Intermittency of turbulence in the stable atmospheric boundary layer: a field data and stochastic modeling approach

Stably stratified atmospheric flows are typically characterized by weak and intermittent turbulence, gravity waves, low level jets and Kelvin-Helmholtz instabilities. These phenomena greatly complicate the modeling and measurements of the stable boundary layer (SBL). The intermittency of turbulence clearly lacks physical understanding and that leads to specific problems in the representation of the SBL in weather and climate models.

In this study we investigate the intermittency of turbulence in very stable conditions by applying new statistical analysis tools to the SnoHATS dataset, collected in Switzerland over the Glacier de la Plaine Morte in 2006 (data collected by the EFLUM laboratory, EPFL). These statistical tools can be used to develop a stochastic parameterization for the SBL for use in weather or climate models.

The SnoHATS dataset includes measurements of atmospheric turbulence collected by horizontal arrays of sonic anemometers. We apply a data driven approach for parameterization of turbulence, based on nonstationary multivariate autoregressive factor models (VARX). The nonstationarity of the turbulence is described by means of a finite element method (FEM) clustering procedure (Horenko 2010). This recently introduced procedure allows us to investigate nonstationary turbulent mixing using 2 approaches:

1- We analyze results of the clustering of vertical velocity fluctuations and other turbulence data using different sets of external influencing factors for the VARX models. The automatic procedure shows successful clustering of the data and helps us gain physical intuition on the underlying causes for higher mixing events in the SBL. For example, we are able to isolate cases where the mixing is largely due to shear from cases where the mixing is mostly due to temperature instabilities.

2- We use prefiltering of the external influences used in the FEM-VARX procedure to assess which scales of motions are largely triggering turbulent mixing. Results show a large influence of the motions of scales larger than 1 minute on the turbulent vertical velocity fluctuations. We subsequently apply conditional sampling on the clustered data to explore the relationship of turbulence quantities to mean flow quantities, such as the Richardson number, for the different clusters.

Ultimately the goal is to derive a stochastic representation of turbulence outbursts that would realistically represent turbulent kinetic energy dissipation under very stable stratification, thereby overcoming some of the difficulties of ABL parameterization in weather and climate models. The technique allows for a description of the linkages between different scales of motions, through the parameterization of the jump process between different locally stationary VARX models.

Horenko, I., 2010: On the Identification of Nonstationary Factor Models and Their Application to Atmospheric Data Analysis. J. Atmos. Sci, 67, 1559–1574, doi:10.1175/2010JAS3271.1.