Industrial surfaces can become rough due to mechanisms such as deposition, erosion, corrosion, icing, etc. Roughness can affect the aerodynamic or heat transfer behaviour of the surfaces leading to, e.g. increased profile loss or heat load in a turbine blade. The stochastic nature of the roughness formed by these mechanisms is a major challenge in engineering prediction of roughness effects. Thus, the existing models and correlations, captured based on limited experimental datasets, suffer a severe lack of universality. Our vision is that Direct Numerical Simulation (DNS) coupled with model reduction techniques can be used to develop a general framework for predicting the aerodynamic characteristics of surface roughness and its sensitivities. The requested funds will be used (1) to support a preliminary study aiming at a major collaborative research proposal (2) to organize an international workshop that promotes a fruitful ideas exchange with top experts in the field and guarantees a high-quality proposal. The invited guests will be experienced researchers in the fluid mechanics field with successful experience in model reduction and adjoint-based methods.