Long-Term Effects of Large Woody Debris Addition on Stream Habitat and Brook Trout Populations Free

The abundance of stream-dwelling salmonids is often correlated with the abundance of large woody debris (LWD [Fausch and Northcote 1992; Berg et al. 1998; Neumann and Wildman 2002]). However, manipulative studies designed to test such correlations have often shown mixed results (Solazzi et al. 2000; Roni and Quinn 2001; Warren and Kraft 2003; Sweka and Hartman 2006). The lack of a significant influence of LWD manipulations on salmonid populations may be due to pseudo-replicated studies (e.g., Cederholm et al. 1997), relatively small spatial scales of habitat manipulation, or postmanipulation monitoring that is not long enough to detect a population change (Hunt 1976, 1988; Sweka and Hartman 2006).

In a previous study, Sweka and Hartman (2006) assessed stream habitat and brook trout Salvelinus fontinalis populations for 1 y prior to, and 3 y after, habitat manipulation and found no significant changes following the addition of LWD. Brook trout populations fluctuated from year to year, but there was no net increase in density. In this paper, we compare data collected in 2006 (6 y postmanipulation) to that collected in 2003 (3 y postmanipulation), the last year of our previous study, to determine long-term changes to stream habitat and brook trout populations due to LWD additions. Monitoring of habitat manipulations on this temporal scale is rare in the literature (except see Gowan and Fausch 1996a), but is needed to evaluate the effectiveness of habitat manipulations in achieving desired habitat and fish population responses.

We added LWD to eight tributaries of the Middle Fork River, Randolph County, West Virginia (see Sweka and Hartman 2006 for complete details). The streams were first- and second-order with temperatures rarely exceeding 20°C with brook trout as the dominant fish species. We divided the study into two experiments: a within-stream experiment and an among-stream experiment. In the within-stream experiment, four streams had LWD added to a single 300-m stream reach with additional 300-m reaches, upstream and downstream of the manipulated reach, serving as controls. Treatment reaches were separated from control reaches by 100 m. We compared stream habitat and brook trout densities between controls and manipulated reaches. The among-stream experiment compared habitat and brook trout densities from the four streams used in the within-stream experiment to four additional streams where LWD was added to three 300-m reaches (separated by 100 m). The purpose of the among-stream experiment was to determine the effect of the scale of habitat manipulation upon stream habitat and brook trout density.

A logging crew added LWD by felling trees into the streambed at a rate of approximately 15 trees/300-m stream length. Each added piece of LWD was individually tagged to assess movement and functionality of individual pieces between years. The felled trees had a minimum diameter of 10 cm and median diameter of 22 cm (Sweka and Hartman 2006). We assessed stream habitat once prior to the addition of LWD (summer 2000) and three times following LWD addition (summers 2001–2003). Brook trout populations were assessed in the autumn and spring of each year. Pretreatment brook trout population assessment occurred in the autumn 1999 and spring 2000, and posttreatment brook trout population assessment occurred from autumn 2000 through spring 2003.

We used the same methodology in 2006 for assessing stream habitat and brook trout populations as in our previous study (Sweka and Hartman 2006). Briefly, habitat was surveyed under summer base-flow conditions and individual habitat units were classified as riffles, runs, or pools. We measured the length of each habitat unit along the thalweg, and wetted and bank-full widths were visually estimated at transects of 0.25, 0.50, and 0.75 of the thalweg length. The widths of 20% of the habitat units were estimated and then verified by measuring to the nearest 0.1 m. The area of each habitat unit was calculated as the mean width multiplied by the length. Verified area was regressed on estimated area and predicted values of the regression equation were used for the area of estimated habitat units.

We considered LWD to be any piece of wood within the bank-full channel with a minimum diameter of 10 cm and minimum length of 1 m. Pieces were classified according to the size criteria proposed by Flebbe and Dolloff (1995) and according to their function within the stream channel (pool formation, sediment storage, organic storage, overhead cover, bank stabilization). The location of each added piece of LWD was determined by measuring the distance of each piece in relation to the upstream or downstream end of each 300-m study reach. Movement of added pieces of LWD was estimated as the difference in locations between years. We also classified the function of each added piece in the same manner as naturally occurring LWD.

Brook trout populations were estimated by multiple-pass removal electrofishing in a randomly chosen 100-m section of each 300-m reach (both control and treatment reaches). Electrofishing occurred in the autumn (October–November) of 2006. Population estimates were generated using the program CAPTURE (White et al. 1982) and density was estimated as the population estimate divided by the area of the electrofishing site.

We compared our 2006 habitat results to those from 2003 to determine whether any changes had occurred since our previous study (Sweka and Hartman 2006). Brook trout density data were compared to those found in the autumn 2002, because our previous study ended in the spring 2003. Total number of pieces of LWD, total number of pools, total pool area, and brook trout density were analyzed as a repeated-measures analysis of variance using the PROC MIXED procedure in SAS® version 9.1. We were specifically interested in the treatment × time interaction in these tests to determine whether any changes had occurred between treatments since 2003. The function of the added LWD was compared between 2003 and 2006 using a chi-square test.

The majority of the added LWD remained within the study areas and continued to serve some function by 2006 (Supplemental Material, Table S1, http://dx.doi.org/10.3996/012010-JFWM-002.S1). Three hundred fifty-four pieces of LWD were added to all the streams in 2000 (Sweka and Hartman 2006). By 2006, 304 (86%) were found, although 107 (30%) pieces moved > 5 m from their original location in 2000. Of the added pieces, 11% were involved in pool formation, 38% stored organic material, 7% stored sediment, 9% provided overhead cover for brook trout, 2% were involved in bank stabilization, and 33% did not serve any apparent function under summer base-flow conditions. The frequencies of these functions were not significantly different from those in 2003 (χ²  =  5.84, df  =  5, P  =  0.32). The frequency of the added LWD involved in pool formation continued to be greater than that of naturally occurring LWD (11% versus 2%), which is likely due to the large size of the added pieces compared to naturally occurring pieces (Sweka and Hartman 2006).

Although the majority of added LWD was retained, stream habitat did not significantly change between 2003 and 2006 in the within-stream experiment. Treatment reaches in the within-stream experiment continued to have an overall greater amount of total LWD (natural + added pieces [Figure 1a; F_1,7  =  19.08, p < 0.01]) than control reaches, but there was no significant interaction with time (F_1,10  =  0.63, P  =  0.44). Reaches where LWD was added tended to have a greater number of pools (Figure 2a; F_1,7  =  4.64, P  =  0.07) and pool area/300 m (Figure 2b; F_1,7  =  5.09, P  =  0.06) compared to control reaches, but these differences were not significant, and there was no time-dependent treatment effect for the number of pools/300 m (F_1,10  =  2.72, P  =  0.13) or pool area/300 m (F_1,10  =  0.15, P  =  0.71).

Stream habitat also did not significantly change in the among-stream experiment. Streams that had LWD added to three 300-m reaches did not have significantly more total LWD (total LWD  =  natural LWD + added LWD) than those where LWD was added to a single 300-m reach (Figure 3a; F  =  1.86, P  =  0.22). However, when considering the largest size classes of LWD (diameter > 10 cm and length > 5 m; Flebbe and Dolloff 1995), the streams where LWD was added to three 300-m reaches continued to have significantly more of this large size class of LWD than streams where LWD was added to only a single 300-m reach (Figure 3b; F_1,6  =  7.93, P  =  0.03). There was also no differential change between treatments through time in the among-stream experiment for total LWD (F_1,6  =  0.25, P  =  0.63) or the largest size classes of LWD (F_1,6  =  0.55, P  =  0.49). In the among-stream experiment, pool number tended to decrease between 2003 and 2006, but this was not significant (Figure 4a; F_1,6  =  4.44, P  =  0.08), and there was no overall effect of the scale of habitat manipulation (F_1,6  =  0.14, P  =  0.72) or a differing effect of scale with time (F_1,6  =  3.60, P  =  0.11). Pool area also decreased significantly between 2003 and 2006 in all streams (Figure 4b; F_1,6  =  12.82, P  =  0.01), but there was no overall effect of the scale of habitat manipulation (F_1,6  =  0.11, P  =  0.75) or a differing effect of scale with time (F_1,6  =  3.30, P  =  0.12).

The added LWD continued to have no effect on age 1+ brook trout density by 2006. In the within-stream experiment, age 1+ brook trout density showed an overall decrease between 2002 and 2006 (Figure 5; F_1,10  =  13.35, P < 0.01), but there was no overall difference between reaches where LWD was added and control reaches (F_1,7  =  4.50, P  =  0.07) and no differential change between treatments with time (F_1,10  =  0.08, P  =  0.78). Likewise, age 1+ brook trout density decreased between 2002 and 2006 in the among-stream experiment (Figure 6; F_1,6  =  9.26, P  =  0.02), but there was no overall effect of the scale of habitat manipulation (F_1,6  =  1.07, P  =  0.34) or a differing effect of scale with time (F_1,6  =  0.14, P  =  0.72). The overall mean age 1+ brook trout density in all eight streams in 2006 was 2.96 ± 1.38 fish/100 m² (± 95% CI) while it was 5.85 ± 3.83 fish/100 m² in 1999 when our previous study began (Sweka and Hartman 2006). Thus, the addition of LWD did not increase brook trout populations in these streams.

Long-term monitoring of trout population response to habitat manipulation is rare in the literature. Hunt (1976, 1988) suggested that population response may take greater than 5 y postmanipulation and we also partly attributed the lack of a sustained increase in brook trout density in our original study (Sweka and Hartman 2006) to the lack of time since the LWD was added. Even with monitoring of populations for 6 y following habitat manipulation, we still did not observe an increase in brook trout density. Brook trout densities were actually lower in 2006 than in 2002. Pool habitat in the among-stream experiment decreased in both scales of habitat manipulation from 2003 to 2006, which may be one potential explanation for decreasing brook trout densities in the among-stream experiment. However, pool area remained the same in the treatment and control reaches in the within-stream experiment and yet brook trout densities still declined. Thus, the cause of the decrease in brook trout density from 2002 to 2006 appears independent of the LWD manipulations and resulting effects on stream physical habitat. Brook trout populations of the Appalachian region commonly show much inter-annual variability (Petty et al. 2005; Sweka and Hartman 2006), making it difficult to differentiate variation due to habitat manipulation from variation due to other factors.

The importance of large woody debris to structuring stream habitat remains a strong tenet of stream restoration, although manipulative studies have had difficulty showing a population-level effect. Riley and Fausch (1995) found trout abundance increased after addition of log structures in Colorado streams, but Gowan and Fausch (1996a) attributed this increase to immigration from outside their study reaches rather than an increase in recruitment and survival. Likewise, Warren and Kraft (2003) did not observe a decrease in brook trout relative abundance following the removal of LWD from Adirondack mountain streams.

The underlying assumption in manipulative studies with LWD, and in stream restoration projects utilizing LWD, is that LWD is a limiting factor for stream habitat complexity and stream fish populations. Perhaps the significant correlations seen between the abundance of LWD and salmonid populations (Fausch and Northcote 1992; Fleebe and Dolloff 1995; Neumann and Wildman 2002) are an artifact of overall past land use within a watershed. For example, disturbed watersheds may have had instream structure removed, increased sedimentation, decreased habitat connectivity, removal of riparian vegetation, and degraded water chemistry. Conversely, in undisturbed watersheds all these factors may still be in pristine condition, with accompanying high abundances of LWD resulting in positive correlations between salmonid abundance and LWD abundance. Although LWD can increase habitat complexity, the underlying geology, stream-channel type (i.e., Rosgen Channel types), and abundance of boulders may have more of an influence on stream habitat than LWD in relatively high-gradient streams (Warrren and Kraft 2003). Also, stream-dwelling salmonids move much more than is typically acknowledged by resource managers (Gowan et al. 1994; Gowan and Fausch 1996b; Logan 2003) and utilize differing habitats seasonally (Petty et al. 2005). Thus, it is not surprising that manipulative studies that modify stream habitat over a relatively small spatial scale (i.e., a few hundred meters of stream length) often fail to show a population-level effect. Successful habitat restoration needs to consider watershed-scale processes and the mobility of target species, such as brook trout (Logan 2003; Petty et al. 2005), in order to result in population-level effects on stream fish.

Although our LWD additions failed to significantly modify stream habitat or increase brook trout populations, our results provide valuable information to others considering such habitat manipulations. For example, the streams of this study had an average gradient of > 3% (Sweka and Hartman 2006) with abundant boulders. In such systems, stream habitat complexity may be governed more by boulders than by LWD (Warren and Kraft 2003). Also, the significant correlations seen in the literature between salmonid abundance and LWD do not necessarily mean that LWD is a limiting factor for species such as brook trout and increasing LWD does not always increase fish abundance.

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.

Table S1. Functions of added LWD in tributaries of Middle Fork River, Randolph County, West Virginia in 2003 and 2006.

Found at DOI: 10.3996/012010-JFWM-002.S1 (9.6 KB XLSX).

Table S2. Amounts of LWD in tributaries of Middle Fork River, Randolph County, West Virginia in 2003 and 2006.

Found at DOI: 10.3996/012010-JFWM-002.S2 (11.7 KB XLSX).

Table S3. Number of pools and area of pool habitats in tributaries of Middle Fork River, Randolph County, West Virginia in 2003 and 2006.

Found at DOI: 10.3996/012010-JFWM-002.S3 (9.96 KB XLSX).

Table S4. Density of brook trout age 1+ in tributaries of Middle Fork River, Randolph County, West Virginia in autumn 2002 and 2006.

Found at DOI: 10.3996/012010-JFWM-002.S4 (8.94 KB XLSX).

We thank the U.S. Forest Service Monongahela National Forest and Northern Research Station, the MeadWestvaco Corporation, West Virginia Division of Natural Resources, and West Virginia University Division of Forestry for funding this research.

We also thank those individuals who aided in data collection during our previous study and J. Studinski, J. Howell, and J. Stolarski for aiding in data collection in 2006. This manuscript was improved with preliminary reviews by J. Mohler, M. Millard, and three anonymous reviewers.