CHARACTERISTICS OF NEAR-BOUNDARY FLOW THROUGH TWO RIPARIAN VEGETATION TYPES: A FLUME STUDY

Transcription

1 CHARACTERISTICS OF NEAR-BOUNDARY FLOW THROUGH TWO RIPARIAN VEGETATION TYPES: A FLUME STUDY L. C. Hopkinson ABSTRACT. Stream restoration designs frequently involve establishing vegetation on the bank face and floodplain, resulting in mixed patterns of riparian vegetation. This study examined the effects of varied patterns of simulated vegetation on time-averaged velocity and turbulence statistics. A flume experiment was conducted with eight submerged vegetation treatments composed of three characteristics: flexible grass height (0.7 and 3.9 cm), percent grass cover (100% and 50%), and the presence of herbaceous tubelings (with and without). Turbulence statistics (e.g., turbulence intensities, turbulent kinetic energy, and Reynolds stresses) were examined using near-bed velocity measurements. The 50% grass cover treatments introduced additional boundaries located at the vegetation edges that were not present in the 100% cover treatments. These additional edges resulted in complex flow patterns and velocity gradients, as well as heterogeneous distribution of turbulence statistics. Elevated Reynolds stresses located at the vegetation patch edges indicated potential for localized particle entrainment, consistent with previous observations of erosion. The presence of dowels also contributed to flow complexity while not significantly elevating overall turbulence. Results indicate that combining dense vegetation with varying lengths and sparsely positioned herbaceous tubelings will promote flow complexity. Keywords. Flume, Stream restoration, Submerged vegetation, Turbulence. Establishing riparian vegetation along streambanks and floodplains is a common technique used to control streambank retreat (Bernhardt et al., 2005; Bernhardt and Palmer, 2007). Bank vegetation increases geotechnical stability through root reinforcement (Simon and Collison, 2002; Wynn and Mostaghimi, 2006; Simon et al., 2006), and interactions between streambank vegetation and near-bank flow result in altered near-bank velocity and shear stress distributions (Thorne and Furbish, 1995; McBride et al., 2007; Hopkinson and Wynn, 2009). Generally, multiple planting methods (e.g., live staking, bareroot plantings, and potted plants) combined with seeding and other soil bioengineering systems are suggested for bank stabilization (USDA-NRCS, 1996; Doll et al., 2003). This mixed planting method results in a distribution of vegetation with varying densities, time of establishment, length, and flexibility. In addition to the desired stability management, this heterogeneous planting scheme may generate flow complexity (e.g., velocity gradients, eddies, etc.) that are desirable habitat features for aquatic organisms (Fausch and White, 1981; Hayes and Jowett, 1994; Crowder and Diplas, 2002). Flow through vegetation has been studied extensively for idealized vegetation in controlled experiments. Flow resistance and turbulence structure have been quantified for flow through an array of mimicked vegetation (Ikeda and Submitted for review in December 2011 as manuscript number SW 9561; approved for publication by the Soil & Water Division of ASABE in October The author is Leslie C. Hopkinson, Assistant Professor, Department of Civil and Environmental Engineering, West Virginia University, P.O. Box 6103, Morgantown, WV 26506; phone: ; Leslie.Hopkinson@mail.wvu.edu. Kanazawa 1996; Nepf, 1999; Garcia et al., 2004; Ghisalberti and Nepf, 2002, 2006; Tanino and Nepf, 2008; Liu et al., 2008). These studies primarily utilized a single vegetation type (e.g., rigid cylinders, plastic strips) located along the flume bed. This work was extended by defining the stepped velocity profile of flow through a combination of short and long rigid dowels (Liu et al., 2010) and through artificial vegetation with frontal densities that varied with depth (Lightbody and Nepf, 2006). Recent research has begun to address the impacts of streamside vegetation and vegetation patches (i.e., regions of vegetation with a finite area) on the flow field. The turbulence structure resulting from idealized vegetation located on the floodplain (Thornton et al., 2000; McBride et al., 2007; Yang et al., 2007), as streamside vegetation (White and Nepf, 2007; Afzalimehr and Dey, 2009; Zong and Nepf, 2010), and on sloping streambanks (Hopkinson and Wynn, 2009) has been quantified. High turbulent shear stresses are moved away from the streambank in the presence of dense, streambank vegetation (Hopkinson and Wynn, 2009). Recent research also characterized flow structure near vegetation/main channel boundaries of vegetation patches (Zong and Nepf, 2010; Siniscalchi et al., 2012). There is a need to expand this knowledge, considering flow transitions at vegetation patch boundaries (Nepf, 2012) and the flow effects resulting from heterogeneous vegetation patterns (Nezu and Sanjou, 2008). The objective of this research was to examine the effects of varied patterns of simulated vegetation on time-averaged velocity and turbulence statistics. Measurements of velocity were taken at a high spatial density within vegetation treatments combining flexible and rigid vegetation mimics that created two vertical layers. Artificial vegetation spacing was determined by a recent restoration plan. Transactions of the ASABE Vol. 55(6): American Society of Agricultural and Biological Engineers ISSN

2 METHODS FLUME SETUP AND VEGETATION TREATMENTS A flume experiment was conducted in a 1.21 m wide 6 m long flume (0% bed slope) at the St. Anthony Falls Laboratory (Minneapolis, Minn.). Flume vegetation models were constructed based on a planting pattern followed by an actual stream restoration site (Stroubles Creek, Blacksburg, Va.). The stream restoration design involved reshaping the banks and establishing bank and riparian vegetation (Thompson et al., 2012). In the Stroubles Creek restoration design, the lower bank of the restoration site was seeded with a wetland grass mixture, and herbaceous tubelings were planted in rows with a 0.15 m spacing and 0.76 m offset (Theresa Thompson, BSE Virginia Tech, personal communication, 27 August 2010). From these characteristics, eight vegetation treatments were constructed involving three vegetation factors: flexible grass length (0.7 and 3.9 cm), percent cover (100% and 50%), and herbaceous plantings (with and without) (fig. 1 and table 1). The grass was simulated using synthetic grass mats (RealGrass) with blade lengths of 0.7 cm (127 stems cm -2 ) and 3.9 cm (22 stems cm -2 ); the two grass lengths represented the wetland grass at two stages of development. Because vegetation often does not develop in a uniform pattern, the flexible vegetation density was reduced by 50% by randomly removing square sections (30.5 cm 30.5 cm) of the grass. This decreased area with vegetation was used to mimic patchy vegetation (figs. 1c and 1d). Wooden dowels were used to represent the herbaceous tubelings and were meant to represent the plants immediately after planting (diameter = 0.6 cm; length = cm). The dowels Table 1. Description of experimental flume runs. Flexible Percent Vegetation Flexible Discharge Stem Length Cover (m 3 s -1 ) (cm) (%) Herbaceous Plantings Run None Dowels None Dowels None Dowels None Dowels None Dowels None Dowels None Dowels None Dowels were positioned vertically in rows with a 0.15 m spacing and a 0.76 m offset, following the Stroubles Creek planting arrangement (Theresa Thompson, BSE Virginia Tech, personal communication, 27 August 2010; figs. 1b and 1d). MICROADV VELOCITY MEASUREMENTS The eight treatments were examined under two discharges (0.02 and 0.03 m 3 s -1 ; tailgate height = 13.3 cm; water depth = 17 to 18 cm). Experimental discharges were guided by previous studies (e.g., McBride et al., 2007; Afzalimehr and Dey, 2009) and by flow rates and velocities expected in the riparian area of Stroubles Creek. Discharge was measured with a sharp-crest weir located at the end of Figure 1. (a-d) Plan views of flume open channel indicating the location of flexible vegetation (shaded areas) and staggered dowels ( ) for one flexible vegetation stem length. The main flow direction is the positive x-direction, and velocity measurement locations (marked as ) are identified in the expanded views of the measurement regions. (e-l) Treatments for both flexible vegetation stem lengths: (e) S-100%, (f) S-100%-d, (g) S-50%, (h) S-50%-d, (i) L-100%, (j) L-100%-d, (k) L-50%, and (l) L-50%-d, where S = short grass (0.7 cm), L = long grass (3.9 cm), and d = dowel. For images (e) through (l), the main flow direction is into the camera TRANSACTIONS OF THE ASABE

3 the open channel and was verified with MicroADV velocity measurements. The discharge of 0.02 m 3 s -1 was the minimum discharge that could be measured with the sharp-crest weir. Flow was fully turbulent for all experiments (Reynolds numbers of the order of 18,700), and vegetation was fully submerged. Due to local fluctuations in the water surface, precise water surface slopes could not be measured over the short longitudinal distance of the open channel. Three-dimensional velocity was measured with a downlooking Sontek 16 MHz MicroADV (San Diego, Cal.) at 25 Hz for 120 s, meeting the minimum record length for calculating turbulence statistics (Buffin-Bélanger and Roy, 2005). Velocity was measured under steady flow in the measurement region ( cm) located approximately 20 flow depths from the flume entrance. Velocity measurements were taken in a grid pattern at one vertical distance per location (n = 98 for treatments with dowels; n = 99 for treatments without dowels; figs. 1a to 1d). Velocity was measured at a distance of 4 cm from the flume bed. This distance was as close to the flume bed as allowed without significant interference of the velocity probe for all treatments. The MicroADV velocity data were filtered to reduce erroneous data that may bias mean velocity values and turbulence characteristics. Instantaneous velocity values with an average signal correlation value less than 70% or an average signal to noise ratio less than 15 were removed (Wahl, 2000; McBride et al., 2007; Hopkinson and Wynn, 2009). All recorded time series were included in the data analysis because more than 75% of the instantaneous values remained after filtering (McBride et al., 2007; Hopkinson and Wynn, 2009). DATA ANALYSIS Each velocity time series was analyzed for turbulence intensities, turbulent kinetic energy, and Reynolds stresses. The root mean square (RMS) of the fluctuating velocities was calculated for each velocity component, representing the turbulence intensities: RMS u RMS v RMS w 2 u' = (1) 2 v' = (2) 2 w' = (3) where u is the downstream velocity component (m s -1 ), v is the lateral velocity component (m s -1 ), and w is the vertical velocity component (m s -1 ). This analysis required separating the instantaneous velocity fluctuations (u, v, and w ) from the time-averaged velocity ( u, v, and w ) for each velocity component. The overall average turbulent kinetic energy (TKE) was calculated: ( ) TKE = 0. 5ρ u' + v' + w' (4) where ρ is the fluid density (kg m -3 ). Reynolds stresses τ xz and τ xy, representing vertical and lateral momentum exchange, respectively, were also calculated: τ xz = ρ u' w' (5) τ xy = ρ u' v' (6) STATISTICAL ANALYSIS Statistical tests were used to determine differences among the near-bed time-averaged velocities and calculated turbulence statistics. First, the distributions of timeaveraged velocities and turbulence statistics were tested for normality using the Shapiro-Wilk test. The normal quartile plots were also examined. The data did not follow a Gaussian distribution; therefore, the Wilcoxon test was used to test for differences in medians. Differences among factors (i.e., discharge, flexible stem length, percent flexible cover, and herbaceous plantings) were tested for each timeaveraged velocity component ( u, v, and w ), TKE, Reynolds stress (τ xz and τ xy ), and turbulence intensity (RMS u, RMS v, and RMS w ). To eliminate the error contributed by position, data were blocked by measurement location (figs. 1a to 1d). Blocks without data (i.e., dowel location; figs. 1b and 1d) were removed from the analysis (Devore, 2000). A significance level of α = 0.05 was assumed for all tests. Statistical analysis was conducted using JMP software (v.9.0.0, SAS Institute, Inc., Cary, N.C.). RESULTS AND DISCUSSION For both discharges, u was typically two orders of magnitude greater than the time-averaged lateral ( v ) and vertical ( w ) velocity components for each individual run. The RMS values represent the strength of the velocity fluctuations and were on the same order of magnitude for all three components (10 0 cm s -1 ). On average, RMS u was an order of magnitude less than u, while RMS v and RMS w were an order of magnitude greater than v and w, respectively (table 2). These results indicate that turbulence fluctuations larger than the average velocity were frequent for the lateral and vertical components of velocity. To further compare the RMS values, relative turbulence intensity values were calculated by dividing the RMS values by the magnitude of the three-dimensional velocity vector component M for each velocity measurement ( M = u + v + w ). The relative turbulence in tensity values varied by location and treatment, but similarities for each component were observed. Excluding measurement locations directly behind a dowel, RMS u /M was greater than both RMS v /M and RMS w /M. Including all runs, relative turbulence intensity ranges were 0.11 to 0.54, 0.11 to 0.46, and 0.07 to 0.22 for RMS u /M, RMS v /M, and RMS w /M, respectively. As such, the downstream velocity component was the dominant contribution to overall turbulence; however, lateral and vertical components also substantially contributed. RMS u, RMS v, and RMS w contributed 46% to 54%, 31% to 36%, and 12% to 16% to 55(6):

4 Table 2. Median, range (in italics), and interquartile range (in parentheses) of velocity and turbulence data for each flume run: u is the timeaveraged downstream velocity; v is the time-averaged lateral velocity; w is the time-averaged vertical velocity; RMS u, RMS v, and RMS w are turbulence intensities; TKE is turbulent kinetic energy; and τ xz and τ xy are Reynolds stresses. Treatments for runs are defined in table 1. Run u v w RMS u RMS v RMS w TKE (N m -2 ) τ xz (Pa) τ xy (Pa) to to to to to to to to to 0.11 (2.2) (0.38) (0.36) (0.24) (0.14) (0.11) (0.08) (0.03) (0.09) to to to to to to to to to 0.15 (0.93) (0.44) (0.26) (0.23) (0.17) (0.08) (0.08) (0.02) (0.10) to to to to to to to to to 0.12 (1.87) (0.39) (0.45) (0.30) (0.15) (0.09) (0.10) (0.03) (0.08) to to to to to to to to to 0.14 (1.93) (0.39) (0.45) (0.29) (0.17) (0.07) (0.09) (0.02) (0.09) to to to to to to to to to 0.13 (0.97) (0.34) (0.26) (0.37) (0.20) (0.13) (0.17) (0.06) (0.09) to to to to to to to to to 0.17 (1.26) (0.36) (0.29) (0.34) (0.23) (0.11) (0.17) (0.05) (0.09) to to to to to to to to to 0.18 (1.04) (0.53) (0.63) (0.75) (0.39) (0.24) (0.37) (0.14) (0.15) to to to to to to to to to 0.16 (1.29) (0.58) (0.73) (0.79) (0.39) (0.23) (0.35) (0.12) (0.13) to to to to to to to to to 0.04 (1.07) (0.20) (0.24) (0.15) (0.12) (0.07) (0.04) (0.01) (0.03) to to to to to to to to to 0.08 (1.63) (0.26) (0.18) (0.30) (0.16) (0.11) (0.08) (0.02) (0.05) to to to to to to to to to 0.05 (1.18) (0.17) (0.30) (0.26) (0.14) (0.06) (0.06) (0.01) (0.03) to to to to to to to to to 0.07 (1.58) (0.34) (0.39) (0.29) (0.18) (0.07) (0.08) (0.02) (0.05) to to to to to to to to to 0.10 (0.93) (0.40) (0.21) (0.30) (0.31) (0.14) (0.17) (0.04) (0.07) to to to to to to to to to 0.11 (1.32) (0.40) (0.24) (0.29) (0.25) (0.12) (0.12) (0.03) (0.06) to to to to to to to to to 0.11 (0.84) (0.47) (0.53) (0.58) (0.43) (0.25) (0.26) (0.08) (0.07) to to to to to to to to to 0.11 (1.41) (0.39) (0.58) (0.70) (0.46) (0.30) (0.28) (0.09) (0.06) the overall TKE magnitude, respectively. For all runs, median τ xz and τ xy values ranged from 0.04 to 0.23 Pa and from to 0.01 Pa, respectively. The median τ xy values suggest little variability among treatments, but the interquartile range (IQR) within the treatments had a value of up to 0.15 Pa (run 7, table 2). This result indicates increased variability within the treatment with 50% flexible cover, as is discussed in the following sections. Lastly, τ xz was on the same order of magnitude as τ xy for the short grass treatments (10-2 Pa), and τ xz was an order of magnitude greater than τ xy for the long grass treatments (table 2), suggesting greater vertical momentum exchange in the long grass treatments. The following sections discuss the impacts of discharge, flexible stem length, vegetation arrangement, and vertical layers on time-averaged velocities, turbulence intensities, turbulent kinetic energy, and Reynolds stresses. EFFECTS OF DISCHARGE The effect of discharge was significant for timeaveraged velocity components u and w (p < ) and turbulence characteristics RMS u, RMS v, RMS w, TKE, and τ xz (p < ) (table 3). Time-averaged velocities for the high discharge runs (0.03 m 3 s -1 ) were 25% to 27% greater than those measured at 0.02 m 3 s -1. Turbulence intensities, TKE, and τ xz measured at the high discharge were 29% to 39%, 56%, and 69% greater than those measured at the low discharge, respectively. While there were differences in magnitude, spatial trends of velocities and turbulence for each vegetation treatment were the same for both levels of 2104 TRANSACTIONS OF THE ASABE

5 Table 3. Median values reported by factor (discharge, flexible stem length, percent flexible cover, and herbaceous plantings); p-values for each pair are reported in parentheses. Discharge Flexible Stem Length Percent Flexible Cover Herbaceous Plantings 0.03 m 3 s m 3 s cm 0.7 cm 100% 50% Plantings None u (cm s ) (<0.0001) (<0.0001) (0.9387) (0.5377) v (cm s ) (0.7775) (<0.0001) (0.0012) (0.7765) w (cm s ) (<0.0001) (<0.0001) (<0.0001) (0.0126) RMS u (<0.0001) (<0.0001) (0.1413) (<0.0001) RMS v (<0.0001) (<0.0001) (0.3404) (0.0004) RMS w (<0.0001) (<0.0001) (<0.0001) (0.0067) TKE (N m -2 ) (<0.0001) (<0.0001) (0.8452) (0.0001) τ xz (Pa) (<0.0001) (<0.0001) (0.0005) (<0.0001) τ xy (Pa) (0.8133) (<0.0001) (0.5040) (0.3397) discharge. These findings are consistent with McBride et al. (2007), who reported that the pattern and distribution of velocities and turbulence did not change with discharge level in a flume study with forested and non-forested riparian vegetation. EFFECTS OF STEM LENGTH The effect of stem length was significant for all velocity components ( u, v, and w ; p < ) and turbulence characteristics (RMS u, RMS v, RMS w, TKE, τ xz, and τ xy ; p < ) (table 3), likely due to the increase in overall drag. The long grass cross-sectional blockage factor (i.e., B = A v /A, where A v is the area of plants in the cross-section, and A is the area of the cross-section) was 5.5 times the short grass cross-sectional blockage factor for both the 100% and 50% flexible cover vegetation treatments. Increasing drag along the streambank will reduce flow within the bank vegetation and direct flow into the main channel. As observed by Rominger et al. (2010), this increased drag and associated flow acceleration at the vegetation edge can result in local scour along the vegetation boundary. A shear layer likely developed above the submerged, flexible vegetation. For the long grass treatments, it is expected that the shear layer was moved away from the flume bed as compared to the short grass treatments. Therefore, the median downstream velocity for the long grass (11.7 cm s -1 ) was less than that measured for the short grass (12.6 cm s -1 ) at the measurement location. Similarly, the median values for all calculated turbulence statistics were greater than those calculated for the short grass treatments. Turbulence intensities, TKE, and τ xz for the long grass treatments were 30% to 16%, 70%, and 150% greater than that measured for the short grass treatments, respectively (table 3). EFFECTS OF VEGETATION ARRANGEMENT (PERCENT FLEXIBLE COVER) The effect of percent flexible cover was significant for w and v (p 0.012), as illustrated in figure 2. The additional vegetation/main channel boundaries introduced by the vegetation patches created complex flow patterns that provide potential habitat benefits. For the treatments with 50% of flexible grass cover, velocity vectors are oriented in the positive z-direction at the dense vegetation patch entrance (x = 34.2 cm) and in the negative z-direction between patches (x = 7.6 to 26.6 cm, y = 0 to 15.2 cm; figs. 2c, 2d, 2g, and 2h). For treatments with 100% dense vegetation, 1% to 8% of the measurement locations had velocity vectors pointed in the negative z-direction; however, 19% to 23% and 44% to 46% of the velocity vectors were pointed in the negative z-direction for the 50% short grass and 50% long grass treatments, respectively. This flow heterogeneity was also illustrated by the dispersion of the data. Figure 2. Three-dimensional velocity vector plots showing mixing in patchy vegetation and outward exchange with the fully covered treatments. The main flow direction is in the positive x-direction. Plots only show the 0.03 m 3 s -1 results; trends were similar for the lowdischarge experiments (S = short grass, L = long grass, and d = dowel; exaggerated vertical scale). 55(6):

6 The IQR for the vertical velocity component was at least 20% larger in patchy treatments (50% flexible cover) as compared to treatments with 100% flexible cover (table 2). Due to the hydraulic resistance introduced by the vegetation, flow was directed over the vegetation patch (x = 32.4 cm; figs. 2c, 2d, 2g, and 2h). However, downstream of the measurement region (x > 40 cm), it is expected that vertical mass and momentum exchange occurred as flow moved over the patch. Assuming there is sufficient resistance to generate an inflection point in the velocity profile, a shear layer will form due to Kelvin-Helmholtz instability (Ghisalberti and Nepf, 2002; Poggi et al., 2004), and vortices will dominate vertical momentum exchange (Ghisalberti and Nepf, 2005). Two turbulence parameters, RMS w and τ xz, were determined to be statistically significant for the effect of percent flexible cover (p ; table 3). Figures 3a to 3h illustrates the distribution of τ xz in the measurement region for the high-discharge runs; trends for the low-discharge measurements were similar in distribution and approximately 40% less in magnitude. Especially apparent for the long grass, treatments with 50% flexible vegetation cover (both with and without dowels) resulted in a localized area of reduced τ xz ( 0.1 Pa; x > 20 cm, y > 15 cm; figs. 3g and 3h) where the vegetation was omitted for the full longitudinal length of the measurement region. Maximum τ xz values were located between long grass vegetation patches (y < 15 cm, x < 30 cm) and were at least double those measured in the area where the vegetation was omitted for the full longitudinal length of the measurement region. The maximum τ xz measured between the vegetation patches was 3.4 times the median τ xz value. Additionally, treatments with 50% long grass (runs 7, 8, 15, and 16) resulted in the greatest median τ xz and IQR at both discharges; the IQR values were more than double those of all other treatments at each discharge (table 2). This result indicates greater variability in the Reynolds stresses for the 50% long grass treatments due to the additional boundaries introduced to the patches of vegetation (i.e., vegetation/main channel edges). However, the τ xz values were of relatively low magnitude, ranging from to 0.26 Pa and from to 0.45 Pa for the low and high discharge runs, respectively. Fluvial entrainment, typically quantified by the excess shear stress model, results when the applied shear stress ex- Figure 3. (a-h) Distributions of the Reynolds stress tensor τ xz (Pa) in the measurement region, and (i-p) distributions of the Reynolds stress tensor τ xy (Pa) in the measurement region (S = short grass, L = long grass, and d = dowel). Only the high-discharge runs (0.03 m 3 s -1 ) of each treatment are shown, as the trends were similar. The diamond-shaped areas with no data were due to dowel interference with the ADV probe TRANSACTIONS OF THE ASABE

7 ceeds the critical shear stress of the soil (Partheniades, 1965). The applied shear stress is often represented by near-boundary Reynolds stresses (Nepf, 2012); therefore, the elevated τ xz values in this study indicate potential localized erosion on the edges of the vegetation patches. This result supports previous observations of erosion along the edge of vegetation patches in tidal flats (Temmerman et al., 2007) and along the edge of a vegetated, sandy point bar (Rominger et al., 2010). This localized erosion may be beneficial by creating complex bank geometry where biota can hide from predators and locate low-velocity refugia. Bank erosion also leads to riparian vegetation succession, essential for riparian diversity (Florsheim et al., 2008). While there was no significant difference in TKE when defined by percent cover, the spatial distribution indicated trends. For the high-discharge runs, contour plots indicate a location of increased TKE along the edges of vegetation patches, which was most apparent for the long grass treatments (TKE > 1 N m -2 ; figs. 4g and 4h). Maximum TKE measured between long grass vegetation patches (y < 15 cm, x < 30 cm) was 1.6 times the median TKE value for treatments with dowels and 1.5 times the median TKE value for treatments without dowels. This flow heterogeneity was also supported by the IQR values. For these treatments (runs 7 and 8), the IQR was more than double the IQR for all other treatments at the high discharge. The IQR for the 50% long grass treatments were 53% greater than for all other treatments at the low discharge (table 2). EFFECTS OF VERTICAL LAYERS The effect of two vegetation layers (with or without dowels) was significant for all turbulence intensities: RMS u, RMS v, and RMS w (p ). For treatments with dowels, median turbulence intensities were 2.23, 1.84, and 1.19 cm s -1 for RMS u, RMS v, and RMS w, respectively. Without multiple layers, median turbulence intensities values were 2.41, 1.97, and 1.27 cm s -1 for RMS u, RMS v, and RMS w, respectively. With two layers of vegetation (i.e., long grass with dowel treatments), the dowels likely reduced the velocity of the flow above the grass. This decreasing momentum exchange resulted in the decrease in RMS u for the treatments with dowels compared to treatments without dowels. The effect of two vegetation layers was also significant for TKE (p < ). Median TKE values decreased with the presence of dowels (table 3); however, localized areas of elevated TKE (TKE > 1 N m -2 ) were present within a downstream distance of 7.5 cm from the dowels (figs. 4b, 4d, 4f, and 4h). While the dense, flexible vegetation likely produced turbulence in the form of a shear layer over the vegetation, the turbulence produced by the dowels was due to vortex shedding from individual dowels. This difference in turbulence generation is illustrated in figure 4. Elevated TKE values were located downstream of the dowels because the TKE calculation incorporates all velocity components (eq. 4). The vortex shedding due to the dowels occurred in the lateral direction, making the velocity component (v) an important component to define turbulence at this location. The effect of multiple layers was not statistically significant for τ xy ; however, spatial patterns of τ xy similar to TKE were observed (figs. 3j, 3l, 3n, and 3p), and the τ xy values with the greatest absolute magnitude (-0.29 to Pa) were located directly downstream of the dowels. While a local area of elevated mean turbulence was present in the dowel treatments, the addition of the dowels did not result in an overall increase in TKE or Reynolds stresses (table 3). These results suggest that the presence of sparse vegetation promoted lateral momentum exchange while not significantly contributing overall instability. Figure 4. Distribution of turbulent kinetic energy (TKE, N m -2 ) in the measurement region of each treatment for the high-discharge flume runs; trends were similar in the low-discharge treatments (S = short grass, L = long grass, and d = dowel). The diamond-shaped areas with no data were due to dowel interference with the ADV probe. CONCLUSIONS Mixed planting methods are often suggested components of stream restoration designs and result in varied patterns of riparian vegetation. This research was intended to improve our understanding of flow complexity resulting from these planting methods. Vegetation patches composed of dense, flexible vegetation created boundaries where the edge of the vegetation met the main channel. These additional edges resulted in complex flow patterns and velocity gradients, as well as heterogeneous distribution of turbulence statistics. Elevated near-bed Reynolds stresses locat- 55(6):

8 ed at the vegetation patch edges, mainly for the long grass treatments, indicate a potential for localized particle entrainment, consistent with previous observations of erosion along vegetation edges. The presence of dowels also contributed to flow complexity while not significantly elevating overall turbulence. Elevated TKE and Reynolds stresses only occurred in close proximity to the dowels. Therefore, sparsely planted herbaceous tubules will contribute to lateral mass and momentum exchange. Dense grass patches at varying stem heights in combination with sparsely planted herbaceous tubelings will likely provide the benefits of flow complexity while minimizing the vegetation/main channel edges that are susceptible to particle entrainment. ACKNOWLEDGEMENTS Support for this work was provided by the STC program of the National Science Foundation via the National Center for Earth-Surface Dynamics under the agreement number EAR and by the National Science Foundation's ADVANCE IT Program under Award HRD Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation. The author would like to thank Dr. Theresa Thompson of the Department of Biological Systems Engineering at Virginia Tech for information about the Stroubles Creek restoration. REFERENCES Afzalimehr, H., and S. Dey Influence of bank vegetation and gravel bed on velocity and Reynolds stress distributions. Intl. J. Sediment Res. 24(2): Bernhardt, E. S., and M. A. Palmer Restoring streams in an urbanizing world. Freshwater Biol. 52(4): Bernhardt, E. S., M. A. Palmer, J. D. Allan, G. Alexander, K. Barnas, S. Brooks, J. Carr, S. Clayton, C. Dahm, J. Follstad- Shah, D. Galat, S. Gloss, P. Goodwin, D. Hart, B. Hassett, R. Jenkinson, S. Katz, G. M. Kondolf, P. S. Lake, R. Lave, J. L. Meyer, T. K. O Donnell, L. Pagano, B. Powell, and E. Sudduth Synthesizing U.S. river restoration efforts. Science 308(5722): Buffin-Bélanger, T., and A. G. Roy min in the life of a river: Selecting the optimal record length for the measurement of turbulence in fluvial boundary layers. Geomorphology 68(1-2): Crowder, D. W., and P. Diplas Vorticity and circulation: Spatial metrics for evaluating flow complexity in stream habitats. Canadian J. Fisheries Aquatic Sci. 59(4): Devore, J. L Probability and Statistics for Engineering and the Sciences. 5th ed. Pacific Grove, Cal.: Duxbury Thomson Learning. Doll, B. A., G. L. Grabow, K. R. Hall, J. Halley, W. A. Harman, G. D. Jennings, and D. E. Wise Stream Restoration: A Natural Channel Design Handbook. Raleigh, N.C.: North Carolina State University, North Carolina Stream Restoration Institute. Fausch, K. D., and R. J. White Competition between brook trout (Salvelinus fontinalis) and brown trout (Salmo trutta) for position in a Michigan stream. Canadian J. Fisheries Aquatic. Sci. 38(10): Florsheim, J. L., J. F. Mount, and A. Chin Bank erosion as a desirable attribute of rivers. BioScience 58(6): Garcia, M. H., F. Lopez, C. Dunn, and C. V. Alonso Flow, turbulence, and resistance in a flume with simulated vegetation. In Riparian Vegetation and Fluvial Geomorphology, S. J. Bennett and A. Simon, eds. Washington, D.C.: American Geophysical Union. Ghisalberti, M., and H. M. Nepf Mixing layers and coherent structures in vegetated aquatic flows. J. Geophysical Res. 107(C2): Ghisalberti, M., and H. M. Nepf Mass transport in vegetated shear flows. Environ. Fluid Mech. 5(6): Ghisalberti, M., and H. M. Nepf The structure of the shear layer in flows over rigid and flexible canopies. Environ. Fluid Mech. 6(3): Hayes, J. W., and I. G. Jowett Microhabitat models of large drift-feeding brown trout in three New Zealand rivers. North American J. Fisheries Mgmt. 14(4): Hopkinson L., and T. Wynn Vegetation impacts on nearbank flow. Ecohydrology 2(4): Ikeda, S., and M. Kanazawa Three-dimensional organized vortices above flexible water plants. J. Hydraul. Eng. 122(11): Lightbody, A. F., and H. M. Nepf Prediction of near-field shear dispersion in an emergent canopy with heterogeneous morphology. Environ. Fluid Mech. 6(5): Liu, D., P. Diplas, J. D. Fairbanks, and C. C. Hodges An experimental study of flow through rigid vegetation. J. Geophysical Res. 113: F04015, doi: /2008JF Liu, D., P. Diplas, C. C. Hodges, and J. D. Fairbanks Hydrodynamics of flow through double-layer rigid vegetation. Geomorphology 116(3-4): McBride, M., W. C. Hession, D. M. Rizzo, and D. M. Thompson The influence of riparian vegetation on near-bank turbulence: A flume experiment. Earth Surface Proc. Landforms 32(13): Nepf, H. M Drag, turbulence, and diffusion in flow through emergent vegetation. Water Resources Res. 35(2): Nepf, H. M Hydrodynamics of vegetated channels. J. Hydraulic Res. 50(3): Nezu, I., and M. Sanjou Turbulence structure and coherent motion in vegetated canopy open-channel flows. J. Hydro- Environ. Res. 2(2): Partheniades, E Erosion and deposition of cohesive soils. J. Hydraulics Div. ASCE 91(1): Poggi, D., L. Ridolfi., J. D. Albertson, and G. G. Katul The effect of vegetation density on canopy sub-layer turbulence. Boundary-Layer Meteorol. 111(3): Rominger, J. T., A. F. Lightbody, and H. M. Nepf Effects of added vegetation on sand bar stability and stream hydrodynamics. J. Hydraul. Eng. 136(12): Simon, A., and J. C. Collison Quantifying the mechanical and hydrologic effects of riparian vegetation on streambank stability. Earth Surface Proc. Landforms 27(5): Simon, A., N. Pollen, and E. Langendoen Influence of two woody riparian species on critical conditions for streambank stability: Upper Truckee River, California. J. AWRA 42(1): Siniscalchi, F., V. I. Nikora, and J. Aberle Plant patch hydrodynamics in streams: Mean flow, turbulence, and drag forces. Water Resources Res. 48: W01513, doi: /2011WR Tanino, Y., and H. M. Nepf Laboratory investigation of mean drag in a random array of rigid, emergent cylinders. J. Hydraul. Eng. 134(1): Temmerman, S., T. J. Bouma, J. Van de Koppel, D. Van der Wal, M. B. De Vries, and P. M. J. Herman Vegetation causes 2108 TRANSACTIONS OF THE ASABE

9 channel erosion in a tidal landscape. Geology 35(7): Thompson, T. W., W. C. Hession, and D. Scott StREAM Lab at Virginia Tech. Resource 19(2): 8-9. St. Joseph, Mich.: ASABE. Thorne, S. D., and D. J. Furbish Influences of coarse bank roughness on flow within a sharply curved river bend. Geomorphology 12(3): Thornton, C. I., S. R. Abt, C. E. Morris, and J. C. Fischenich Calculating shear stress at channel-overbank interfaces in straight channels with vegetated floodplains. J. Hydraul. Eng. 126(12): USDA-NRCS Chapter 16: Streambank and shoreline protection. In National Engineering Handbook Series, Part 650. Washington, D.C.: USDA Natural Resources Conservation Service. Wahl, T Analyzing ADV data using WinADV. In Proc Joint Conf. on Water Resources Engineering and Water Resources Planning and Mgmt. Minneapolis, Minn.: ASCE. White, B. L., and H. M. Nepf Shear instability and coherent structures in shallow flow adjacent to a porous layer. J. Fluid Mech. 593: Wynn, T., and S. Mostaghimi The effects of vegetation and soil type on streambank erosion, southwestern Virginia, USA. J. AWRA 42(1): Yang, K., S. Cao, and D. W. Knight Flow patterns in compound channels with vegetated floodplains. J. Hydraul. Eng. 133(2): Zong, L., and H. Nepf Flow and dispersion in and around a finite patch of vegetation. Geomorphology 116(3-4): (6):