Structure extraction by stochastic estimation with adaptive events

Various aspects of turbulence structure can be found by a new class of stochasticestimation methods in which the conditional events that define the stochastic estimate are systematically varied. Methods are presented to find the length scale of large periodic structures, the form of structures that have specified geometric constraints such as two-dimensionality, and the structure of small-scale motions embedded in large-scale motions. These methodologies are demonstrated in high Rayleigh number turbulent convection by extracting both the large-scale roll-cell and coherent thermal plumes. A method of compressed representation using a stochastic estimate given data on optimally chosen points is also demonstrated.