Parametrizing the number of formed ice crystals in contrails from hydrogen combustion

Hydrogen-powered aircraft have the potential to reduce CO2 emissions to zero. However, a significant portion of the global warming attributed to aviation arises from non-CO2 effects, including contrails. The thermodynamic state and microphysical pathways that form these contrails differ substantially between hydrogen and conventional kerosene combustion. Therefore, the overall climate impact of contrails formed by hydrogen combustion is not yet known and needs to be assessed by Global Circulation Models (GCMs). The contrail parametrization in a GCM cannot resolve the contrail formation processes. However, these early processes have a large influence on the contrail life cycle and should therefore be included in the contrail initialization of a GCM. Here, a crucial ingredient is the number of ice crystals formed during the jet phase.

In this study, we present a parametrization that provides a link between the outcome of a high-resolution model and the contrail initialization in a GCM. For that, we performed contrail formation simulations with the particle-based Lagrangian Cloud Module (LCM) in a box model approach. We assume that contrail droplets and ice crystals form solely on entrained ambient aerosols. With our simulation setup, we aim to cover the entire parameter space relevant for contrail formation in the case of hydrogen combustion. This involves varying background meteorological conditions, ambient aerosol properties, and engine exit conditions, resulting in more than 20,000 simulations.

The simulation results show that the number of formed ice crystals is mostly sensitive to a variation of the ambient aerosol background concentration, followed by a variation of the ambient temperature. We identify a parameter subspace where the number of ice crystals becomes almost independent of the size and chemical composition of the ambient aerosols.

Furthermore, we performed simulations with two coexisting background aerosol ensembles differing in mean size and/or solubility. The simulation results show that coarse mode particles have neither a direct nor an indirect influence on the number of formed ice crystals if their number concentration is 2-3 orders of magnitude lower than that of a coexisting Aitken/accumulation mode. Furthermore, the ice crystal number from simulations with the two coexisting background aerosol ensembles can be reconstructed by a weighted mean of two single simulations, each containing only one of the two aerosol ensembles. This allows to construct a simpler parametrization still covering the case of two coexisting aerosol ensembles.

We used the simulation results to train a shallow feed-forward neural network that maps the box model input parameters to the number of formed ice crystals. This neural network serves as a fit function of our simulation results, which can be implemented in a GCM for contrail initialization.

This work contributes to the collaborative effort of the German Aerospace Center (DLR) and Airbus in assessing the climate impact of H2 contrails.