Advancing process understanding and modeling of contrail formation with emphasis on hydrogen combustion

Kurzfassung

The use of hydrogen (H2) produced from renewable energy sources could eliminate aviation-related CO2 emissions entirely. For present-day aviation based on fossil fuels, however, roughly two-thirds of the climate impact are attributable to non-CO2 effects, with the largest contribution arising from long-lived contrail cirrus (ice clouds generated by aviation). The radiative forcing of contrail cirrus is estimated using general circulation models (GCMs) with integrated contrail modules. As small-scale processes below the grid resolution of these models cannot be resolved explicitly, they must be parameterized. This includes the processes governing contrail formation that occur within the first seconds in the exhaust plume behind an aircraft engine. A key quantity is the number of ice crystals formed, as it strongly influences the lifetime and radiative impact of long-lived contrails. For conventional kerosene combustion, droplet activation followed by freezing into ice crystals occurs primarily on emitted soot particles. A suitable parameterization of ice crystal number already exists for this case. For hydrogen combustion, however, both the thermodynamic conditions due to increased water vapor emissions and the nature of the condensation nuclei differ fundamentally. In the ideal case, the exhaust is particle-free, so that ice crystals form exclusively on entrained ambient aerosol particles. These fundamental differences require a parameterization specifically tailored to hydrogen combustion. The aim of this work is to develop such a parameterization for the hydrogen combustion case. The parameterization is based on a comprehensive dataset of contrail formation simulations performed with a box model including particle-based microphysics. Using these simulations, the influence of ambient conditions, aerosol properties, and engine parameters on ice crystal number is systematically investigated and physically interpreted. Shallow neural networks are trained and used as a fitting function to represent the simulated ice crystal number. Despite their simple architecture, these networks are sufficiently flexible to accurately represent nonlinear dependencies. Since their application requires only a small number of computational operations, they can be readily integrated into large-scale models. Physically motivated analytical scalings further ensure the broad applicability of the parameterization. In addition, the potential influence of volatile particles that may form from lubrication oil emissions is investigated. The scenarios under which these particles may dominate ice crystal formation are analyzed for both kerosene and hydrogen combustion. The parameterization developed can be integrated into large-scale models and thus provides a basis for estimating the radiative impact of contrail cirrus produced by a potential future fleet of hydrogen-powered aircraft. Moreover, the methodology presented together with the theoretical understanding gained could be used to derive a comparable parameterization for sustainable aviation fuels (SAFs).