Brush piles (i.e., trees and large woody debris) are often installed in reservoirs to supplement fish habitat. The retention and dimensional change of brush piles after installation is important information that can be used to maximize the effectiveness of this management action. We evaluated the retention and dimensional change of 70 eastern red cedar Juniperus virginiana and bald cypress Taxodium distichum brush piles in an embayment of a drawdown reservoir up to four annual cycles of submergence and exposure. We used satellite imagery to supplement our onsite measurements of retention. We also examined spatial patterns of brush pile retention and dimensional change. Brush piles were lost at 10% per year, and their volume was lost at 14% per year. We compared our rates of brush pile retention and dimensional change with those from a holdout data set of 50 brush piles. Estimates between data sets did not differ statistically. Spatial patterns of retention and dimensional change coincided with morphological features in our study area, suggesting that retention and dimensional change is influenced by variable physical forces (e.g., wave action and flow) at installation locations. Our estimates of brush pile retention and dimensional change can be used to generally sustain desirable brush densities. For example, to maintain a fixed total volume of brush in our study embayment, roughly 23% of the total brush volume installed would need to be replaced annually. Similar research in reservoirs managed for other purposes is needed, as length and cycle of inundation could lead to variable rates of retention and dimensional change. Additionally, advancements into computer-assisted detection and volume estimation could reduce the time and effort needed to monitor brush piles.

Habitat degradation is a common problem across the aging population of U.S. reservoirs (Miranda et al. 2010; Kaufmann et al. 2014). Loss of structural complexity and the formation of mudflats are prominent issues in flood control reservoirs (Miranda and Krogman 2015; Krogman and Miranda 2016) and can have several undesirable impacts on reservoir fisheries, such as low recruitment and decreased fish abundance (Bolding et al. 2004; Willis et al. 2004; Carmignani and Roy 2017; Miranda 2017). With over 600 large reservoirs (>100 ha) in the contiguous United States used for flood control plus recreation or fish habitat, or both (Rodgers 2017), loss of structural complexity and mudflat formation are widespread concerns.

Supplementing fish habitat in areas void of structural complexity and cover with aggregates of woody debris (hereafter referred to as brush piles) is a common practice in fisheries management (Tugend et al. 2002; Stone et al. 2012; Miranda 2017). The addition of brush piles to degraded systems and their effects on fisheries (e.g., increased cover for fishes and higher catch rates) are well studied and reported in the literature (Johnson and Lynch 1992; Barwick 2004; Bolding et al. 2004; DeBoom and Wahl 2013; Allen et al. 2014; Miranda 2017; Hatcher et al. 2019). Brush piles are popular for their ready availability, low cost, easy installation, and ability to naturally degrade (Magnelia et al. 2008; Allen et al. 2014). However, the natural degradation of brush piles is both an advantage and a disadvantage; on the one hand, brush piles do not introduce microplastics into freshwater systems as with synthetic, slowly degrading materials (e.g., polyvinyl chloride; Eerkes-Medrano et al. 2015), but on the other hand, brush piles require planning and upkeep to maintain their benefits (Baumann et al. 2016).

Although their effects are well studied, the retention (i.e., remaining at or near the installation site) and dimensional change (i.e., change in volume) of brush piles after installation have not been examined closely (Czarnecka 2016) and create ambiguity when planning maintenance of brush piles to meet management objectives. Previous literature on brush pile retention and dimensional changes has relied on qualitative expert opinions and anecdotal evidence (Mabbott 1991; Bolding et al. 2004; Stone et al. 2012; Baumann et al. 2016; Miranda 2017). Quantitative rates of retention and dimensional change are needed for managers to develop a schedule that efficiently mitigates reduced fisheries due to loss of structural complexity. Because water levels in flood control reservoirs change drastically, there is opportunity to quantify retention and structural dimensions of brush piles when they are exposed.

In this study, we used onsite surveys to quantify retention and dimensional change of brush piles in a drawdown reservoir one to four flood cycles after installation and hypothesized that retention and dimensions would decrease over time. We supplemented our onsite surveys of retention with visual inspections of satellite imagery taken one and two flood cycles after installation. We chose this methodology because the use of satellite imagery can potentially reduce the time and costs to monitor brush pile retention or enhance monitoring efforts. Additionally, basic visual inspection of satellite and aerial imagery to monitor fish habitats at small scales provides a proof of concept for computer-assisted assessments at larger scales (Finkbeiner et al. 2001; Molthan and Jedlovec 2011; Ward et al. 2014; Grimm et al. 2016). We used measurements from our surveys to characterize the spatial relationships of brush pile retention and dimensional change and associated spatial patterns with reservoir features. We hypothesized that rates of brush pile loss would be greater in deeper water and closer to the main stem of Long Branch Creek. We expected the opposite relationship between dimensional change and water depth (i.e., higher rates of volume loss in shallower depths). We hypothesized that dimensional changes would vary spatially but that neighboring brush piles would experience similar dimensional change. Last, we used estimated rates of brush pile retention and dimensional change to demonstrate scheduling of brush pile replenishment in a management scenario where the objective is to maintain the initial total volume of brush pile habitat.

Study site and data collection

We conducted our study in the Long Branch Creek embayment of Enid Lake in northwest Mississippi (Figures 1a, 1b, and 1d). Enid Lake is a 68-year-old U.S. Army Corp of Engineers (USACE) flood control reservoir that drains 1,450 km2 with a mean and maximum depth of 15.5 and 22.5 m, respectively. Enid Lake experiences extensive mudflat formation due to annual winter drawdowns that change the surface area from 6,500 ha at normal pool (76 m) over spring and summer to 2,500 ha at conservation pool (70 m) over fall and winter (Figure 1c; Hatcher 2018; Hatcher et al. 2019). Approximately 40% of the drawdown area lies between the 74- and 76-m elevation contours. Other hydrological forces in the embayment result from runoff into Long Branch Creek and wind-driven waves, which are typically between 0.15 and 0.30 m in height with a northwest wind (4.83 km fetch) and between 0.05 and 0.20 m with a southeast wind (805 m fetch; CERC 1984; Minar 2011; Data S1, Code S1, and Methods S1, Supplemental Material). During conservation pool on 17 February 2017, 196 brush piles were installed in the 72- to 76-m contours on either side of Long Creek Branch (Figure 1d) by volunteers during a public engagement event coordinated by the USACE (Hatcher 2018). Brush piles consisted of one to five small (average of 5.61-m tall; SD, 1.84) eastern red cedar Juniperus virginiana and bald cypress Taxodium distichum secured to the reservoir bottom with concrete masonry blocks (i.e., cinder blocks).

Figure 1.

Map of Enid Lake, Mississippi, Long Branch Creek embayment (a and b), water levels during water years (October–September) from 2016 to 2020 compared with the guide curve (c), and areas of brush pile (fish habitat/attractors) installation in spring 2017 (d).

Figure 1.

Map of Enid Lake, Mississippi, Long Branch Creek embayment (a and b), water levels during water years (October–September) from 2016 to 2020 compared with the guide curve (c), and areas of brush pile (fish habitat/attractors) installation in spring 2017 (d).

Close modal

On 17 March 2017, four weeks after installation (no flood cycles elapsed), we recorded a global positioning system (GPS) location for each brush pile and measured volume (m3) as maximum length (m) × maximum width (m) × maximum height (m) with a tape measure (Hatcher 2018). We revisited 70 of the brush pile locations on the east side of Long Branch Creek (Figure 1d, Installation area E) on 18 December 2019 (three flood cycles after installation) and measured brush pile retention (i.e., remaining at or near the installation site) and volume. We considered a brush pile retained if we found it within a 3-m radius of its original GPS location (3 m was the accuracy of our GPS/wide area augmentation system receiver; buy.garmin.com/en-US/US/p/87771) and attached to cinder blocks. If a brush pile was retained, we estimated volume as originally measured by Hatcher (2018). The process of locating and measuring brush piles took approximately 3.5 min/brush pile. We could not revisit additional brush pile locations on the west side of Long Creek Branch (Figure 1d, Installation area W) due to flooded and soggy conditions, which prevented access. On 9 November 2020 (four flood cycles after installation), we revisited the same 70 locations on the east side of Long Creek Branch and again measured brush pile retention and volume. On this occasion, we were able to visit an additional 50 locations on the west side of Long Branch Creek. We withheld measurements from these 50 additional locations (holdout data) from calculations of summary statistics and spatial analyses and used these measurements to evaluate estimated coefficients of brush pile retention rates and rates of dimensional change (see below). Measurements from the onsite surveys described above can be found in Data S1, Supplemental Material.

To supplement our onsite surveys, we used satellite imagery to monitor brush pile retention after installation and before our last onsite survey (i.e., photogrammetry). To do this, we first uploaded the 120 GPS locations of the brush piles we had visited during onsite surveys to Google Earth Pro (version 7.3.3; www.google.com/earth). We then used the historical imagery tool to select reference imagery (the most recent date before brush pile installation on 17 February 2017) and satellite imagery for analysis (between brush pile installation and the last onsite survey on 9 November 2020). We limited all imagery to the months during conservation pool when brush piles would be exposed (i.e., November to February). Imagery was available on 8 December 2016 for use as reference imagery (Figure 2a). Imagery for visual assessment of brush pile retention was available on two occasions between installation and the last survey date: 10 November 2017 (Figure 2b; one flood cycle after installation) and 15 January 2019 (two flood cycles after installation). We visually inspected each GPS location for the presence of a brush pile and considered a brush pile retained if any portion of it was within a 3-m radius of its GPS location (Data S1, Supplemental Material). The process of locating brush piles in satellite imagery took approximately 45 s/brush pile.

Figure 2.

Satellite imagery from Tributary A in Figures 1 and 4 in the Long Branch Creek embayment of Enid Lake, Mississippi, before brush pile installation (8 December 2016; top) and one flood cycle after installation (10 November 2017; bottom). Arrows highlight brush pile installation locations; not all locations are highlighted. White circles illustrate a 3-m radius around brush pile locations. Note that the brush pile in location BP-018 was lost between installation on 17 March 2017 and the date when the first satellite image was taken after installation (10 November 2017; bottom); other locations that are highlighted retained brush piles.

Figure 2.

Satellite imagery from Tributary A in Figures 1 and 4 in the Long Branch Creek embayment of Enid Lake, Mississippi, before brush pile installation (8 December 2016; top) and one flood cycle after installation (10 November 2017; bottom). Arrows highlight brush pile installation locations; not all locations are highlighted. White circles illustrate a 3-m radius around brush pile locations. Note that the brush pile in location BP-018 was lost between installation on 17 March 2017 and the date when the first satellite image was taken after installation (10 November 2017; bottom); other locations that are highlighted retained brush piles.

Close modal

To assess that our methodology for determining retention of brush piles from satellite imagery was valid, we compared retention of brush piles in our satellite surveys with the onsite survey conducted on 9 November 2020, four flood cycles after installation. We presumed that locations with brush piles present in the onsite survey must have brush piles at those locations in the satellite surveys and that locations without brush piles in the satellite surveys should not have brush piles at those locations in the onsite surveys. If our methodology passed the validity checks, we would use retention measurements from satellite surveys for the 70 locations on the east side of Long Branch Creek in calculations of summary statistics and spatial analyses and in the estimation of brush pile retention rates. Also, we would add retention measurements from satellite surveys for the 50 locations on the west side of Long Branch Creek to the holdout data.

Summary statistics and rates of loss

We used the number of brush piles lost (i.e., not retained) from satellite (if applicable) and onsite surveys at brush pile locations on the east side of Long Branch Creek (Figure 1d, Installation area E; i.e., not the holdout data) to calculate the percentage of brush piles that were lost from their location since installation (17 February 2017). We assessed brush pile dimensional change as the difference between the initial volume (measured on 17 March 2017) and volume measured during the onsite surveys (18 December 2019 [three flood cycles] and 9 November 2020 [four flood cycles]) divided by the initial volume. We then calculated the mean, standard deviation (SD), minimum, and maximum dimensional change for each onsite survey (Data S1, Supplemental Material). Microsoft Excel was used to calculate summary statistics.

Using the number of brush piles retained since installation in satellite (if applicable) and onsite surveys, including 17 March 2017 (no flood cycles), we estimated the rate of brush pile loss. Using the mean volume of brush piles from onsite surveys, again including 17 March 2017, we estimated the rate of dimensional change for the average brush pile. We assumed a constant rate of brush pile loss and estimated instantaneous brush pile loss ZN using a log-linear regression (i.e., linearized catch curve analysis; Miranda and Bettoli 2007), with the loge number of brush piles as the response variable and number of flood cycles since installation as the independent variable. Similarly, we assumed a constant rate of loss in brush pile volume and estimated instantaneous loss in volume ZV using a log-linear regression, with loge average brush pile volume as the response and number of flood cycles since installation as the independent variable. We then used the holdout data to estimate instantaneous brush pile loss (ZNholdout) and instantaneous loss in brush pile volume (ZVholdout) to corroborate our estimates. If ZNholdout and ZVholdout fell within the 95% confidence intervals (95% CI) of ZN and ZV, respectively, we considered ZN and ZV as valid in similar settings. We calculated annual rates of brush pile loss and loss in brush pile volume using the general form 100 × (1 – e−Z). We used the lm function in the stats R package (R Core Team 2019) for log-linear regressions and the confint function (stats package) to calculate confidence intervals.

Spatial analyses

We assigned each brush pile location to one of four ordinal depth categories relative to normal pool (76 m): >75, 75–74.1, 74–73.1, and <73 m (bathymetry to derive continuous measures of depth was not available). We determined each brush pile depth category by using the depth contours in the i-Boating Marine Charts App for Windows (http://www.gpsnauticalcharts.com/). GPS coordinates were uploaded using the Load GPX/KML function in the Route Manager tool. We tested for an association between brush pile retention among depth categories using logistic regression (glm function in the stats R package [R Core Team 2019]), and the number of flood cycles since installation, depth category, and their interaction as independent variables. We tested for differences in brush pile dimensional change among depth categories using two-factor analysis of variance (ANOVA; glm function in the stats R package [R Core Team 2019]), with number of flood cycles since installation, depth category, and their interaction as independent variables. We checked residuals in both models for spatial dependence using Moran's I test (Plant 2012; lm.morantest function in the spdep R package [Bivand et al. 2013]). If we detected spatial autocorrelation, we refit the logistic regression and two-factor ANOVA with weighted-sum auto-covariates (lag.listw function in the spdep R package [Bivand et al. 2013]; i.e., auto-logistic and auto-normal models), respectively, to account for spatial dependence (Plant 2012; Bardos et al. 2015). We used the three nearest brush piles as spatial connections when constructing spatial connections in auto-covariate construction (knearneigh, knn2nb, make.sym.nb, and the nb2listw [style = “B” for valid {i.e., weighted-sum} neighborhood weightings; see Bardos et al. 2015 for details] functions in the spdep R package [Bivand et al. 2013]).

Next, we used join-count analysis (joincount.mc function in the spdep package [Bivand et al. 2013]) to test if locations where brush piles were lost neighbored one another. We analyzed retention data from each survey separately (we used satellite surveys only if applicable). The join-count analysis compares the spatial connectedness of neighboring GPS locations with an observed categorical variable (i.e., lost or retained) to a set of permutations of the categorical variable across all available locations (Plant 2012). We calculated a probability value (P value) by ranking the join-count statistic of the observed pattern among statistics calculated from the set of permutations. We then divided the rank of the join-count statistic for the observed pattern by the total number of permutations plus the observation (e.g., 985/(999+1) = 0.985). We then tested if dimensional change showed spatial correlation using Moran's I test (we analyzed each survey separately; Plant 2012; moran.mc function in the spdep package [Bivand et al. 2013]). Moran's I test is similar to the join-count analysis except that it analyzes correlations between spatial connections for a continuous variable. In the join-count analyses and Moran's I tests, we used the three nearest brush piles as spatial connections, and we constructed 9,999 permuted sets for comparisons and P value calculations, as described above for the join-count analysis. We performed mapping using the maps (Becker et al. 2018) and ggmap (Kahle and Wickham 2013) packages in R (R Core Team 2019). We interpreted the strength of statistical evidence for all statistical analyses by the size of the P value, with smaller values indicating greater strength.

Application to a management scenario

We present a hypothetical management scenario in our study embayment to illustrate how rates of brush pile retention and dimensional change can be used to estimate the number of brush piles needed to maintain a fixed volume of brush pile habitat. We estimate the total volume of brush piles that must be replenished every t years (Rt) to maintain the original or desired volume as
where 0 is the initial average brush pile volume, N0 is the original number of brush piles installed, and ZN and ZV are estimates of instantaneous brush pile loss and instantaneous loss in brush pile volume, respectively (see above). We estimated the number of brush piles that must be replenished every t years (t) by dividing Rt by 0. We also used the upper and lower 95% CI values for ZN and ZV to create a range of plausible values for Rt and brush piles t to replenish. In our hypothetical management scenario, we calculate Rt and t given 0 = 50 m3 and N0 = 100 for an initial total volume of 5,000 m3. We performed calculations in R (Code S1, Supplemental Material; R Core Team 2019); we also provided a Microsoft Excel spreadsheet that can be used to execute basic calculations (Table S1, Supplemental Material).

Validity of satellite imagery to monitor brush pile retention

Where brush piles were present in the 9 November 2020 onsite survey, they were also evident in the satellite surveys (10 November 2017 and 15 January 2019). Where locations did not have brush piles in the satellite surveys, they were absent in the 9 November 2020 onsite survey. Because the validity checks were passed, we included retention measurements for the 70 locations on the east side of Long Branch Creek (Figure 1d, Installation area E) in calculations of summary statistics and spatial analyses and in estimation of brush pile retention rates, and we added the retention measurements for the 50 locations on the west side of Long Branch Creek (Figure 1d, Installation area W) as the holdout data.

Summary statistics and rates of loss

An average of 6.5 (SD, 3.7; minimum–maximum, 3–11) brush piles were lost from their installation locations between flood cycles; those retained lost an average volume of 6.3 m3 (SD, 1.4; min–max, 5.2–7.3; Table 1). On 18 December 2019, 78.2% (43/55) of the brush piles that were retained decreased in volume by an average of 40.1% (SD, 18.8%; min–max, 2.7–98.4%). However, 21.8% (12/55) of the brush piles retained increased in volume by an average of 18.7% (SD, 15.0%; min–max, 1.7–55.2%). By 9 November 2020, 84.1% (37/44) of remaining brush piles had decreased in volume by an average of 50.3% (SD, 24.1%; min–max, 7.6–99.2%); 15.9% (7/44) of brush piles increased in volume by an average of 13.6% (SD, 8.5%; min–max, 4.0–28.0%). Instantaneous brush pile loss ZN was 0.10 (95% CI = [0.16, 0.05]; t = −6.17, P < 0.01, r = 0.96), and instantaneous loss in volume ZV was 0.15 (95% CI = [0.25, 0.05]; t = −19.10, P = 0.03, r = 0.99; Figures 3a and 3b). Estimates for ZNholdout (0.09) and ZVholdout (0.14) fell within the 95% CI values of our log-linear regressions for ZN and ZV, respectively (see Data S1, Supplementary Material, for the summary statistics of holdout data used to estimate ZNholdout and ZVholdout). This suggests that the estimates for ZN and ZV are valid for similar settings. Given the estimates of ZN and ZV, we estimated annual brush pile loss at 10.0% and annual decrease in brush pile volume at 13.9% using the general equation 100 × (1 – eZ). We extrapolated estimates of the number of brush piles lost and the decrease in volume of the average brush pile for t 1–10 years in Figures 3a and 3b.

Table 1.

The number and percent of brush piles lost in the Long Creek Branch embayment of Enid Lake (Mississippi) from 2017 to 2020 by depth category over the length of this study are given in the third column, average brush pile volume with standard deviation (SD) is given in the fourth column, and average dimensional change with SD is given in the fifth column.

Figure 3.

Brush pile loss curves (i.e., not retained; a) and loss in volume curves (b) of the initial number (N0 = 70) and volume (0 = 47.9 m3) of brush piles in the Long Branch Creek embayment, Enid Lake, Mississippi, from 2017 to 2020. The solid lines in subfigures represent estimates, while dotted lines represent the upper and lower 95% confidence intervals. The dashed lines represent the estimates from the holdout data set.

Figure 3.

Brush pile loss curves (i.e., not retained; a) and loss in volume curves (b) of the initial number (N0 = 70) and volume (0 = 47.9 m3) of brush piles in the Long Branch Creek embayment, Enid Lake, Mississippi, from 2017 to 2020. The solid lines in subfigures represent estimates, while dotted lines represent the upper and lower 95% confidence intervals. The dashed lines represent the estimates from the holdout data set.

Close modal
Figure 4.

Map of brush pile locations (fish habitat or attractors) installed in the Long Branch Creek embayment of Enid Lake, Mississippi, in spring 2017. In (a) and (b), open diamonds represent locations where brush piles were retained; × represents locations where brush piles were lost by (a) 10 November 2017 and (b) 15 January 2019. In (c) and (d), filled circles indicate brush piles with increased volume, unfilled circles indicate brush piles with decreased volume, and × represents where brush piles were lost. The size of circles represents relative difference in dimensional change (c) from 17 March 2017 to 18 December 2019 and (d) from 17 March 2017 to 9 November 2020. Tributaries are labeled as “A” and “B”, the long-dash line represents the shoreline, and the short-dash line represents the 75-m elevation contour.

Figure 4.

Map of brush pile locations (fish habitat or attractors) installed in the Long Branch Creek embayment of Enid Lake, Mississippi, in spring 2017. In (a) and (b), open diamonds represent locations where brush piles were retained; × represents locations where brush piles were lost by (a) 10 November 2017 and (b) 15 January 2019. In (c) and (d), filled circles indicate brush piles with increased volume, unfilled circles indicate brush piles with decreased volume, and × represents where brush piles were lost. The size of circles represents relative difference in dimensional change (c) from 17 March 2017 to 18 December 2019 and (d) from 17 March 2017 to 9 November 2020. Tributaries are labeled as “A” and “B”, the long-dash line represents the shoreline, and the short-dash line represents the 75-m elevation contour.

Close modal

Spatial patterns

Brush piles were installed unequally across depth categories, but brush piles in all depth categories were lost (Table 1). We detected spatial autocorrelation in the logistic regression (Moran's I = 0.06; P = 0.02), so we fit an auto-logistic model with a weighted-sum auto-covariate. There was strong statistical support that brush piles were lost across number of flood cycles since installation (λLR = 12.65; P < 0.01), while depth category (λLR = 6.37; P = 0.09) and the interaction of the number of flood cycles and depth category (λLR = 7.24; P = 0.06) received only some statistical support. We detected spatial autocorrelation in our two-factor ANOVA (Moran's I = 0.17; P < 0.01), so we refit using an auto-normal model with a weighted-sum auto-covariate. Depth category (F = 2.59; P = 0.06) and number of flood cycles since installation (F = 2.69; P = 0.10) received some statistical support, but the interaction of the number of flood cycles since installation and depth category was not supported (F = 0.04; P = 0.99).

Spatial patterns of brush pile losses were statistically supported for all surveys (satellite: 10 November 2017 [one flood cycle] join-count statistic = 3.5 and P = 0.02, and 15 January 2019 [two flood cycles] join-count statistic = 5.5 and P = 0.02; onsite 18 December 2019 [three flood cycles] join-count statistic = 8.5 and P = 0.02, and 9 November 2020 [four flood cycles] join-count statistic = 19 and P = 0.03). In general, brush piles were lost and lost volume at higher rates closer to the main stem of Long Branch Creek and from higher elevation contours (Figure 4). The majority of the eight brush piles that were lost on 10 November 2017 (satellite survey [one flood cycle]) were from locations adjacent to Long Branch Creek (5/8; Figure 4a). The rest of the losses were from the upstream section of tributary A (3/8). Additional losses in the satellite survey on 15 January 2019 (two flood cycles after installation) and the onsite survey on 18 December 2019 were again adjacent to Long Branch Creek and the upstream section of tributary A (Figures 4b and 4c). Data from the 9 November 2020 onsite survey (four flood cycles) indicated that many of the additional losses were in the area between tributary A and tributary B, above the 75-m contour interval (Figure 4d). Patterns in dimensional change were present in both the 18 December 2019 (Moran's I = 0.21; P = 0.01) and 9 November 2020 (Moran's I = 0.20; P = 0.01) onsite surveys, indicating that percent change in volume was spatially correlated. Of note, most brush piles that increased in volume were adjacent to tributary A and closer to County Road 43 (10/12 on 18 December 2019 and 6/7 on 09 November 2020).

Management scenario

In our hypothetical management scenario, where N0 = 100 and 0 = 50.0 m3 for an initial total volume of 5,000 m3, we estimated Rt = 1,125 m3 = 100 × 50 m3 (1 − e−(0.10+0.15)×1). This translates to an annual replenishment of roughly 1 = 23 brush piles averaging 50 m3 (i.e., 1,125 m3/50 m3; equation 1) in the Long Branch Creek embayment. Replenishment options for our management scenario at longer intervals (t) are illustrated in Figure 5. Custom management options can be explored with equation (1) and using Code S1 and Table S1, Supplemental Material.

Figure 5.

Brush pile replenishment curve to maintain the initial or desired total volume of brush piles for a hypothetical management scenario (i.e., N0 = 100, 0 = 50 m3) in the Long Branch Creek embayment of Enid Lake, Mississippi, from 2017 to 2020. The solid line represents the average estimate, while dotted lines represent the upper and lower 95% confidence intervals. The numbers of brush piles that would need to be replenished at t flood cycles after installations 1, 2, 4, 8, and 16 are given in the inset table.

Figure 5.

Brush pile replenishment curve to maintain the initial or desired total volume of brush piles for a hypothetical management scenario (i.e., N0 = 100, 0 = 50 m3) in the Long Branch Creek embayment of Enid Lake, Mississippi, from 2017 to 2020. The solid line represents the average estimate, while dotted lines represent the upper and lower 95% confidence intervals. The numbers of brush piles that would need to be replenished at t flood cycles after installations 1, 2, 4, 8, and 16 are given in the inset table.

Close modal

Our results provide a quantified assessment of brush pile retention and dimensional change in a flood control reservoir and suggest that physical and, perhaps, anthropogenic factors (see below) result in brush piles alterations after installation. When contours of reservoirs are inundated, wave action presents a mechanical force that constantly breaks down twigs and branches of brush piles (Czarnecka 2016; Miranda 2017). Exposure to wind and atmospheric oxygen during drawdown further increases the rate of breakdown and decomposition (Bilby 2003; Spänhoff and Meyer 2004; Czarnecka 2016). Additionally, drawdown and precipitation events contribute to dynamic water levels and flow, likely resulting in dynamic rates of retention and variability in brush pile dimensional change (Czarnecka 2016). The hypotheses that brush piles are lost over time and lose volume over time was strongly supported by our results. Additionally, the hypotheses that the rates at which brush piles are lost and loss in brush pile volume exhibit spatial patterns were supported by our results.

Surprisingly, some brush piles in the Long Branch Creek embayment increased in volume. We presume this is likely due to wind, waves, and, potentially, other hydrologic forces and anthropogenic factors. In our study area, winds from the northwest create waves, typically 0.15–0.30 m in height, which are strong enough to push woody debris toward the shoreline (CERC 1984; Minar 2011; Czarnecka 2016). Because of the raised contour to the south of tributary A, some woody debris may become trapped in this location, ultimately adding volume to brush piles. Indeed, most brush piles that increased in volume in the Long Branch Creek embayment were in this area. Additionally, debris from other locations may have gotten caught in installed brush piles after being moved by hydrologic forces (e.g., runoff and eddy currents; Czarnecka 2016). Anthropogenic factors are also potential contributors to alterations in brush pile dimensional change and retention, as people relocate or add material to brush piles in the Long Creek Branch embayment near county road 43 (K. Meals, Mississippi Department of Wildlife, Fisheries, and Parks, personal communication).

The typical brush pile in our study, however, was either lost or decreased in volume over time, as expected. Previous authors noted that brush piles can be swept away by hydrological forces, slowly sink into the substrate, decay, and collapse, settle, and spread, causing an overall loss in supplemental brush pile habitat over time (Johnson and Lynch 1992; Bolding et al. 2004; Stone et al. 2012; Czarnecka 2016). Again, anthropogenic factors might influence retention and dimensional change of brush piles, as boating and fishing create wakes that can degrade brush piles and break twigs from snagging while angling and from contact from boat hulls and propellers. The loss of brush piles or brush pile volume was greater in some areas in the Long Branch Creek embayment than in others, likely due to unequal intensity, duration, or frequency of waves and high flow events. For example, all but one brush pile near Long Branch Creek, where hydrological forces are presumably more dynamic and intense, were either lost completely or had lost volume by 18 December 2020. Indeed, the hypothesis that brush pile retention and dimensional change express spatial patterns was supported by tests of autocorrelation and was somewhat supported by tests of brush pile loss and dimensional change in different depth categories or an interaction effect that included depth category.

The management scenario we present provides an example of how managers can apply our results to plan brush pile replenishment (see Figure 5). However, managers might monitor brush pile retention and dimensional change in their own waterbodies and use our analytical approaches to estimate rates of retention and dimensional change for their specific setting (see Code S1 and Table S1, Supplemental Material). In general, managers may consider that brush piles near hydrologically dynamic areas experience reduced retention, increased structural degradation, and diminished benefits to fish over time (Melillo et al. 1983; Miranda et al. 2010; Baumann et al. 2016; Czarnecka 2016; Hatcher et al. 2019). Given that brush piles near Long Branch Creek in our study lost numbers and volume at higher rates, managers may also consider 1) installing a greater proportion of brush piles in areas where they are expected to be more readily lost, 2) replenishing brush piles in these areas more frequently (i.e., subannually), 3) devoting a larger proportion of annual replenishment brush piles to such areas, or 4) avoiding such areas altogether. Strategies 1–3 would be more costly than strategy 4 but would provide supplemental habitat more uniformly over a given area. However, more research linking hydrodynamics to brush pile retention and dimensional changes is needed before any one strategy can be recommended. Furthermore, different arrangements, densities, and configurations of brush piles could influence overall rates of retention and dimensional change, and spatial patterns of retention and dimensional change (e.g., density dependence). As for now, estimation of Rt and t as described in our study serves as a simple yet easy-to-use way for managing total brush pile volume in a given area.

Log-linear regression models of ZN and ZV fit well (R2 = 0.93 and 0.99, respectively), and their respective validation Z estimates (i.e., ZNholdout and ZVholdout) fell well within each model's 95% CIs. Future studies might monitor brush pile retention with onsite surveys at shorter time intervals and compare models (hypotheses) of brush pile loss and changes in volume (e.g., constant decline, log-linear regression versus threshold decline, change-point regression). Also, while natural supplemental habitat is much less costly to install than their artificial counterparts ($1.60 per juniper tree versus$22.05 per plastic pipe attractor [Magnelia et al. 2008]; $209.05 [total cost] evergreen trees versus$344.30+ artificial habitats [Baumann et al. 2016]), comparisons of loss and degradation between natural and artificial supplemental habitat are needed to fully assess the cost per unit over time (but see Baumann et al. 2016). Additionally, our study did not quantify the change in interstitial space once needles and twigs are lost, although this was apparent when viewing photographs from 17 March 2017 and 18 December 2019 (Figure S1, Supplemental Material). Future studies might also quantify the rate of interstitial space growth and how such changes influence fish assemblages (Hatcher et al. 2019).

Although we found no instances of false negatives or false positives when comparing satellite imagery to onsite survey measurements, using satellite imagery to monitor brush pile retention needs further validation. Future studies might consider monitoring brush pile retention at more frequent intervals because it may provide the best opportunity to directly compare with satellite imagery. Further demonstrations of reliably monitoring brush pile retention using satellite imagery can lead to quicker, computer-assisted assessments that can be scaled to multiple study areas or waterbodies. Indeed, applications using remotely sensed data to detect and quantify woody debris volume have been made in forest plantation and lotic settings (Marcus et al. 2003; Ortega-Terol et al. 2014; Windrim et al. 2019; Queiroz et al. 2020) and can likely be extended to include flood control reservoirs and other lentic settings.

With over 600 large reservoirs (>100 ha) aiming to provide flood control and recreation or fish habitat in the contiguous United States (Rodgers 2017), loss and maintenance of structural complexity will remain a priority for fisheries managers. Brush piles will continue to be used as supplemental habitat for logistic and economic reasons (Magnelia et al. 2008; Allen et al. 2014), and managers may choose to avoid or replace supplemental habitat constructed of plastic if they are concerned that these contribute to microplastic pollution in aquatic systems (Eerkes-Medrano et al. 2015). However, brush piles degrade more quickly than their plastic counterparts (Baumann et al. 2016) and require replenishment to maintain, which may counter initial cost savings, but more research is needed. Simple measurements, such as the presence of brush piles, can potentially be monitored using satellite imagery in a geographic information system (e.g., Google Earth Pro) and can help managers make informed planning decisions to meet habitat objectives. Onsite surveys are more time consuming (3.5 min/brush pile versus 45 s/brush pile) but can provide additional information, such as volume, not yet available through satellite imagery (but see above). Additionally, onsite surveys lend themselves to involving local stakeholder groups and training them to collect the required information. Alternatively, managers may use the results of our study, in full or in part, as an approximation when planning brush pile replenishment to persistently meet fisheries habitat objectives.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Code S1. Brush piles are often used as fish habitat and attractors in areas void of structural complexity. Because they biodegrade, managers often need to replenish brush piles to meet management objectives. This R file contains code to estimate instantaneous brush pile loss (ZN), instantaneous loss in volume (ZV), the total volume of brush piles that must be replenished every t years (Rt) to maintain initial total volume, and the number of brush piles that must be replenished every t years (t) to maintain initial total volume. Data needed for estimates are embedded in the file and derived from measurements in the Long Creek Branch embayment of Enid Lake, Mississippi, from 2017 to 2020 (see Data S1, Supplemental Material).

Available: https://doi.org/10.3996/JFWM-21-033.S1 (12 KB R)

Data S1. Brush piles are often used as fish habitat and attractors in areas void of structural complexity. Because they biodegrade, managers often need to replenish brush piles to meet management objectives. This Excel file contains the ID, location, retention, and dimensional measurements of brush piles along with summary statistics of brush piles and wind data used in this study. Measurements were taken from the Long Creek Branch embayment of Enid Lake, Mississippi, from 2017 to 2020.

Available: https://doi.org/10.3996/JFWM-21-033.S2 (17.331 MB XLSX)

Methods S1. Brush piles are often used as fish habitat and attractors in areas void of structural complexity. Because they biodegrade, managers often need to replenish brush piles to meet management objectives. This Word document contains the steps to summarize prevalent wind direction and speed using data retrieved from weather stations nearby the study site (Long Branch Creek embayment of Enid Lake, Mississippi) during the study period (2017–2020); measure fetch across Enid Lake to the study site using Google Maps, and calculate the typical range of wave height at the study site using CERC (1984).

Available: https://doi.org/10.3996/JFWM-21-033.S3 (13 KB DOCX)

Table S1. Brush piles are often used as fish habitat and attractors in areas void of structural complexity. Because they biodegrade, managers often need to replenish brush piles to meet management objectives. This Excel file allows users to input their own brush pile counts and average brush pile volume to estimate the total volume of brush piles Rt and the total number of brush piles t that must be replenished annually to maintain the initial total volume. This is limited to counts and volumes at two time points (see Code S1, Supplemental Material, for >1 time point). Default values are calculated from measurements taken from the Long Creek Branch embayment of Enid Lake, Mississippi, from 2017 to 2020 (see Data S1, Supplemental Material).

Available: https://doi.org/10.3996/JFWM-21-033.S4 (11 KB XLSX)

Figure S1. Brush piles are often used as fish habitat and attractors in areas void of structural complexity. Because they biodegrade, managers often need to replenish brush piles to meet management objectives. Photographs in this figure show brush piles on 17 March 2017 and 18 December 2019 from the Long Branch Creek embayment of Enid Reservoir, Mississippi. Visual comparison of photos shows the loss of foliage, twigs, and small branches, leading to decreased interstitial space.

Available: https://doi.org/10.3996/JFWM-21-033.S5 (58.762 MB TIF)

Reference S1.[CERC] Coastal Engineering Research Center. 1984. Shoreline protection manual volume I. 4th edition. Washington, D.C.: Waterways Experiment Station, U.S. Army Corps of Engineers.

Available: https://doi.org/10.3996/JFWM-21-033.S6 (11.507 MB PDF)

Reference S2. Finkbeiner M, Stevenson B, Seaman R. 2001. Guidance for benthic habitat mapping: an aerial photographic approach. Charleston, South Carolina: U.S. NOAA Coastal Services Center. NOAA/CSC/20117-PUB.

Available: https://doi.org/10.3996/JFWM-21-033.S7 (8.129 MB PDF)

Reference S3. Mabbott LB. 1991. Artificial habitat for warmwater fish in two reservoirs in southern Idaho. Pages 62–65 in Cooper JL, Hamre RH, technical coordinators. Warmwater fisheries symposium I. Fort Collins, Colorado: USDA Rocky Mountain Forest and Range Experiment Station. General Technical Report RM-207.

Available: https://doi.org/10.3996/JFWM-21-033.S8 (337 KB PDF)

Reference S4. Rodgers KD. 2017. A reservoir morphology database for the conterminous United States. Reston, Virginia: U.S. Geological Survey. Data Series 1062.

Available: https://doi.org/10.3996/JFWM-21-033.S9 (167 KB PDF) and https://pubs.usgs.gov/ds/1062/

Funding was provided by the Mississippi Department of Wildlife, Fisheries and Parks and Reservoir Fisheries Habitat Partnership. We thank Andy Turner and Jeremy Shiflet for their helpful comments on an earlier version of this manuscript and the reviewers and editors for their time and effort in helping us refine this manuscript to its current form. This publication is a contribution of the Forest and Wildlife Research Center at Mississippi State University.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Allen
MJ,
Bush
SC,
Vining
I,
Siepker
MJ.
2014
.
Black bass and crappie use of installed habitat structures in Table Rock Lake, Missouri
.
North American Journal of Fisheries Management
34
:
223
231
.
Bardos
DC,
Guillera-Arroita
G,
Wintle
BA.
2015
.
Valid auto-models for spatially autocorrelated occupancy and abundance data
.
Methods in Ecology and Evolution
6
:
1137
1149
.
Barwick
DH.
2004
.
Species richness and centrarchid abundance in littoral habitats of three southern U.S. reservoirs
.
North American Journal of Fisheries Management
24
:
76
81
.
Baumann
JR,
Oakley
NC,
McRae
BJ.
2016
.
Evaluating the effectiveness of artificial fish habitat designs in turbid reservoirs using sonar imagery
.
North American Journal of Fisheries Management
36
:
1437
1444
.
Becker
RA,
Wilks
AR,
Brownrigg
R,
Minka
TP,
Deckmyn
A.
2018
.
maps: draw geographical maps. R package version 3.3.0
.
Available: https://CRAN.R-project.org/package=maps (December 2019)
Bilby
RE.
2003
.
Decomposition and nutrient dynamics of wood in streams and rivers
.
Pages
135
147
in
SV,
Gregory
Boyer
KL,
Gurnell
AM,
editors.
The ecology and management of wood in world rivers
.
Bethesda, Maryland: American Fisheries Society. Symposium 37.
Bivand
RS,
Pebesma
E,
Gomez-Rubio
V.
2013
.
Applied spatial data analysis with R. 2nd edition
.
New York
:
Springer
.
Bolding
B,
Bonar
S,
Divens
M.
2004
.
Use of artificial structure to enhance angler benefits in lakes, ponds, and reservoirs: a literature review
.
Reviews in Fisheries Science
12
:
75
96
.
Carmignani
JR,
Roy
AH.
2017
.
Ecological impacts of winter water level drawdowns on lake littoral zones: a review
.
Aquatic Sciences
79
:
803
824
.
[CERC] Coastal Engineering Research Center.
1984
.
Shoreline protection manual volume I. 4th edition
.
Washington, D.C
.:
Waterways Experiment Station, U.S. Army Corps of Engineers (see Supplemental Material, Reference S1)
Czarnecka
M.
2016
.
Coarse woody debris in temperate littoral zones: implications for biodiversity, food webs and lake management
.
Hydrobiologia
767
:
13
25
.
DeBoom
CS,
Wahl
DH.
2013
.
Effects of coarse woody habitat complexity on predator–prey interactions of four freshwater fish species
.
Transactions of the American Fisheries Society
142
:
1602
1614
.
Eerkes-Medrano
D,
Thompson
RC,
Aldridge
DC.
2015
.
Microplastics in freshwater systems: a review of the emerging threats, identification of knowledge gaps and prioritization of research needs
.
Water Research
75
:
63
82
.
Finkbeiner
M,
Stevenson
B,
Seaman
R.
2001
.
Guidance for benthic habitat mapping: an aerial photographic approach. Charleston, South Carolina: U.S. NOAA Coastal Services Center. NOAA/CSC/20117-PUB
(see Supplemental Material, Reference S2).
Grimm
AG,
Brooks
CN,
Binder
TR,
Riley
SC,
Farha
SA,
Shuchman
RA,
Krueger
CC.
2016
.
Identification of lake trout Salvelinus namaycush spawning habitat in northern Lake Huron using high-resolution satellite imagery
.
Journal of Great Lakes Research
42
:
127
135
.
Hatcher
HR.
2018
.
Establishing and evaluating agricultural plantings and supplemental cover on reservoir mudflats as a means to increase juvenile game fish abundance and growth. Master's thesis
.
Mississippi State, Mississippi
:
Mississippi State University
.
Hatcher
HR,
Miranda
LE,
Colvin
ME,
Coppola
G,
Lashley
MA.
2019
.
Fish assemblages in a Mississippi reservoir mudflat with low structural complexity
.
Hydrobiologia
841
:
163
175
.
Johnson
DL,
Lynch
WE
Jr.
1992
.
Panfish use of and angler success at evergreen tree, brush, and stake-bed structures
.
North American Journal of Fisheries Management
12
:
222
229
.
Kahle
D,
Wickham
H.
2013
.
ggmap: spatial visualization with ggplot2
.
The R Journal
5
:
144
161
.
Kaufmann
PR,
Peck
DV,
Paulsen
SG,
Seeliger
CW,
Hughes
RM,
Whittier
TR,
Kamman
NC.
2014
.
Lakeshore and littoral physical habitat structure in a national lakes assessment
.
Lake and Reservoir Management
30
:
192
215
.
Krogman
RM,
Miranda
LE.
2016
.
Rating of U.S. reservoirs relative to fish habitat condition
.
Lake and Reservoir Management
32
:
51
60
.
Mabbott
LB.
1991
.
Artificial habitat for warmwater fish in two reservoirs in southern Idaho
.
Pages
62
65
in
JL,
Cooper
Hamre
RH,
technical coordinators.
Warmwater Fisheries Symposium I
.
Fort Collins, Colorado: USDA Rocky Mountain Forest and Range Experiment Station. General Technical Report RM-207
(see Supplemental Material, Reference S3).
Magnelia
SJ,
De Jesus
MJ,
Schlechte
JW,
Cummings
GC,
Duty
JL.
2008
.
Comparison of plastic pipe and juniper tree fish attractors in a Central Texas reservoir
.
Proceedings of the Annual Conference of the Southeastern Association of Fish and Wildlife Agencies
62
:
183
188
.
Marcus
WA,
Legleiter
CJ,
Aspinall
RJ,
Boardman
JW,
Crabtree
RL.
2003
.
High spatial resolution hyperspectral mapping of in-stream habitats, depths, and woody debris in mountain streams
.
Geomorphology
55
:
363
380
.
Melillo
JM,
Naiman
RJ,
Aber
JD,
Eshleman
KN.
1983
.
The influence of substrate quality and stream size on wood decomposition dynamics
.
Oecologia
58
:
281
285
.
Minar
N.
2011
.
Wind history map. Daedalus Bits, LLC
.
Available: http://windhistory.com/ (August 2021)
Miranda
LE,
Bettoli
PW.
2007
.
Mortality
.
Pages
229
277
in
CS,
Guy
Brown
ML,
editors.
Analysis and interpretation of freshwater fisheries data
.
Bethesda, Maryland
:
American Fisheries Society
.
Miranda
LE,
Spickard
M,
Dunn
T,
Webb
KM,
Aycock
JN,
Hunt
K.
2010
.
Fish habitat degradation in U.S. reservoirs
.
Fisheries
35
:
175
184
.
Miranda
LE,
Krogman
RM.
2015
.
Functional age as an indicator of reservoir senescence
.
Fisheries
40
:
170
176
.
Miranda
LE.
2017
.
Reservoir fish habitat management
.
Totowa, New Jersey
:
Lightning Press
.
Molthan
A,
Jedlovec
G.
2011
.
NASA satellite data assist in tornado damage assessments
.
Eos
92
:
337
339
.
Ortega-Terol
D,
Moreno
MA,
Hernández-Lopez
D,
Rodríguez-Gonzálvez
P.
2014
.
Survey and classification of large woody debris (LWD) in streams using generated low-cost geomatic products
.
Remote Sensing
6
:
11770
11790
.
Plant
RE.
2012
.
Spatial data analysis in ecology and agriculture using R
.
Boca Raton, Florida
:
CRC Press/Taylor & Francis Group
.
Queiroz
GL,
McDermid
GJ,
J,
Hopkinson
C,
Kariyeva
J.
2020
.
Estimating coarse woody debris volume using image analysis and multispectral LiDAR
.
Forests
11
:
141
.
R Core Team.
2019
.
R: a language and environment for statistical computing
.
Vienna
:
R Foundation for Statistical Computing
.
Rodgers
KD.
2017
.
A reservoir morphology database for the conterminous United States
.
Reston, Virginia
:
U.S. Geological
Survey. Data Series 1062 (see Supplemental Material, Reference S4).
Spänhoff
B,
Meyer
EI.
2004
.
Breakdown rates of wood in streams
.
Journal of the North American Benthological Society
23
:
189
197
.
Stone
N,
Morris
JE,
Smith
B.
2012
.
Managing the pond environment
.
Pages
113
143
in
JW,
Neal
Willis
DW,
editors.
Small impoundment management in North America
.
Bethesda, Maryland
:
American Fisheries Society
.
Tugend
KI,
Allen
MS,
Webb
M.
2002
.
Use of artificial habitat structures in U.S. lakes and reservoirs: a survey from the Southern Division AFS Reservoir Committee
.
Fisheries
27
(5)
:
22
27
.
Ward
DP,
Petty
A,
Setterfield
SA,
Douglas
MM,
Ferdinands
K,
Hamilton
SK,
Phinn
S.
2014
.
Floodplain inundation and vegetation dynamics in the Alligator Rivers region (Kakadu) of northern Australia assessed using optical and radar remote sensing
.
Remote Sensing of Environment
147
:
43
55
.
Willis
TC,
Bremigan
MT,
Hayes
DB.
2004
.
Variable effects of habitat enhancement structures across species and habitats in Michigan reservoirs
.
Transactions of the American Fisheries Society
133
:
398
410
.
Windrim
L,
Bryson
M,
McLean
M,
Randle
J,
Stone
C.
2016
.
Automated mapping of woody debris over harvested forest plantations using UAVs, high-resolution imagery, and machine learning
.
Remote Sensing
11
:
733
.

The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.

Author notes

Citation: Aldridge CA, Norris DM, Hatcher HR, Coppola G, Colvin ME, Miranda LE. 2022. Retention and dimensional changes of evergreen brush piles within a flood control reservoir. Journal of Fish and Wildlife Management 13(1):223–235; e1944-687X. https://doi.org/10.3996/JFWM-21-033