In-flight icing is a dangerous phenomenon that poses significant risks to aircraft safety, resulting from the accumulation of ice on surfaces due to supercooled water droplets. This ice formation can lead to reduced visibility, engine power loss, blocked probes and vents, and adverse effects on aerodynamics. Current methods for assessing icing scenarios rely heavily on costly wind tunnel and flight tests highlighting the urgent need for validated simulation tools. The existing reliance on experimental measurements as the sole "ground truth" in the validation process often leads to biased model assessments, compounding uncertainties that can adversely affect statistical reliability. This action aims to revolutionize the validation process of numerical simulations for in-flight icing by directly addressing these uncertainties and biases. I propose a novel statistical validation framework that incorporates uncertainties at every stage of analysis, enabling a clearer identification of physical model gaps. The proposed methodology encourages a dynamic and continuous exchange between experimental data and numerical models, allowing for mutual refinement. The research objectives include: (1) Characterizing uncertainties in experimental data related to ice accretion through focused campaigns, leading to reliable datasets; (2) Developing robust statistical metrics that comprehensively capture the variability inherent in both computational predictions and experimental outcomes; and (3) Benchmarking the application of these methodologies to evaluate existing ice accretion models. This innovative approach promises to bridge the gap between experiments and simulations, providing reliable tools for aerospace design and certification. If successful, this work will establish a new standard in certification by analysis, reshaping future validation practices in complex multi-physics applications such as combustion, heat transfer, and fluid-structure interactions.

Many Earth system processes involving multi-physics, multi-phase conditions extend over several orders of magnitude in length- and time-scales. Engineering science, in pursuit of deeper process understanding and solution-oriented design, has used scaling theories to address scale-afflicted, complex processes through experimental work in laboratory environment at reduced scale. The standard scaling approach, the Buckingham π-theorem, is especially deficient when multi-physics and multi-phase processes require the choice of more than a single non-dimensional number, resulting in severe scale effects and typically meaning that accuracies at reduced scale are inadequately quantified. Hence, we choose a demonstrably complex multi-physics, multi-phase process for the investigation of scaling accuracies – the progressive collapsing of residential buildings and the associate debris transport, evolving from extreme flow events from natural hazards, such as flash floods or tsunami. ANGRYWATERS seeks to achieve a breakthrough in modelling these complex processes by deriving novel scaling laws that will be developed in the framework of the Lie group of point scaling transformations. Scaling requirements will be applied to the combined fluid-structure interaction at various scales, developing sophisticated building specimens; here, we employ 3D-printing and appropriately engineered materials to match the scaling requirements. We conduct a comprehensive experimental campaign, using medium- and large-scale facilities, subjecting the specimens to extreme flow conditions in the form of dam-break waves. We consider sub-assemblages, single and multiple buildings, enhancing the understanding of energy losses and debris production upon collapse, elaborating reduced scale accuracies. High-fidelity numerical modelling will complement our experiments, deepening our process understanding; a depth-averaged model with novel debris advection model crucially enhances predictive capabilities.