Error-Growth Dynamics and Predictability of Surface Thermally Induced Atmospheric Flow

Abstract

Using the CSU Regional Atmospheric Modeling System (RAMS) in its nonhydrostatic and compressible configuration, over 200 two-dimensional simulations with ?x = 2 km and ?x = 100 m are performed to study in detail the initial adjustment process and the error-growth dynamics of surface thermally induced circulations including the sensitivity to initial conditions (i.e., the traditional predictability), boundary conditions, and model parameters, and to study the predictability as a function of the size of surface heat patches under a calm mean wind. It is found that the error growth (at least at the stage when the surface forcing is strong) is not sensitive to the characteristics of the initial perturbations. The numerical smoothing has a strong impact on the initial adjustment process and on the error-growth dynamics. The predictability is variable dependent. The mesoscale flow is insensitive to lateral and top boundary conditions. Among the conclusions regarding the influence of the boundary-layer structures and model parameters on the predictability and flow structures, it is found that the vertical velocity field is strongly affected by the mean wind, and the flow structures are quite sensitive to the initial soil water content. The transition from organized flow to the situation in which fluxes are dominated by noncoherent turbulent eddies under a calm mean wind is quantitatively evaluated and this transition is different for different variables. The relationship between the predictability of a realization and of an ensemble average is discussed. The predictability and the coherent circulations modulated by the surface inhomogeneities are also studied by computing the autocorrelations and the power spectra. The three-dimensional mesoscale and large-eddy simulations are performed to verify the above results. It is found that the two-dimensional mesoscale (or fine resolution) simulation yields very close or similar results regarding the predictability as those from the three-dimensional mesoscale (or large eddy) simulation. The horizontally averaged quantities based on two-dimensional fine-resolution simulations are insensitive to initial perturbations and agree with those based on three-dimensional large-eddy simulations.