Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-16T04:54:47.771Z Has data issue: false hasContentIssue false

On the surface expression of a canopy-generated shear instability

Published online by Cambridge University Press:  26 March 2019

Tracy L. Mandel*
Affiliation:
Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305, USASchool of Natural Sciences, University of California, Merced, CA 95343, USA
Saksham Gakhar
Affiliation:
Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305, USA
Hayoon Chung
Affiliation:
Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305, USA
Itay Rosenzweig
Affiliation:
Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305, USA
Jeffrey R. Koseff
Affiliation:
Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305, USA
*
†Email address for correspondence: tmandel2@ucmerced.edu

Abstract

Results are presented from a laboratory study on the free-surface signal generated over an array of submerged circular cylinders, representative of submerged aquatic vegetation. We aim to understand whether aquatic ecosystems generate a surface signature that is indicative of both what is beneath the water surface as well as how it is altering the flow. A shear layer forms over the canopy, generating coherent vortex structures which eventually manifest in the free-surface slope field. We connect the vortex properties measured at the surface with measurements of the bulk flow, and show that correlations between these quantities are adequate to create a parameterized model in which the interior velocity profile can be predicted solely from measurements taken at the free surface. Experimental surface observations yield a Strouhal number that is twice the most amplified mode predicted by linear stability theory, suggesting that vortices may evolve between generation at the canopy height and their manifestation at the water surface. Additionally, the surface signal continues evolving with distance downstream, with vortices becoming spread farther apart and the passage frequency gradually decreasing. By the trailing edge of the canopy, surface-impacting boils emerge for canopies with higher submergence ratios, suggesting a transition from coherent two-dimensional rollers to transversely varying structures.

Information

Type
JFM Papers
Copyright
© 2019 Cambridge University Press 
Figure 0

Figure 1.Schematic of experimental set-up and variables.

Figure 1

Table 1.Ranges of canopy height, density and flow conditions varied across experiments, with associated non-dimensional parameters. Estimates of uncertainty are as follows: canopy height is accurate to $\pm 0.3$ cm; canopy density to within $\pm 5~\text{stems}~\text{m}^{-2}$; depth-averaged velocity in an unobstructed flow to $\pm 1~\text{cm}~\text{s}^{-1}$.

Figure 2

Figure 2.Snapshots of perturbations in streamwise surface slope over time for case 20A2. Black-edged circles indicate the location of the vegetation elements, and shading that fills the circles represents the streamwise surface slope $\unicode[STIX]{x2202}\unicode[STIX]{x1D702}/\unicode[STIX]{x2202}x$. The surface slope is interpolated between elements to better show features. The horizontal axis indicates the streamwise direction $x$, and vertical axis represents the transverse direction $y$. Snapshots are of the centre 30 cm of a 60 cm wide test section. The apparent displacement of the tips of the cylindrical elements is very small, of the order of 1–2 pixels, so positions of the elements appear identical in all frames.

Figure 3

Figure 3.Vortex speed, as measured at the surface, for varying canopy height, density and flow speed. Measurements at the free surface are compared to LDA measurements of depth-averaged velocity $\overline{U}$ and $U_{h_{c}}$. It is apparent that the speed of the vortex at the free surface is a sufficient analogue for common representations of the shear layer velocity.

Figure 4

Figure 4.Ratio of vortex velocity to velocity at the canopy height as a function of submergence ratio. The velocity ratio is highest for the densest canopy case (black circles), indicating that vortex centres are more significantly displaced upwards when the relative porosity of the canopy is lower. The dashed line at $U_{v}/U_{h_{c}}=1.8$ indicates the asymptotic limit postulated by Nepf (2011).

Figure 5

Figure 5.Vortex length scales and predominant frequency at the free surface for case 20A1 versus distance downstream. (a) Length scales are normalized by the canopy height $h_{c}$. A coherent signal emerges at the free surface around $x/L_{c}\sim 0.4$, at which time spatial correlations become significant. The vortex size $L_{v}$ remains fairly fixed over the canopy length, while the vortex spacing $\unicode[STIX]{x1D706}_{v}$ grows at a constant rate. The vortex spacing has almost doubled by the end of the canopy. (b) The predominant frequency observed at the surface shows a similar linear trend with downstream distance. Shaded error bars indicate the order of magnitude of error in each measurement point.

Figure 6

Figure 6.Vortex length scales as a function of canopy geometry. (a) The vortex size shows an inverse proportionality to the canopy height $h_{c}$. While previous work found that in unconfined canopies, vortex size ${\sim}h_{c}$, these results in more strongly submerged canopies find the inverse, that vortex size decreases for larger canopy heights; $h_{c}$ of 10 cm corresponds to a submergence ratio $H/h_{c}$ of 3.1, while $h_{c}$ of 20 cm corresponds to $H/h_{c}$ of 1.6. (b) The vortex spacing shows a similar dependence on the canopy height, in addition to dependence on the canopy density, here represented by the dimensionless planform density $\unicode[STIX]{x1D706}_{p}$. Denser canopies generate a larger vortex spacing relative to sparser canopies.

Figure 7

Figure 7.Comparison of vortex size estimates from this study and previous studies. Data for rigid model vegetation are shown as filled symbols, and data for flexible vegetation as open symbols. The data of Okamoto & Nezu (2009) are comprised of the integral length scale at the canopy height for both flexible and rigid vegetation. Okamoto et al. (2016) also computed the integral length scale, but at the canopy height and near the surface. Finally, the data of Ghisalberti & Nepf (2002) are estimates of vertical vortex size based on $\unicode[STIX]{x1D703}$ and $t_{ml}$. The data for the present study are the vortex size estimates $L_{v}$, based on the decorrelation length. Best fit lines in (b) show an exponential fit of the form $y=a\text{e}^{bx}$ for data from previous work ($a=5.5$, $b=-0.71$), and all work including the data from the present study ($a=2.5$, $b=-0.48$).

Figure 8

Figure 8.Shear length scale versus vortex length and spacing. (a) The shear length scale shows very little correlation with the vortex size at the surface. (b) The shear length correlates reasonably well with the vortex spacing at the end of the canopy. A linear relation is found with $R^{2}=0.39$.

Figure 9

Figure 9.Correction of the inflection point height $h_{i}$ for sample velocity profiles 20A1, 15A1 and 10A1. Black circles indicate mean velocities measured using LDA. The solid blue line shows the model of $U(z)$ using the canopy height as the location of inflection. The dot-dashed black line shows (3.5), where a corrected inflection height has been fitted from the data. Both models use values of $\unicode[STIX]{x0394}U$, $\unicode[STIX]{x1D703}$ and $\overline{U}$ from subsurface measurements. The horizontal dotted line indicates the canopy height.

Figure 10

Figure 10.Non-dimensional growth rate and oscillation frequency as a function of wavenumber for all 27 experimental cases. Colour bars show non-dimensional planar density parameter $\unicode[STIX]{x1D706}_{p}$, a representation of canopy density. Canopies with the highest density experience faster growth rates of the instability, and shed vortices at a lower frequency.

Figure 11

Figure 11.Observed instability frequency from measured surface slope spectrum versus computed values of the instability frequency, given measured instability wavelengths. Black circles show the unstable mode computed in the above analysis. Red squares show the commonly used Strouhal shedding frequency defined in Ho & Huerre (1984).

Figure 12

Figure 12.Non-dimensional instability growth rate versus range of Strouhal numbers for three sample velocity profiles. Lines indicate the theoretical growth rate of the most unstable mode for a range of forcing wavenumbers/frequencies. Shaded region indicates the relative frequency of occurrence of experimentally observed Strouhal numbers, with darker regions indicating the most common $St$.

Figure 13

Table 2.Parameterization of interior flow variables with surface variables.

Figure 14

Figure 13.Correlation of surface measurements with interior flow measurements to obtain parameterizations for the velocity profile, equation (3.5). While in (a), (c) and (d), all densities are plotted as black circles, (b) shows how scatter in $\unicode[STIX]{x0394}U$ is a function of canopy density $\unicode[STIX]{x1D706}_{P}$. We expect this single linear fit to perform best for the intermediate density, $\unicode[STIX]{x1D706}_{P}=0.08$.

Figure 15

Figure 14.Performance of surface model parameterization versus LDA data for the intermediate Reynolds number case. Black circles represent velocity measurements taken using LDA. The canopy height is indicated by the light dashed horizontal line. The dot-dashed black line represents the hyperbolic tangent model of (3.5) using inputs based on LDA data, and an optimized inflection point height. The dashed red line represents the hyperbolic tangent model using inputs based solely on surface measurements using FS-SS. The surface-based model performs reasonably well in predicting the measured velocity profile, with the main error occurring in predictions of $\unicode[STIX]{x0394}U$, the difference between the free-stream and within-canopy velocities.

Figure 16

Figure 15.Performance of surface model parameterization versus measured values for all experimental cases. The root-mean-squared error for each case, normalized by $\unicode[STIX]{x0394}U$, is plotted against the non-dimensional canopy frontal area $\unicode[STIX]{x1D706}_{f}$. The model performs reasonably well, with RMSE ${\leqslant}10\,\%$ for the majority of the cases. The largest error occurs for the lowest $Re$ cases.

Figure 17

Figure 16.Power spectral density at the measured peak instability frequency versus downstream distance $x/L_{c}$ for all experimental cases. Rows show differing canopy heights, and columns show different flow speeds. Markers indicate the canopy density. The densest canopy generates the most power at the instability frequency. Note the significant dropoff in power that occurs for many cases between $x/L_{c}=0.5$ and 1.

Figure 18

Figure 17.Images of a representative ‘boil’ structure impacting the free surface and propagating downstream (from right to left) towards the trailing edge of the canopy for experimental case 20A2. The side wall of the flume is in the lowermost portion of the images; the boil structure at the top of the image is approximately 30 cm from the side wall, moving along the centreline of the flume. By image (v), the centre boil structure is approximately 5.5 cm in diameter. The circumference of the primary upwelling structure is marked with a dotted circle. Small waves can be seen radiating away from the centre of the structure, causing optical distortion beyond its edges, so that the area of disturbed fluid is larger than the boil itself. Note that these photos were taken with a wider-angle lens than those used for actual FS-SS measurements. The arrow points to the region of lensing due to surface disturbance.

Figure 19

Figure 18.Distance to vortex–surface interaction. (a) PSD dropoff distance $L_{PSD}$ versus estimated boil eruption point $L_{1}$ computed using naive single-value estimate of boil diameter $b$. (b) $L_{1}$ computed using fitted canopy height-dependent estimate of boil diameter. Dashed lines indicate a 1:1 correspondence.

Figure 20

Figure 19.Estimates of adjustment lengths versus canopy density for the full range of experimental conditions in this study. Three different estimates of adjustment length are used. The first two, $X_{D}$, estimate the distance required for the mean flow to adjust to the presence of a canopy. The third, $X_{\ast }$, estimates the distance for the mixing layer to develop. The dotted line indicates the end of the canopy at $X/L_{c}=1$.