Extend universal ordinary differential equations to comprehensive cloud-microphysics schemes

Determine whether universal ordinary differential equation approaches can be applied successfully to more comprehensive cloud microphysics schemes beyond the idealized collision–coalescence warm-rain system.

Background

The study demonstrates that a universal ordinary differential equation can learn an approximate warm-rain closure from trajectories of bulk prognostic variables while retaining the ordinary-differential-equation structure. The experiments, however, are restricted to an idealized box model, a single collection kernel, and collision–coalescence processes.

A remaining unresolved issue is whether the same operator-learning strategy can scale to more comprehensive cloud microphysics schemes that include additional processes and more complex closures.