Complexity analysis of turbulent flow around a street canyon

Abstract

We analyzed the results of two-component velocity measurements at several points in a wind tunnel model representing a simplified, idealized urban environment. The velocity fluctuations were treated as a marked point process. The statistical distribution of the interarrival times were characterized by a lognormal fit. The time series were transformed into symbol sequences, utilizing the quadrant method. The information content of the “quadrantified” symbol sequences was investigated by comparing the number of words and normalized entropy levels in case of the measured and several artificially generated periodic, random and noisy periodic symbol sequences. We found that artificially generated periodic series with noise show qualitatively similar entropy distribution to that of the measured signal. Surrogate sequences were produced based on first and higher order Markov–statistics, the entropy levels of which were also compared to those of the measured sequences. We demonstrated that the information content of the velocity fluctuation series can be better captured by higher order Markov–chains. The quadrant transformation was also performed in rotated quadrant coordinate systems, and we found that the entropy is close to minimal in the principal axes system of the velocity fluctuation pairs.