Publications

High-resolution numerical weather prediction simulations are able to reproduce observed stochastic scale invariant behavior of atmospheric wind and water fields down to the effective model resolution, which is shown to be a process-dependent transient property that varies with the underlying dynamics. The effective resolution gain in dynamical downscaling of convective regimes is substantially smaller than the grid size decrease indicating that improvements in the model's capacity to resolve small-scale processes require consistent adjustments including both numerical formulation and physical parameterizations. Instantaneous realizations of simulated atmospheric wind and water fields exhibit robust multifractal properties with intrinsically transient scaling behavior depending on the underlying atmospheric state. In particular, a sharp transition in the scaling parameters between nonconvective and convective conditions is found, which explains different scaling regimes reported in the literature for atmospheric wind, temperature, and moisture observations. Spectral slopes around 2–2.3 arise under nonconvective or very weak convective conditions, tightly related to the scaling behavior of the underlying topography. In convective situations the transient scaling exponents remain under 5/3 in agreement with the Kolmogorov turbulent regime accounting for the intermittency correction. These findings have important implications for stochastic downscaling and the implementation of stochastic subgrid scale parameterizations using fractal methods. Specifically, it is shown that, based on scaling arguments, subgrid scale probability distributions of atmospheric moisture can be obtained from the coarse resolution information alone. Our results suggest that fractal methods can be used for estimating temporally and spatially varying regime-based subgrid scale statistics (and realizations of moisture fields) in real time and in a computationally efficient manner that could be useful in climate models.