Turbulent Scales in the Boundary Layer: A Year-long LES

We present results of a single, continuous Large-Eddy Simulation of actual weather conditions during the timespan of a full year, made possible through recent computational developments. The run is forced on the basis of a regional weather model, in order to provide an LES dataset that is representative of the daily weather of the year 2012 around Cabauw, the Netherlands. This location is chosen such that LES results can be compared with both the regional weather model and observations from the Cabauw observational supersite (Cabauw Experimental Site for Atmospheric Research, CESAR).

The run was made possible by porting our Large-Eddy Simulation program to run completely on the GPU, resulting in GALES (Schalkwijk et al, 2012). GPU adaptation allows us to reach much improved time-to-solution ratios (i.e. simulation speedup versus real time). As a result, GALES can perform runs with a much longer timespan than previously feasible.

The dataset resulting from the LES run provides many avenues for further study. First, it can provide a more statistical approach to boundary-layer turbulence than the more common case-studies by simulating a diverse but representative set of situations, as well as the transition between situations. This has obvious advantages in designing and evaluating parameterizations. On the other hand, the set-up can provide valuable information on the quality of the LES when applied to a wide range of cases. In this talk, we aim to briefly explain and evaluate the results as well as present a number of highlights in the dataset, which is available on request.

Specifically, the year-long LES run provides a novel opportunity to study timescales ranging from seconds up to seasons on the basis of a single, uninterrupted simulation. For the first time, this allows us to construct time spectra of temperature, humidity and velocity from second to year-scale on the basis of LES time series. Since the full three dimensional turbulence field is known in the LES, we compare these time spectra with spatial spectra to find the length and time scales most important to turbulent transport in the boundary layer. This also allows us to investigate the presence of the (co-)spectral gap.