Spatio-temporal consistency of cloud-microphysical parameter sensitivity in a warm-conveyor belt

Abstract

A good representation of clouds and precipitation processes is essential in numerical weather and climate models. Subgrid-scale processes, such as cloud physics, are parameterized and inherently introduce uncertainty into models. Traditionally, the sensitivities of the model state to specific uncertain parameters are quantified through perturbations to a few selected parameters, limited by computational resources. Algorithmic Differentiation (AD) enables the efficient and simultaneous estimation of sensitivities for a large number of parameters, thereby overcoming the previous limitations and significantly enhancing the efficiency of the analysis. This framework provides an objective way to identify processes where more precise representations have the largest impact on model accuracy. AD-estimated sensitivities can also address the underdispersiveness of perturbed ensemble simulations by guiding the parameter selection or the perturbation itself. In our study, we applied AD to 169 uncertain parameters identified in the two-moment microphysics scheme of the numerical weather prediction (NWP) model ICON of the German Weather Service. This application of AD allowed us to evaluate the sensitivities of specific humidity, latent heating, and latent cooling along several thousand warm conveyor belt trajectories. This coherent, strongly ascending Lagrangian flow feature is crucial for the cloud and precipitation structure and the evolution of extratropical cyclones. The quantification of individual parameter sensitivities shows that only 38 parameters are of primary importance for the investigated model state variables. These parameters are associated with rain evaporation, hydrometeor diameter-mass relations, and fall velocities. Moreover, the parameter sensitivities systematically vary with different microphysical regimes, ascent behavior, and ascent stages of the WCB airstream. Finally, several parameters impact an extended region in the extratropical cyclone, illustrating the spatiotemporal consistency of cloud microphysical parameter uncertainty in the applied NWP model and microphysics scheme.