TY - JOUR
T1 - Parametric modelling of turbulence
JF - Philosophical Transactions of the Royal Society of London. Series A:
Physical and Engineering Sciences
JO - Philos Trans Phys Sci Eng
SP - 439
LP - 455
M3 - 10.1098/rsta.1990.0125
VL - 332
IS - 1627
AU -
AU -
AU -
Y1 - 1990/09/15
UR - http://rsta.royalsocietypublishing.org/content/332/1627/439.abstract
N2 - Some steps are taken towards a parametric statistical model for the velocity and velocity derivative fields in stationary turbulence, building on the background of existing theoretical and empirical knowledge of such fields. While the ultimate goal is a model for the three-dimensional velocity components, and hence for the corresponding velocity derivatives, we concentrate here on the stream wise velocity component. Discrete and continuous time stochastic processes of the first-order autoregressive type and with one-dimensional marginals having log-linear tails are constructed and compared with two large data-sets. It turns out that a first-order autoregression that fits the local correlation structure well is not capable of describing the correlations over longer ranges. A good fit locally as well as at longer ranges is achieved by using a process that is the sum of two independent autoregressions. We study this type of model in some detail. We also consider a model derived from the above-mentioned autoregressions and with dependence structure on the borderline to long-range dependence. This model is obtained by means of a general method for construction of processes with long-range dependence. Some suggestions for future empirical and theoretical work are given.
ER -