Publications & Events

pubs home | list all entries

An Intercomparison of T-REX Mountain-Wave Simulations and Implications for Mesoscale Predictability

Abstract

Numerical simulations of flow over steep terrain using 11 different nonhydrostatic numerical models are compared and analyzed. A basic benchmark and five other test cases are simulated in a two-dimensional framework using the same initial state, which is based on conditions during Intensive Observation Period (IOP) 6 of the Terrain-Induced Rotor Experiment (T-REX), in which intense mountain-wave activity was observed. All of the models use an identical horizontal resolution of 1 km and the same vertical resolution. The six simulated test cases use various terrain heights: a 100-m bell-shaped hill, a 1000-m idealized ridge that is steeper on the lee slope, a 2500-m ridge with the same terrain shape, and a cross-Sierra terrain profile. The models are tested with both free-slip and no-slip lower boundary conditions. The results indicate a surprisingly diverse spectrum of simulated mountain-wave characteristics including lee waves, hydraulic-like jump features, and gravity wave breaking. The vertical velocity standard deviation is twice as large in the free-slip experiments relative to the no-slip simulations. Nevertheless, the no-slip simulations also exhibit considerable variations in the wave characteristics. The results imply relatively low predictability of key characteristics of topographically forced flows such as the strength of downslope winds and stratospheric wave breaking. The vertical flux of horizontal momentum, which is a domain-integrated quantity, exhibits considerable spread among the models, particularly for the experiments with the 2500-m ridge and Sierra terrain. The differences among the various model simulations, all initialized with identical initial states, suggest that model dynamical cores may be an important component of diversity for the design of mesoscale ensemble systems for topographically forced flows. The intermodel differences are significantly larger than sensitivity experiments within a single modeling system.

View Citation