Skykomish River Braided Reach Restoration Assessment Fish Use

Skykomish River Braided Reach Restoration Assessment

Fish Use Analysis

Draft Final Report

June 28, 2006

Suggested citation: Drucker, E.G. 2006. Skykomish River Braided Reach Restoration Assessment: Fish Use Analysis. Draft Final Report, June 28, 2006, prepared by Washington Trout for Snohomish County Surface Water Management, Everett, WA.

Introduction Snohomish County Surface Water Management and partners have proposed to identify and prioritize opportunities to restore reach-level channel processes within the braided section of the Skykomish River extending from the cities of Gold Bar to Sultan, WA. The braided reach (“study reach”) includes over ten miles of mainstem channel between “Big Eddy” in Gold Bar (rm 43.3) and Sultan (rm 33), and nearly seven miles of interconnected side channels (Figs 1–7). In 2004 Washington Trout (WT) surveyed the braided reach in support of the following project goals:

• to describe patterns of current fish use within the study reach to define a baseline against which future conditions can be compared; • to identify critical fish habitats so that restoration actions with high likelihood of creating and maintaining these habitats can be prioritized; • to evaluate the biological cost and benefit of proposed restoration actions.

Washington Trout’s 2004 field work served two general tasks designed to evaluate fish use within the braided reach. First, fish species composition and relative abundance were surveyed seasonally within a subsample of habitat units chosen to represent the overall habitat-type distribution in the study reach. The specific objectives of these surveys were (i) to compare species diversity and relative density of adult and juvenile fishes in the study reach’s mainstem and in associated side channel networks; (ii) to identify physical habitat features significantly related to fish relative density, including habitat unit type and size, and within-unit survey location (e.g. channel center versus edge habitat; unit middle versus upstream–downstream unit interface; Fig. 8); and (iii) to examine seasonal and diel variation in species composition, distribution, and relative abundance.

The second task related to fish-use assessment was characterization of salmon spawning activity in the study reach. Redd and carcass surveys were performed in order to (i) quantify the incidence and timing of spawning activity throughout the braided reach; (ii) to document species-specific patterns in the spatial distribution of spawning; and (iii) to examine the relationship between spawning activity and physical habitat characteristics.

Patterns of fish use in the Skykomish River braided reach provided guidance for developing both a general strategy and specific projects for restoring and protecting habitat. Recommendations for designing and prioritizing such projects are presented in Chapter X.

Methods Reach Nomenclature The braided reach was partitioned into distinct habitat units (pools, riffles, glides) according to the Snohomish County Preliminary Habitat Analysis of July 2004 (Figs 1–7). To facilitate comparison among units, unique identifier numbers were assigned to each mainstem and side channel habitat unit surveyed. Side channel networks were named from upstream to downstream as follows: SCA (right-bank channel near the city of Gold Bar) (Fig. 3), SCB (left-bank network near Gold Bar) (Fig. 3), SCC (left-bank network between the cities of Startup and Sultan) (Figs 5, 7), SCD (right-bank network near Sultan) (Fig. 6).

Preliminary habitat-type designations, based on inspection of channel morphology in aerial photographs taken during summer low-flow conditions, were ground-truthed by Washington

Trout in August and September 2004. During these summer surveys, WT crews qualitatively assessed each mapped unit’s physical characteristics including water depth, velocity, and surface turbulence. These observations were measured against the geomorphic definitions of each unit type following Cramer (2001):

Pool: a unit with residual depth and no surface turbulence except at its inflow. Glide: a unit with no residual depth, little surface turbulence, and relatively uniform flow velocity. Riffle: a unit with discernable gradient and surface turbulence.

Several of the preliminary habitat type designations in the mainstem channel were modified on the basis of WT’s on-site surveys. When correction of the original habitat unit type resulted in two consecutive units of the same type, the two units were consolidated into a single larger unit and renamed accordingly (e.g. mainstem unit “70+71”). In side channels, where most mapped habitat units were pools, WT crews identified and named a number of additional units not appearing on the preliminary maps. Note that the fish use surveys, which relied on these corrected preliminary maps, preceded the final habitat unit delineation based on quantitative, on- site physical surveys (cf. Appendix X). Fish survey areas and final habitat unit boundaries were largely congruent; in cases where areal discrepancies were large, associated fish use data were excluded from analysis.

Species Composition and Relative Abundance Snorkel surveys were performed seasonally (Table 1) to provide estimates of fish species diversity and relative abundance within the braided reach. Survey teams were comprised of three snorkelers who floated abreast in a downstream direction during daylight hours (summer, fall and winter 2004), and at night (winter 2004), recording in notebooks the species, life stage (young-of-year and yearling juvenile versus adult) and number of all fishes observed within each habitat unit. Waterproof habitat unit maps and aerial photographs (Figs 2–6) were used to assist in identifying habitat unit boundaries in the field.

Surveyors occupied three longitudinal snorkel lanes, one located in the center of the channel and one at each channel edge (Fig. 8) (cf. Fig. 3 in Pess et al., 2002; Cramer, 2001 p. 42). Snorkelers changed lanes frequently between habitat units to minimize potential spatial bias in fish counts. The width of each snorkel lane was defined by visibility on the day of the survey. Although adult fish could be discerned underwater at distances up to 25 feet in the summer and up to 15 feet in the fall (Table 1), the maximum distance at which the smallest juvenile fish could be confidently identified was approximately 5 feet (cf. Hillman and Chapman, 1989). Each snorkel lane, therefore, was conservatively estimated as 10 feet wide. Within each habitat unit polygon (Figs 2–6) the ten-foot wide curvilinear corridor along the unit’s longitudinal midline was used to calculate snorkel lane area. To approximate relative fish densities for each habitat unit, total fish counts were divided by the estimated total wetted area surveyed (i.e. three-times snorkel lane area). As emphasized previously (Cramer, 2001), fish relative density calculated from a large-river census of this type must be considered representative only of the surveyed fraction of the total habitat area.

To examine patterns of within-unit spatial variation in the relative density of fish species observed, each mainstem habitat unit was partitioned into nine survey subareas. The three longitudinal snorkel lanes defined left-bank, center, and right-bank areas; units were further

2 partitioned into upper, middle and lower thirds (Fig. 8). The latter divisions were approximated in the field by consulting aerial photographs. The area of channel edge habitat surveyed within a unit (AE) was taken as the sum of the left-bank and right-bank subareas. Habitat unit interface area (AI) was calculated as the lower one-third of a given unit’s survey area plus the upper one- third of the adjacent downstream unit’s survey area. Relative densities (count m-2) then were calculated for (i) upper, middle and lower survey areas within units by dividing corresponding fish counts by one-third of the total unit survey area; (ii) edge habitat areas within units by dividing the sum of left- and right-bank fish counts by AE (i.e. two-thirds of the total unit survey area); and (iii) habitat unit interface areas using fish counts within AI. This sampling approach allowed testing of the hypothesis that relative fish density in habitat unit subareas with relatively great physical heterogeneity (channel edge and habitat unit interface) exceeds that in more homogeneous habitats (channel center and middle) (cf. Fig. 8).

For both mainstem and side channel survey areas, fish densities were calculated individually for juveniles and adults of each species observed, and for five species categories: (1) all fish species; (2) all salmonids (juvenile and adults); (3) adult salmonids; (4) juvenile salmonids; and (5) ESA- listed salmonids (Chinook salmon and bull trout, adults and juveniles). For salmonid categories 2–4, mountain whitefish (Prosopium williamsoni) were excluded on the basis of the species’ relatively high abundance and unthreatened status (Snohomish River Basin Salmon Recovery Forum, 2005).

After an initial reconnaissance snorkel of the mainstem in late July 2004, a baseline of fish use data for summer 2004 was obtained over the course of five days (early August to early September) by snorkeling every mapped mainstem habitat unit between rm 34.5 and rm 43.3 (units 1–74 consecutively; Figs 2–6) as well as the majority of mapped units within side channel networks SCB (22 units) and SCC (30 units) (Tables 1, 3). In the subsequent fall 2004 survey, mainstem habitat units were sampled by a stratified (proportional allocation) subsampling protocol (EPA, 2002). Specifically, habitat units of different types were selected randomly for survey in proportion to the frequency of their occurrence within two contiguous segments of the study reach with distinct geomorphological profiles (cf. Appendix X): “Big Eddy” in Gold Bar to Startup and Startup to the mouth of the Sultan River (Table 2). The mainstem was subsampled using this protocol on a single day in fall 2004 (mid November). Selected units in side channel network SCC were surveyed in winter 2004 (late December) to allow seasonal comparisons with summer fish use data from this network. Preliminary observations on diel variation in fish distribution and relative abundance were also made during this winter survey (cf. Table 1). WT crews attempted to snorkel side channel network SCD and the adjacent mainstem reach (i.e. below the mouth of the Sultan River; Skykomish rm 33–34.5) in December 2004 but extremely poor visibility in both areas precluded the collection of fish use data. Snorkel sites and habitat unit sample sizes for all 2004 side channel surveys are listed in Table 3.

Spawning Surveys Washington Trout coordinated with Washington Department of Fish and Wildlife personnel to identify the range of dates that typically encompasses the spawning runs of Chinook, chum and coho salmon in the Skykomish River braided reach. On four days within this range during fall 2004 (between September and December; Table 1), WT field crews conducted surveys throughout the mainstem channel and side channel networks SCA, SCB, SCC and SCD (rm 33– 43.3) to record evidence of salmon spawning activity. Surveys were performed by boat and on

3 foot; polarized sunglasses were worn to facilitate observation of salmon redds and carcasses. At each redd, the following data were collected: • redd location (using GPS); • redd habitat (pool, riffle, glide, tailout, channel edge/bank); • redd substrate (cobble, gravel, sand); • presence (and number) or absence of live spawners. When reduced water visibility precluded identification of the species of fish on a redd, or when a redd was unoccupied, WT crews made presumptive species assignments based on observed physical characteristics of the redd. The “chinook (assumed)” assignment was given to isolated redds exceeding 4 feet in length and occurring in large-cobble mainstem habitat; “chum (assumed)” was assigned to clustered redds, each less than 4 feet in length, occurring in side channel habitat units.

From each salmon carcass observed, the following data were collected: • carcass location (using GPS); • species; • sex; • body length; • presence/absence of adipose fin; • spawning status (when practical, body cavities of female fish were opened to inspect the degree of egg retention).

Statistical Analyses To facilitate comparison with other fish use studies, average relative fish densities in the Skykomish River were calculated from fish counts and survey areas of all habitat units surveyed, regardless of fish presence or absence. For the purpose of statistical analysis, however, adjusted relative densities were calculated following Pess et al. (2002). In the Skykomish braided reach, as in other large river systems of the northwest thus far studied, many fish species and age classes are absent from the majority of habitat units sampled by snorkelers (G. Pess, NOAA Fisheries, Northwest Fisheries Science Center, pers. comm.). As a result, relative abundance data are commonly dominated by zero counts; such data skew greatly reduces the power to detect statistically significant differences in density when fish are present. Except where noted otherwise (e.g. Table 4), this study reports adjusted relative fish density, defined as the fish count per square meter of survey area excluding habitat units in which no fish were observed. These ‘where present’ density estimates are presented graphically together with data on the proportion of surveyed habitat units with fish present (cf. Pess et al., 2002). Note that fish densities are reported only for summer 2004, since habitat unit areas were not measured in other seasons during which snorkel surveys were performed. In this report, fall and winter 2004 relative abundance data appear as absolute counts.

The assumptions of parametric statistical methodology were evaluated by conducting K-S normality tests and Bartlett’s tests for homogeneity of variance (StatView, SAS Institute Inc.). When appropriate, one-way analysis of variance (ANOVA) was performed for each life stage of each species to test whether adjusted relative density varies significantly with river system (mainstem versus side channel and among side channel systems), habitat unit type (pool, riffle, glide), within-unit survey area, and habitat-unit interface type (latter two factors evaluated only

4 in mainstem). Scheffe’s tests were used for post-hoc comparisons. These analyses were conducted on summer 2004 mainstem fish densities within three reaches: (1) entire study reach (mainstem habitat units 1–74); (2) mainstem above the Startup levee (units 1–42); (3) mainstem below the Startup levee (units 43–74) (see Figs 2–6). Units 42–43 spanning the levee were selected as marking the approximate transition from higher gradient and faster flows to lower gradient and slower flows in the mainstem (cf. Appendix X). Univariate ANOVA was used to test for seasonal variation in relative abundance within the mainstem (summer versus fall 2004) and within side channel network SCC (summer versus winter 2004). This method of analysis was also applied to day versus night snorkel-survey data from side channel network SCC to assess the significance of diel variation in fish counts. When assumptions of parametric testing were not met, Kruskal-Wallis tests were used to investigate significant effects of spatial, temporal and habitat-related variation on relative fish abundance and density.

Results Habitat Designations Preliminary habitat unit maps used during the 2004 fish use surveys are shown in Figures 1–7. Adjustments to these maps made in the field (cf. Reach Nomenclature section of Methods) were restricted to the mainstem channel; no changes in habitat unit type were recorded in the side channel networks. Modifications of preliminary unit type in the mainstem are not presented graphically here (see final habitat unit maps in Appendix X), but are reflected in the total count of pools, riffles and glides surveyed (Table 2). Typical mainstem habitat surveyed is illustrated in Figure 9.

The four side channel networks within the study reach differ markedly in size and complexity. SCA (Fig. 3) is a single, broad channel connecting mainstem habitat units 12 (rm X) and 19 (rm X). Network SCB (Fig. 3) extends downstream from mainstem unit 15 (rm X) via a well-defined and continuously wetted channel (SCB) which rejoins the mainstem at unit 22 (rm X). Channel SCB also contributes to an intermittently wetted channel (SCB1) which rejoins the mainstem downstream at unit 30 (rm X). Network SCC (Figs 5, 7), the most complex within the Skykomish braided reach, contains two distinct systems: a series of channels whose flow originates from surface and possibly also subsurface discharge from the mainstem (channels SCC, SCC1, SCC2, SCC3) and an apparent spring brook system with flow fed by groundwater discharge (channels SBA, SBA1, SBA2, SBA3, SBA4). Collectively, this side channel network connects mainstem habitat units 49 and 58D (rm X and Y, respectively). Network SCD (Fig. 6) is comprised of a main channel extending from the mainstem pool immediately downstream of unit 74 (rm X) to the downstream terminus of the study reach, and associated side channels SCD1–SCD5. Representative habitats surveyed within the side channel networks are illustrated in Figure 9.

Species/Age-Class Composition and Relative Abundance Reach-Level Patterns Washington Trout field crews made over 38,000 individual fish observations within the Skykomish River study reach during snorkel surveys in 2004. Over the course of summer, fall and winter surveys, ten species were documented in the mainstem channel and side channel networks, of which seven were salmonids: bull trout, Salvelinus confluentus; Chinook salmon, Oncorhynchus tshawytscha; chum salmon, Oncorhynchus keta; coho salmon, Oncorhynchus kisutch; coastal cutthroat trout, Oncorhynchus clarki clarki; rainbow trout/steelhead,

Oncorhynchus mykiss; and mountain whitefish, Prosopium williamsoni. Also observed were largescale sucker, Catostomus macrocheilus; threespine stickleback, Gasterosteus aculeatus; and sculpin, Cottus sp. The relative abundance of salmonids varied among habitat units from roughly one-quarter to nearly 100% of the total fish count. In the mainstem and side channels, snorkel survey area represented 19±10% and 26±21% (mean±S.D.), respectively, of total habitat unit wetted area (Fig. 10).

Over 8,500 fish belonging to eight species were observed in the braided reach mainstem, with mountain whitefish and largescale sucker among the most abundant during summer and fall surveys (Fig. 11; Tables 5, 6). Adults of these two species were observed in a large proportion of all mainstem habitat units surveyed (83% and 43%, respectively, during the summer) (Fig. 12C). Anadromous salmonids were dominated by coho, Chinook and chum salmon, and resident salmonids were represented in greatest numbers by juvenile trout (Fig. 11). Fishes found only in the mainstem during the intensive summer snorkel surveys (i.e. not also observed within side channel networks) were juvenile and adult bull trout, and adult Chinook, steelhead, and largescale sucker (Fig. 12).

Approximately 30,000 fish belonging to ten species were observed in the braided reach side channel networks, with juvenile coho salmon far outnumbering all other fishes in the majority of channels during summer and winter surveys (Fig. 13; Tables 7–10). In side channel networks SCB and SCC, juvenile coho were documented in 69% of all habitat units surveyed during the summer (Fig. 12A). Other salmonids observed in side channels, in order of decreasing relative abundance, were rainbow and cutthroat trout, mountain whitefish, and juvenile Chinook salmon (Fig. 13). Adult stickleback were observed at high relative densities but were patchily distributed in the side channels, occurring in just 5% of habitat units surveyed; this species was found only in the side channels (i.e. not also observed in the mainstem) (Fig. 12C).

Reach-level analysis of the summer 2004 snorkel-survey data revealed distinct patterns of spatial variation in fish relative density. For several species and life stages, adjusted relative density (i.e. density evaluated in habitat units with fish present) was significantly higher in the side channel networks than in the mainstem channel. This pattern was found for juvenile Chinook and juvenile coho salmon, adult rainbow trout, and juvenile rainbow/cutthroat trout (Fig. 12A,B). A similar pattern was found for all species-summary categories examined. Although fish were present in approximately 90% of all habitat units surveyed in both the mainstem and side channels, the adjusted relative densities of all species observed, all salmonids, and ESA-listed salmonids were significantly higher in the off-channel networks (Fig. 14).

Species diversity, age-class composition and relative abundance also varied between and within the side channel networks. The two major networks, SCB and SCC, contained the same seven fish species during summer snorkel surveys, but Oncorhynchus and Prosopium were represented only by juveniles in SCB while SCC held both juvenile and adult trout and whitefish (Tables 7– 9). In addition, SCB during the summer contained significantly higher densities of juvenile largescale sucker (t-test: d.f.=6; P<0.05) while SCC contained significantly higher densities of juvenile coho salmon (t-test: d.f.=25; P<0.05) (Fig. 15). Within side channel network SCC, fish species diversity and relative abundance also varied among pools, many of which were isolated from adjacent habitat units during summer low-flow conditions (Fig. 7). Much of this variability was associated with physical differences between the two major channel systems within the SCC network. The spring brook system contained two fewer species than the surface flow channels and, notably, lacked juvenile Chinook during summer surveys (Fig. 13; Tables 8, 9). The surface 6 flow system also contained significantly more fish (i.e. higher average count of all fish species within pools) during the summer than either the spring brook or mainstem (t-tests: d.f.=23, 32; P<0.05) (Fig. 22A). Among-pool variability in side channel network SCC can also be partitioned into a within-system component. Pools in the main surface-flow channel SCC, for instance, contained nearly 20-times as many fish on average than pools in the adjacent surface- flow channel SCC3 (Fig. 7; Table 8).

Habitat Unit Effects For a number of species and age classes, summer fish density showed a significant dependence on habitat unit size and type. In the mainstem, the adjusted relative density of adult rainbow trout and steelhead exhibited an inverse relationship with unit wetted area; a similar pattern was observed for juvenile trout in the side channels (Fig. 16). Significantly higher densities of adult salmonids (excluding Prosopium) were observed in mainstem pools than in mainstem riffles or glides (Fig. 17A). In the single mainstem glide where adult bull trout were observed (unit #48; Fig. 4), bull trout density was higher than in other types of units examined during summer mainstem snorkel surveys (Fig. 17B). Juvenile salmonids in the mainstem were found in highest densities within riffles, a pattern statistically significant for juvenile trout throughout the mainstem (units 1–74) and for juvenile Chinook salmon below the Startup levee (units 43–74) (Fig. 17C,D). Within the side channel networks, no significant relationship between fish density and habitat unit type was detected.

Survey Area Effects Summer fish densities measured in the Skykomish mainstem also exhibited distinct patterns of within- and between-unit spatial variation. In three of the four species/age class categories for which significant differences were detected, survey areas in the center of habitat units contained higher adult fish densities than unit edges (Figs 8, 18A–C). Juvenile salmonids, by contrast, aggregated more densely in edge habitat (Fig. 18D). In four mainstem habitat units, large woody debris and riprap armoring on the right bank supported large populations of both juvenile and adult fish. In two additional units, woody debris jams on the left-bank were also sites of high fish density. On average, 71% of the fish counted within each of these units was associated with edge habitat, of which juvenile trout and salmon comprised the greatest proportion (Table 11).

In all five species/age class categories for which significant differences were detected, survey areas in the middle of habitat units were characterized by significantly higher fish densities than adjacent unit interfaces (cf. Fig. 8). The proportion of interfaces with fish present, however, was uniformly higher than the proportion of associated unit middles containing fish (Fig. 19). In other words, adult and juvenile salmonids were more frequently observed in the transition between habitat units, but within these interfaces fish densities were relatively low. Pool-riffle transitions contained significantly higher salmonid densities, on average, than either riffle-glide or glide-pool transitions (t-tests: d.f.=38,15; P<0.05) (Fig. 20).

Seasonal and Diel Variation Comparison of summer and fall 2004 snorkel survey data reveals marked seasonal variation in the relative abundance, species diversity, and age class distribution of fishes in the Skykomish mainstem. Mean fish counts in pools, riffles and glides declined sharply between summer and fall during daylight hours (Fig. 21). This reduced daytime population in the fall lacked four species observed in the summer (bull trout, Chinook salmon, and rainbow and cutthroat trout),

and overall was comprised of three fewer species of salmonids than the summer population (Tables 5, 6). Chum salmon succeeded Chinook as the most prevalent adult anadromous fish (Fig. 11A,B). In addition, the ratio of juvenile-to-adult trout and salmon observed in pools fell from approximately 10 to 0.01 over this period, reflecting a pronounced change in life history- stage usage of the mainstem during the day.

Snorkel survey data collected during summer and winter 2004 in side channel network SCC highlights additional seasonal variation in fish relative abundance, diversity, and age class structure. Mean daytime fish counts in surface-flow and spring-brook side channel pools showed a significant decrease between the two survey seasons, a pattern also observed in mainstem pools between summer and fall (Fig. 22A). The daytime winter population in side channel network SCC lacked five species observed in the summer (rainbow and cutthroat trout, sculpin, largescale sucker and mountain whitefish) but contained adult chum salmon (Table 10). As in the mainstem, a seasonal reduction in the relative abundance of juvenile salmonids was measured in side channel network SCC. Between summer and winter, the ratio of juvenile-to- adult trout and salmon observed in pools during daylight hours fell roughly 2000-fold in the surface-flow side channel system, and 20-fold in the spring brook channels. Associated with these patterns of seasonal variation in fish use was an 8–13 °C average decline in daytime water temperature in mainstem and side channel pools (Fig. 22B).

Paired day–night snorkel surveys were performed in side channel network SCC in December 2004 to test the hypothesis that fish are in fact present within side channel pools during the winter, but take cover during daylight hours. These surveys revealed that both juvenile and adult trout and salmon, presumably seeking refuge during the day, indeed emerge during the night. Both surface-flow and spring-brook side channel pools showed a trend toward increasing overall relative fish abundance (i.e. juveniles and adults of all species) at night; this pattern was statistically significant for pools fed by surface flow (t-test: d.f.=14; P<0.05) (Fig. 23). Mean fish counts also increased at night for virtually every individual species/age class observed (Table 10). This diel variation was most pronounced for the “unidentified juvenile salmonid” category; the less distinctive winter coloration of juvenile trout and salmon, together with backscatter from underwater illumination at night, often precluded definitive species identification. Adult bull trout, which were not observed during either summer or winter daytime surveys of side channel network SCC, were present at night in the spring brook system in December (Table 10).

Spawning Activity During fall 2004, Washington Trout crews recorded evidence of spawning activity by three species of salmon in the braided reach. A total of 905 redds and 477 adult carcasses were documented. Redd and carcass counts for Chinook salmon were highest in late September, chum salmon spawning began in early October and peaked in early November, and coho carcasses first appeared in early December (Table 12). A summary of the Washington Department of Fish and Wildlife’s 2004 aerial redd surveys of the Skykomish River is presented in Table 13 for comparison to the redd count data collected in this study.

Carcass sex ratios, adipose fin presence/absence, and mean body lengths are given in Table 14. The majority of carcasses examined were female fish of presumed natural origin as evidenced by predominately intact adipose fins. Pre-spawning mortality (PSM) (i.e. death prior to egg deposition) was documented in females of all three salmon species identified. PSM rates for

Chinook and chum were 14% (1 of 7 carcasses examined) and 16% (10 of 63 carcasses), respectively. The single female coho carcass found in the braided reach was also a PSM candidate, bearing full and intact egg skeins.

The locations of all salmon redds and carcasses observed during the spawning surveys are shown in Figs 2–7. The general pattern evident from these maps is that the distribution of salmon spawning activity is species-specific. Chinook redds were built primarily in the mainstem, often near the channel bank (e.g. units #7–13 [Fig. 2] and unit #45 [Fig. 4]), while chum redds were most commonly found within the four side channel networks (Figs 3, 5, 6, 7, 24A). Several exceptions to this pattern of non-overlapping spawning areas are noted: Chinook and chum redds were found together within mainstem units #18–20 at the mouth of side channel network SCA (Fig. 3). The only instance of confirmed chum redds outside the immediate vicinity of a side channel was in mainstem unit #35 (Fig. 3). Chinook and chum carcasses showed a similar distribution to that of redds (Fig. 24B). Although largely restricted to distinct spawning areas, carcasses of the two species were found in proximity within side channel SCB, resembling the adjacent mainstem in its habitat unit diversity (Fig. 3), as well as in a seasonal backwater channel in the vicinity of mainstem units #42–44 (Fig. 4) and in the lower reaches of side channel SCC (Figs 5, 7). A large proportion of chum redds were occupied by live spawners (93% of mainstem redds and 86% of side channel redds), whereas Chinook redds contained live fish less frequently (22% in mainstem; 57% in side channels) (Fig. 24A).

The habitat preferences of spawning salmon in the braided reach were examined in detail by analyzing the physical characteristics of redd sites. The majority of Chinook redds were observed in glides, while a large proportion of chum redds were built in pools; for both species, riffle habitats were also frequently selected for spawning (Fig. 25A,B). Chinook redds were constructed almost exclusively in large cobble; chum salmon utilized a variety of substrates, including cobble, gravel and sand (Fig. 25C,D). The depth of water in which redds were constructed did not differ significantly between Chinook (mean 13.4 in.) and chum (mean 16.4 in.) (t-test: d.f.=116; P=0.09).

Washington Trout’s spawning surveys were generally more extensive in their geographic coverage than were snorkel surveys within a given channel system. Accordingly, the spawning surveys provided information about fish use that helped refine the distributional data obtained from snorkel surveys earlier in the year. For example, the small left-bank braid of the mainstem at habitat units #38–39 was dry during summer 2004 and excluded from snorkel surveys. In November, the same channel was found to contain the highest density of chum salmon redds within the braided reach (Fig. 4).

Discussion Fish Abundance and Density Sources of Sampling Bias Fish counts obtained by snorkel survey typically underestimate the “true” number of fish present in a habitat unit as determined by more accurate methods of measuring fish abundance (e.g. electrofishing/removal, mark–recapture) (Rodgers et al., 1992). The calibration recommended by Hankin and Reeves (1988) involving adjustment of snorkel counts by the ratio of “true” fish numbers to snorkel estimates was not performed in this study because exhaustive removal methods were not feasible in the large habitat units surveyed (see also Thompson and Lee, 2000;

Burnett, 2001). Accordingly, the Skykomish River snorkel counts are assumed to be negatively biased, but since the amount of bias is unknown they are used only to provide estimates of relative abundance and density.

Snorkel-survey area represented an approximately constant proportion of total habitat unit wetted area for small and intermediate-sized units in both the mainstem and side channels. For the largest units sampled (>25,000 m2 in the mainstem and >7,000 m2 in the side channels), however, survey area comprised a disproportionately small fraction of total unit area (Fig. 10). Hence, fish densities recorded for these largest units may be artificially low due to sampling bias. The inverse relationship between Oncorhynchus mykiss density and habitat unit area (Fig. 16) suggests preferential use of smaller units by juvenile and adult trout, but it is important to recognize that true densities in the largest units may be higher than those observed.

Water velocities in the Skykomish mainstem prevented snorkelers from surveying in an upstream direction, as is typical in studies of smaller river systems (e.g. Hankin and Reeves, 1988) to avoid disturbing or displacing fish. This study, therefore, was subject to potential observer bias in fish distribution within habitat units. However, a consistent pattern of declining relative abundance from upstream to downstream within units, an expected result of such bias, was not observed.

Variation in water clarity during summer and fall snorkel surveys must also be taken into consideration when interpreting the relative fish abundance patterns described here. The marked seasonal decline in fish counts within the mainstem and side channels (Figs 21, 22) is likely due in part to reduced daytime visibility during surveys performed later in the year (Table 1).

Patterns of Fish Density in the Skykomish and Other Northwest Rivers Comparison of fish abundance data among river systems is complicated by variation in enumeration methodology used in different studies (Lister and Genoe, 1970; Rodgers et al., 1992). Such comparison is further confounded by sampling performed at different times of year and by the use of different means for calculating relative fish density. To facilitate accurate comparison of relative density data from the Skykomish River braided reach and from other large northwest river systems we examine below densities derived from snorkel surveys of the same season and consider different density metrics separately.

The traditional index of mean relative fish density, widely reported in fish use studies, is the average number of individuals observed per unit area in all habitat units surveyed, including those units with fish absent. For juvenile salmon and trout in the Skykomish braided reach, this traditional density measure falls within the ranges reported for other northwest river systems. Specifically, the average relative densities of juvenile trout and Chinook salmon measured during summer in the Skykomish mainstem (Table 4) match those documented in other large river mainstems (i.e. >50 m bankfull width), including the Wenatchee (WA) and Grande Ronde (OR) (Table 15) (see also review by Bartz et al., 2006).

An alternative mean density index, which excludes from consideration all surveyed habitat units with fish absent, can be employed when fish are patchily distributed among units to facilitate parametric statistical analyses (cf. Methods). This adjusted relative fish density, or “where present” density index, is not currently in widespread use but here is compared among sites for which data are available. The mean adjusted relative densities of adult Chinook salmon,

10 steelhead and whitefish measured in the Skykomish River mainstem during summer snorkel surveys (Fig. 12A–C) are very similar to those reported by Pess et al. (2002) for the North Fork Stillaguamish River (WA). In addition, juvenile salmon and trout exhibited mean summer “where present” densities in the Skykomish mainstem (Fig. 12A,B) that closely match those recorded for the mainstem Elwha River (WA) (Pess et al., 2002). In off-channel habitat, mean adjusted relative densities of juvenile salmonids from the Skykomish and Elwha in summer show general agreement with the exception of juvenile Chinook, for which average summer “where present” densities were an order of magnitude higher in the Skykomish side channel networks (cf. Fig. 12A, Table 15).

Patterns of spatial and temporal variation in fish density in the Skykomish braided reach are also largely consistent with those documented in other large northwest rivers. The relatively great abundance of juvenile coho salmon in side channel network SCC during summer (Fig. 13B,C) reflects the well-documented importance of off-channel floodplain habitat for coho rearing (e.g. Nickelson et al., 1992). Juvenile coho counts in side channel network SCC exceeding those in side channel network SCB and in the Skykomish mainstem (Figs 11A, 13) may be related to site-specific variation in water temperature. Mean summer temperatures measured in pools were up to 8 °C cooler in side channel network SCC than in network SCB or the mainstem, reflecting the possible use by juvenile coho of groundwater-influenced off-channel habitat in network SCC as thermal refugia (cf. Brown, 2002).

The marked decline in relative fish abundance between summer and fall/winter measured in the Skykomish during daylight hours (Fig. 22A) is likely related to the reduced swimming performance of fish at low temperatures. As reviewed by Brown (2002), juvenile salmonids may migrate to protected sites, including the interstices of river bed substrate, to avoid predation and to reduce energy expenditure when swimming ability is compromised. The diel variation in fish counts recorded in December 2004 within Skykomish side channel network SCC (Fig. 23) matches a general behavioral pattern documented for salmonids during winter: residing in cover during the day and emerging to feed at night (Lister and Genoe, 1970; Hillman et al., 1989; Peters et al., 1998; Brown, 2002).

The hypothesis that habitats with relatively great physical heterogeneity support higher densities of fish than more homogeneous habitats received partial support from this study. In the Skykomish mainstem, juvenile salmonids exhibited significantly higher relative densities at the edges of habitat units, especially those with marginal woody debris and riprap, than in the center of units (Fig. 18D, Table 11). A similar pattern has been noted for juvenile salmonids in other Pacific Northwest rivers (e.g. Lister and Genoe, 1970; Cramer, 2001; Jeanes and Hilgert, 2001; Brown, 2002; Beamer et al., 2005). The concentration of adult fish in the less physically complex centers of habitat units (Fig. 18A–C) may be a function of the relative scarcity of large woody debris in the Skykomish mainstem and of the ability of adults to navigate more effectively the higher water velocities of the thalweg. Despite the presumed physical heterogeneity of habitat unit interfaces (cf. Fig. 8), neither juvenile nor adult fish showed their highest relative densities in these areas, typically aggregating more densely in the adjacent, longitudinal “middles” of units (Fig. 19). It is notable, however, that salmonids were more frequently observed in these transitional habitats, raising the possibility that unit interfaces support relatively low-density, evenly distributed populations rather than function as discrete distributional “hot spots.” Further study of transitional habitats in the Skykomish and other river systems is warranted to improve our understanding of the relationship between fish use and habitat complexity. 11

Patterns of Salmon Spawning Activity Spawning surveys of the Skykomish braided reach performed independently by Washington Trout and by Washington Department of Fish and Wildlife (WDFW) in fall 2004 allow a valuable comparison of redd count data collected by different techniques. During the two weeks for which data are available from both sources, redd counts obtained by boat and on foot (WT) closely match those obtained by aerial survey (WDFW). Specifically, 118 Chinook redds were counted by WDFW on September 26, 2004 between “Big Eddy” in Gold Bar, WA to the mouth of the Sultan River; 85 Chinook redds were counted by WT four days later in the same reach. At the end of the Chinook spawning run, WT and WDFW each recorded zero Chinook redds (surveys on November 9 and November 10, 2004, respectively) (Table 12; Jackson, 2005b). The total number of Chinook redds documented by WT in the Skykomish braided reach during fall 2004 (136), however, represents roughly one-third of the total tally reported by WDFW (418) (Jackson, 2005a). This discrepancy may reflect differences in total area surveyed, survey effort (i.e. number of flights/surveys), error in salmon species-assignment of redds, and/or technique- specific variation in sensitivity of redd detection.

The Skykomish River 2004 redd survey data underscore the importance of the braided reach as a productive salmon spawning ground. Of 840 Chinook redds documented by WDFW in fall 2004 between the mouth of the Skykomish River and the anadromous fish barrier at Sunset Falls (rm 51.5), approximately 50% were observed in the braided reach. The number of Chinook redds observed per river mile in the braided reach (49.2) exceeded that in all other WDFW survey reaches on the Skykomish River in 2004 (Table 13).

The documentation of pre-spawning mortality in female Chinook, chum and coho salmon within the braided reach (see Results) adds to a growing set of PSM observations across western Washington. To date, the phenomenon has been noted for several salmonid species with intensive spawning studies focused on both rural/residential habitats and urban creeks (e.g. Glasgow et al., 2005; McMillan et al., 2006; N. Scholz, NOAA Fisheries, Northwest Fisheries Science Center, pers. comm.). Ongoing work by Washington Trout and by NOAA Fisheries seeks to refine understanding of the current geographic extent of PSM in the Pacific Northwest and to identify causal factors, including patterns of land use and water quality, related to PSM incidence.

Summary Fish species composition and relative abundance: • During summer, fall and winter 2004 snorkel surveys of the Skykomish River braided reach, ten fish species were documented in the mainstem channel and side channel networks, of which seven were salmonids and two were ESA-listed (bull trout and Chinook salmon). • Relative fish densities measured by snorkel survey during summer in the Skykomish braided reach were generally similar to those recorded for other large northwest rivers. • Relative fish density was generally higher in the side channel networks than in the mainstem channel during summer. This pattern was significant for all ESA-listed salmonids encountered. Side channel pools supported relatively more fish than mainstem pools during winter. • Juvenile coho salmon far outnumbered all other fishes in the majority of side channels surveyed during summer and winter 2004, occurring in nearly 70% of all habitat units examined. • Fish species diversity, age-class composition and relative abundance varied significantly between and within the Skykomish River side channel networks. • The adjusted relative density (i.e. density measured only for habitat units with fish present) of adult and juvenile rainbow trout/steelhead exhibited an inverse relationship with habitat unit wetted area during summer. • Significantly higher densities of adult trout and salmon were observed in mainstem pools than in mainstem riffles or glides during summer. Juvenile salmonids in the mainstem were found in highest densities within riffles during summer, but in the side channel networks, no significant relationship between fish density and habitat unit type was detected. • Summer fish densities measured in the Skykomish mainstem exhibited distinct patterns of within- and between-unit spatial variation: (a) adult salmonids were typically found in higher densities within the center of habitat units than at unit edges, while juvenile salmonids aggregated more densely in edge habitat including large woody debris and riprap bank armoring; (b) adult and juvenile salmonids were more frequently observed in the transition area between habitat units than in the longitudinal “middle” of units, but within these interfaces fish densities were relatively low. Pool-riffle and riffle-pool transitions contained significantly higher salmonid densities, on average, than all other habitat unit transitions studied. • Mean daytime fish counts in both the mainstem and side channels declined sharply between summer and fall/winter. During winter, mean counts increased at night for most fish species and age classes examined.

Summary, continued Salmon spawning activity: • During fall 2004 spawning surveys of the Skykomish braided reach, the majority of salmon carcasses encountered were female Chinook and chum of presumed natural origin (intact adipose fins). Pre-spawning mortality (i.e. death prior to egg deposition) was documented in 14% of Chinook females and 16% of chum females. • The distribution of salmon spawning activity was species-specific: Chinook redds were built primarily in the mainstem, often near the channel bank, while chum redds were most commonly found within side channels. • Redd counts obtained by boat survey and on foot approximated those obtained by WDFW aerial surveys in fall 2004. Both datasets highlight the Skykomish River braided reach as an important salmon spawning ground.

References

Bartz, K. K., Lagueux, K. M., Scheuerell, M. D., Beechie, T., Haas, A. D. and Ruckelshaus, M. H. (2006). Translating restoration scenarios into habitat conditions: an initial step in evaluating recovery strategies for Chinook salmon (Oncorhynchus tshawytscha). Can. J. Fish. Aquat. Sci. 63: 1578-1595.

Beamer, E. M., Hayman, B. and Smith, D. (2005). Linking freshwater rearing habitat to Skagit Chinook salmon recovery. Appendix C of the Skagit Chinook Recovery Plan 2005. Skagit River System Cooperative, La Conner, WA, 24 pp.

Brown, T. G. (2002). Floodplains, flooding, and salmon rearing habitats in British Columbia: a review. Fisheries and Oceans Canada, Canadian Science Advisory Secretariat. Research Document 2002/007, 155 pp. & appendices.

Burnett, K. M. (2001). Relationships among juvenile anadromous salmonids, their freshwater habitat, and landscape characteristics over multiple years and spatial scales in the Elk River, Oregon. Ph.D. thesis, Oregon State University, Corvallis, OR.

Cramer, S. P. (2001). The relationship of stream habitat features to potential for production of four salmonid species, draft report. Prepared by S.P. Cramer & Associates, Inc. February 2001, Gresham, OR., viii+182 pp.

EPA (2002). Guidance on choosing a sampling design for environmental data collection. EPA QA/G-5S.

Snohomish River Basin Salmon Recovery Forum. (2005). Draft Final Snohomish Basin Salmon Conservation Plan. June 2005. Snohomish County Surface Water Management Division, Everett, WA.

Glasgow, J., Drucker, E. and Russell, D. (2005). Land use and coho prespawning mortality in the Snohomish watershed, Washington. Poster presentation at WDFW Habitat Program's All Hands Meeting, April 26–27, 2005, Wenatchee, WA.

Hankin, D. G. and Reeves, G. H. (1988). Estimating total fish abundance and total habitat area in small streams based on visual estimation methods. Can. J. Fish. Aquat. Sci. 45: 834-844.

Hillman, T. W. and Chapman, D. W. (1989). Abundance, growth, and movement of juvenile Chinook salmon and steelhead. In Summer and winter ecology of juvenile Chinook salmon and steelhead trout in the Wenatchee River, Washington. Final Report to Chelan County Public Utility District, Wenatchee, WA, pp. 1-41. Don Chapman Consultants, Inc., Boise, ID.

Hillman, T. W., Chapman, D. W. and Griffith, J. S. (1989). Seasonal habitat use and behavioral interaction of juvenile Chinook salmon and steelhead. I. Daytime habitat selection. In Summer and winter ecology of juvenile Chinook salmon and steelhead trout in the Wenatchee River, Washington. Final Report to Chelan County Public Utility District, Wenatchee, WA, pp. 43-82. Don Chapman Consultants, Inc., Boise, ID.

Jackson, C. (2005a). Email to E. Drucker reporting unpublished WDFW redd count data, December 5, 2005.

Jackson, C. (2005b). Email to E. Drucker reporting unpublished WDFW redd count data, June 22, 2005.

Jeanes, E. D. and Hilgert, P. J. (2001). Juvenile salmonid use of lateral stream habitats, Middle Green River, Washington, 2000 Data Report. Prepared by R2 Resource Consultants, Inc. July 19, 2001, Redmond, WA, viii+63 pp. & appendix.

Jonasson, B. C., Carmichael, R. W. and Keefe, M. (1997). Investigations into the early life history of naturally produced spring Chinook salmon in the Grande Ronde River basin. Oregon Department of Fish and Wildlife, Annual Progress Report, 1 September 1995 to 31 August 1996, 28 pp.

Keefe, M., Carmichael, R. W., Jonasson, B. C., Messmer, R. T. and Whitesel, T. A. (1995). Investigations into the early life history of naturally produced spring Chinook salmon in the Grande Ronde River basin. Oregon Department of Fish and Wildlife, Annual Progress Report, 1 September 1993 to 31 August 1994.

Lister, D. B. and Genoe, H. S. (1970). Stream habitat utilization by cohabiting underyearlings of chinook (Oncorhynchus tshawytscha) and coho (O. kisutch) salmon in the Big Qualicum River, British Columbia. J. Fish. Res. Bd Can. 27: 1215-1224.

McMillan, B., Crabb, D., Staller, F., Drucker, E. and Glasgow, J. (2006). Unexpected abundance: coastal cutthroat trout (Oncorhynchus clarki clarki) as the inheritors of Seattle urban creeks in the declining presence of other wild salmonids. In review for Proceedings of the 2005 Coastal Cutthroat Trout Symposium: Biology, Status, Management, and Conservation, Port Townsend, WA.

Nickelson, T. E., Solazzi, M. F., Johnson, S. L. and Rodgers, J. D. (1992). Effectiveness of selected stream improvement techniques to create suitable summer and winter rearing habitat for juvenile coho salmon (Oncorhynchus kisutch) in Oregon coastal streams. Can. J. Fish. Aquat. Sci. 49: 790-794.

Pess, G. R., Liermann, M., McHenry, M., Bennett, T., Peters, R., Kiffney, P. and Coe, H. (2002). Juvenile and adult salmonid response to the placement of logjams in the Elwha and Stillaguamish Rivers: Preliminary results. Final draft submitted by NOAA Northwest Fisheries Science Center to Stillaguamish Tribe, Lower Elwha Klallam Tribe and Washington Trout, 49 pp.

Peters, R. J., Missildine, B. R. and Low, D. L. (1998). Seasonal fish densities near river banks stabilized with various stabilization methods: First year report of the flood technical assistance project. Prepared by the U.S. Fish and Wildlife Service December 1998, Lacey, Washington. 34 pp.

Scarnecchia, D. L. and Roper, B. B. (2000). Large-scale, differential summer habitat use of three anadromous salmonids in a large river basin in Oregon, U.S.A. Fish. Manag. Ecol. 7: 197- 209.

Rodgers, J. D., Solazzi, M. F., Johnson, S. L. and Buckman, M. A. (1992). Comparison of three techniques to estimate juvenile coho salmon populations in small streams. N. Am. J. Fish. Manag. 12: 79-86.

Thompson, W. L. and Lee, D. C. (2000). Modeling relationships between landscape-level attributes and snorkel counts of chinook salmon and steelhead parr in Idaho. Can. J. Fish. Aquat. Sci. 57: 1834-1842.

Figure Legends for maps on pages 19–25 Fig. 1. Overview of the Skykomish River braided reach with preliminary habitat type designations. Contiguous segments of the study reach are illustrated in greater detail in Figures 2–7. Note that final habitat type designations are presented in Appendix X.

Fig. 2–6. Preliminary habitat type designations and observed salmon spawning activity within the Skykomish River braided reach. Redd and carcass data were collected between September and December 2004 (Tables 1, 12).

Fig. 7. Detailed view of side channel network SCC (cf. Fig. 5).

Figure 8. Schematic illustration of the snorkel-survey subareas within each mainstem habitat unit. Two consecutive units are depicted. Ten-foot wide snorkel lanes (gray shading) defined left, center, and right subareas. Upper, middle and lower subareas of equal size were defined by transverse boundaries (approximated in the field). The area of edge habitat surveyed within each unit was taken as the sum of left and right subareas. Habitat unit interface area was taken as the sum of adjacent lower and upper subareas in consecutive units.

26 Figure 9 Representative Skykomish River braided reach habitats surveyed in 2004

Snorkel survey of mainstem habitat unit #48 on left-bank Snorkel survey of isolated pool in lower reach of (November 2004). side channel SCB1 (August 2004).

Snorkel survey of pool within side channel network SCC (December 2004).

Figure 9 (continued) Representative Skykomish River braided reach habitats surveyed in 2004

Upstream view of confluence of side channel SCC (at left) Chum redd with live fish (in background) within and SBA1 (October 2004). side channel SCC (October 2004).

Spawning survey of side channel SCD (December 2004).

Figure 10. Relationship between snorkel-survey area and total habitat unit wetted area in Skykomish River braided reach during summer 2004. Linear regression lines are shown for small to intermediate sized habitat units in the mainstem (y=0.150x+323.0; r2=0.66) and in the side channels (y=0.133x+63.3; r2=0.87). In the largest units, survey area comprised a disproportionately small fraction of total unit area. Side channel data are pooled values from networks SCB and SCC.

Figure 11. Fish species composition and relative abundance in mainstem Skykomish River braided reach assessed by seasonal snorkel surveys. Pie charts present counts of largescale sucker, mountain whitefish and ‘other’ fishes (sculpin and unidentified non-salmonid juveniles) for all pools, riffles and glides surveyed. Bar charts illustrate relative abundance of all salmonids excluding whitefish; juveniles and adults of these species are totaled separately (shown above and below dashed lines, respectively). Counts of largescale sucker and mountain whitefish are the sum of juvenile and adult counts for each species. Fish count totals: 8,206 in summer 2004 (76 units surveyed); 326 in fall 2004 (20 units). All counts made during daylight hours.

Figure 12 (legend on p. 33) 31

Figure 13 (legend on p. 33)

Figure 12. Relative densities of all fish species and age classes observed in the Skykomish River braided reach during summer 2004 snorkel surveys. The adjusted (“where present”) density of each species/age class is shown together with the proportion of surveyed habitat units with fish present. All data presented as mean±S.E.M.; densities calculated from an average of 14 habitat units per species/age class (range: 1–60 units). Significant differences in relative density between the mainstem (MS) and side channels (SC) (pooled data from networks SCB and SCC) were assessed by univariate ANOVA: *, P<0.05; **, P<0.001. n/o, not observed; †juvenile rainbow/cutthroat trout.

Figure 13. Fish species composition and relative abundance in Skykomish River braided reach side channel pools. Data from summer 2004 snorkel surveys. Pie charts present counts of juvenile coho salmon; bar charts illustrate relative abundance of all other species/age classes observed. (A) Network SCB (channels SCB and SCB1; Fig. 3): 4,941 fish in 11 pools; (B) network SCC receiving surface flow from the mainstem (channels SCC and SCC3; Fig. 7): 22,020 fish in 14 pools; (C) network SCC spring brook system (channels SBA and SBA1; Fig. 7): 1,141 in 4 pools. Counts of mountain whitefish are the sum of juveniles and adults. ‘Other’: sculpin and unidentified non-salmonid juveniles. All counts made during daylight hours.

Figure 14. Relative fish densities for five species categories (defined in text) from summer 2004 snorkel surveys of the Skykomish River braided reach. The adjusted (“where present”) density of each category is shown together with the proportion of surveyed habitat units with fish present. All data presented as mean±S.E.M.; densities calculated from an average of 46 habitat units per category (range: 29–67 units). Significant differences in relative density between the mainstem and side channels (pooled data from networks SCB and SCC) were assessed by univariate ANOVA: *, P<0.05; **, P<0.001.

Figure 15. Comparison of adjusted relative densities of (A) juvenile coho salmon and (B) juvenile largescale sucker in pools within side channel networks SCB and SCC during summer 2004. Data are presented as mean±S.E.M. with number of pools sampled. Asterisks indicate significant differences in relative density at P<0.05 (univariate ANOVA). Densities of all other species/age classes observed (see Figs 12, 14) were not significantly different in the two side channel systems.

Figure 16. Relationship between adjusted relative fish density and habitat unit area for trout and steelhead in the Skykomish mainstem and side channels (pooled data from networks SCB and SCC) during summer 2004.

Figure 17. Relative fish densities for different habitat unit types in the Skykomish braided reach mainstem during summer 2004. Adjusted (“where present”) densities are presented for selected species as mean±S.E.M. with associated proportions of units with fish present and number of units sampled. Asterisks indicate significant differences in relative density at P<0.05 (univariate ANOVA). No significant differences in density among habitat unit types were detected for species/age classes other than those presented here (see Figs 12, 14). Data in A–C are from mainstem habitat units 1–74 and in D are from units 43–74 (below the Startup levee).

Figure 18. Relative fish densities for habitat unit edges and centers (cf. Fig. 8) in the Skykomish braided reach mainstem during summer 2004. Adjusted (“where present”) densities are presented for selected species as mean±S.E.M. with associated proportions of survey subareas with fish present and number of subareas sampled. Asterisks indicate significant differences in relative density at P<0.05 (univariate ANOVA). No significant differences in density between unit edges and centers were detected for species/age classes other than those presented here.

Figure 19 (legend on p. 40)

Figure 19. Relative fish densities for habitat unit middles and adjacent unit interfaces (cf. Fig. 8) in the Skykomish braided reach mainstem during summer 2004. Adjusted (“where present”) densities are presented for selected species as mean±S.E.M. with associated proportions of survey subareas with fish present and number of subareas sampled. Asterisks indicate significant differences in relative density at P<0.05 (univariate ANOVA). No significant differences in density between unit middles and interfaces were detected for species/age classes other than those presented here. Data in A–C are from mainstem habitat units 1–74 and in D and E are from units 43–74 (below the Startup levee).

Figure 20. Relative fish densities for habitat unit interfaces in the Skykomish braided reach mainstem during summer 2004. Adjusted (“where present”) densities are presented as mean±S.E.M. for all salmonid species (adults and juveniles collectively excluding Prosopium) with associated proportions of interfaces with fish present and number of interfaces sampled. Each interface category contains both possible transitions between unit types (e.g. ‘pool-riffle’ includes pool-to-riffle and riffle-to-pool). Mean salmonid density was significantly higher in pool-riffle transitions than in either of the other two transition types examined (univariate ANOVA: P<0.05).

Figure 21. Seasonal variation in Skykomish mainstem fish counts (juveniles and adults of all species; daytime surveys). Data are presented as means with S.E.M. errors bars and habitat unit sample sizes. Mean fish counts in all unit types sampled declined significantly from summer to fall (t-tests: d.f.=19, 40, 31; P<0.05).

Figure 22. Spatial and seasonal variation in (A) daytime fish counts (juveniles and adults of all species) and (B) corresponding daytime temperatures in Skykomish braided reach pools. Data are presented as means with S.E.M. errors bars and pool sample sizes for the surface flow and spring brook systems within side channel network SCC (summer and winter) and for the mainstem (summer and fall) (cf. Table 1). Mean summer counts were significantly higher in the surface-flow side channels than in either the spring brook side channels or the mainstem (univariate ANOVA: P<0.05). For all three sites, mean relative fish abundance declined significantly from summer to fall/winter (t-tests: d.f.= 26, 9, 20; *, P<0.05; **, P<0.001).

Figure 23. Diel variation in winter fish counts (juveniles and adults of all species) for side channel pools within the Skykomish braided reach. Data from December 2004 snorkel surveys of side channel network SCC are presented as means with S.E.M. error bars and pool sample sizes. Asterisk indicates significant difference in average daytime and nighttime fish counts within surface-flow side channel pools (univariate ANOVA: P<0.05).

Figure 24. Distribution of salmon redds and carcasses in the Skykomish braided reach. MS, mainstem; SC, side channels.

Figure 25. Salmon redd counts by habitat unit type (A, B) and by substrate type (C, D). Redds for which habitat and substrate determinations were not recorded during spawning surveys are not enumerated here; see Table 12 for total redd counts.