowest CV (0.4 %). For future GLOF modelling, we recommend carefully considering the output uncertainty stemming from the sensitive input parameters identified here, some of which cannot be constrained before a GLOF and which must be addressed using statistical approaches.