To Scale or Not to Scale: The Question is Truly How to Characterize the Integral Scales of Turbulence.

Direct computation of turbulence length and time scales from observational data has become increasingly possible, and thus more prevalent. Estimation of turbulence length and time scales is important for many reasons, including turbulence closure model development. While measurement technology and observational techniques have improved our ability to more accurately measure quantities of interest, our approach towards harmonizing theory and observations has more or less stagnated for estimating the integral length and time scales in a turbulent flow from said data. Since it is impossible to integrate over infinite limits, there exists a multitude of methods and practical work-arounds for using observational data to estimate turbulence scales, making it difficult for researchers to make and justify an appropriate choice as applied to their situation. In an attempt to begin a thoughtful conversation within the community on 'best practices', we use 1 Hz temperature measurements taken at five heights on eight spatially separated masts aligned along the space between vineyard rows on a slope to infer turbulence length and time scales. We investigate how the choice of integral scale computation affects reported results, and what these effects imply. We find that choice of methodology does not affect the comparative behaviour of turbulent scales with height and distance, but that the overall value of reported scales is indeed affected, while still remaining of the same magnitude. This presents an issue to the community when an 'expected' numerical value of a turbulence scale is unknown as there is no truthful way to quantify the accuracy of a reported result and thus no clear preferred method. We propose that in these cases, a method that minimizes analysis bias and appropriately represents known flow dynamics be employed. In this way, the focus is shifted from numerical values to overall dynamical behaviour, and thus method preference bias does not enter the more important comparison of dynamical behaviour between different studies or sites.