Abstract
Numerically modeled turbulence simulated by the Advanced Research Weather Research and Forecasting Model (ARW) is evaluated with turbulence measurements from NOAA’s high-resolution Doppler lidar on the NOAA Research Vessel Ronald H. Brown during the Variability of the American Monsoon Systems (VAMOS) Ocean–Cloud–Atmosphere–Land Study—Regional Experiment (VOCALS-Rex) field program. A nonprecipitating nocturnal marine stratocumulus case is examined, and a nudging technique is applied to allow turbulence to spin up and come into a statistically stationary state with the initial observed cloud field. This “stationary” state is then used as the initial condition for the subsequent free-running simulation. The comparison shows that the modeled turbulence is consistently weaker than that observed. For the same resolution, the turbulence becomes stronger, especially for the horizontal component, as the length of the horizontal domain increases from 6.4 to 25.6 km. Analysis of the power spectral density shows that, even for the largest domain, the horizontal component of the turbulence is limited by the upper limit of the domain size; supporting evidence from past studies is provided. Results suggest that convergence is expected for (i) energy spectra of turbulence with a sufficiently large domain and (ii) liquid water path with an adequately large domain and fine resolution. Additional tests are performed by changing momentum advection and turning off subgrid-scale diffusion. These exhibit more significant changes in turbulence characteristics compared to the sensitivity to domain size and resolution, suggesting that the model behavior is essentially established by the configuration of the model dynamics and physics and that the simulation only gradually improves when domain size and resolution are increased.