NAM-SCA: A Nonhydrostatic Anelastic Model with Segmentally Constant Approximations

An atmospheric convective system may be modeled as an ensemble of discrete plume elements. A representation of decomposited plumes provides the basis for mass-flux convective parameterization. A dry version of such a prototype model is constructed in a two-dimensional horizontally periodic domain. Each discrete plume element is approximated by a horizontally homogeneous segment such that the whole system is given by segmentally constant approximations (SCA) in the horizontal direction for each vertical level in a nonhydrostatic anelastic model (NAM). The distribution of constant segments is highly inhomogeneous in space and evolves with time in a highly adaptive manner. The basic modeling strategy from a physical point of view is to activate new segments vertically upward with time when a convective plume is growing and to deactivate segments when a plume event is over. The difference in physical values crossing segment interfaces is used as a criterion for numerically implementing this strategy. Whenever a large difference is found, the given interface is stretched vertically by subdividing an existing segment into two. In turn, when a segment interface difference is found below the threshold, the given interface is removed, thereby merging the two segments into one. This nonhydrostatic anelastic model with segmentally constant approximations (NAM SCA) is tested on an idealized atmospheric convective boundary layer. It successfully simulates the evolution of convective plumes with a relatively limited number of segments (i.e., high compression) and with a much scarcer distribution of segments over nonplume environments (i.e., extremely inhomogeneous distribution of segments). Overall, this method compresses the size of the model up to 5 times compared to a standard NAM with homogeneous grid distribution without substantially sacrificing numerical accuracy.