Skip to article frontmatterSkip to article content
Site not loading correctly?

This may be due to an incorrect BASE_URL configuration. See the MyST Documentation for reference.

Correlation Analysis

Open from ‘3. Analysis →\to Correlation Analysis‘.

The two-point spatial correlation function

Ruu(Δx,Δy)=⟨u′(x,t) u′(x+Δ,t)⟩σ(x) σ(x+Δ)R_{uu}(\Delta x, \Delta y) = \frac{\langle u'(\mathbf{x}, t)\, u'(\mathbf{x}+\boldsymbol{\Delta}, t) \rangle}{\sigma(\mathbf{x})\,\sigma(\mathbf{x}+\boldsymbol{\Delta})}

measures the degree to which velocity fluctuations at two different locations are statistically related. When RuuR_{uu} decays from 1 at Δ=0\Delta = 0 to 0, the separation at which it crosses zero defines the integral length scale LL, a measure of the size of the largest energy-containing eddies (Pope, 2000). The temporal autocorrelation R(τ)R(\tau) plays an analogous role in time, with the integral time scale T=∫0∞R(τ) dτT = \int_0^{\infty} R(\tau)\,\mathrm{d}\tau.

Reference: Pope, S.B. (2000). Turbulent Flows. Cambridge University Press.

Selecting a reference

Spatial correlation

In Point mode you see the full 2D correlation map and two independent 1D slice panels. In ROI mode only the 1D panels are shown. The computed LxL_x and LyL_y are displayed in the readout strip at the bottom of the window. A vertical dashed line marks the crossing lag used by the selected scale method; for the zero-crossing method, this line is placed at the lag where RR first crosses zero, while LL is reported as the integral of RR up to that crossing. Click ‘Show Diagnostic Plot‘ to see the cumulative integral of R(Δr)R(\Delta r).

Correlation window in ROI mode showing R_{uu} in the x- and y-directions. The vertical dashed red line marks the zero-crossing lag used to compute the integral length scale. L_x and L_y are reported in the readout strip at the bottom.

Figure 1:Correlation window in ROI mode showing RuuR_{uu} in the xx- and yy-directions. The vertical dashed red line marks the zero-crossing lag used to compute the integral length scale. LxL_x and LyL_y are reported in the readout strip at the bottom.

Table 1:Integral scale methods.

MethodWhen to use
Zero crossingClean, well-resolved correlations with a clear zero crossing. Standard choice.
Exponential fitNoisy correlations without a clear zero crossing. Robust to noise in the tail.
1/e pointReturns the lag at which RR first drops to e−1≈0.368e^{-1} \approx 0.368 (interpolated).
Domain integralWhen no zero crossing exists and you want a conservative upper bound.

Temporal correlation (TR only)

Results show R(τ)R(\tau) vs time lag τ\tau in milliseconds, with integral time scale TT and Taylor microscale λt\lambda_t marked.

Multi-case comparison

Spatial 1D slice and temporal correlation curves support CSV import and overlay (Chapter ??).