Aquatic and riparian ecosystem recovery from debris flows in two western Washington streams, USA

Abstract An exceptionally powerful storm struck southwestern Washington in December 2007 causing large debris flows in two adjacent streams. The two affected streams had been studied prior to the storm, providing a rare opportunity to examine ecosystem recovery. We monitored the streams and their riparian zones for six years after the disturbances to determine whether recovery rates of biota, physical habitat, and water temperature differed, and if so, what factors affected resilience. Along both streams, the debris flows removed wide swaths of soil, rock, and coniferous riparian forests, widening the active channel and increasing solar exposure and summer water temperatures. Initially depauperate of vegetation, after four years red alder trees dominated the riparian plant communities. The warmer water, greater solar radiation, and unstable substrates likely contributed to variable benthic insect and tailed frog tadpole densities over time, although benthic insect communities became more similar after three years. The debris flows also decreased channel slopes and removed channel step barriers such that cutthroat trout were able to rapidly occupy habitats far upstream, but sculpins were slower to recolonize and both fish species exhibited some differences in recovery between the two streams. Crayfish were severely impacted by the debris flows; this may be due to attributes of their life history and the timing of the flows. Overall, we found that recolonizing aquatic species exhibited varying levels of resilience and recovery after the disturbances being related to the influence of physical habitat conditions, species dispersal ability, and the presence of nearby source populations.

with water and accumulates additional substrate and wood from the stream channel and riparian area as it moves downstream. This forms a fast-flowing slurry of fine and coarse sediment combined with trees and other organic debris (Takashi, 2014), often decimating biota in the stream and adjacent riparian area. However, debris-flow delivery of coarse sediment and wood can increase wood-associated pools known to be beneficial to salmonids (Benda et al., 2003;Reeves, Benda, Burnett, Bisson, & Sedell, 1995). As a result, the spacing and network pattern of debris-flow-prone headwater streams exhibits a high degree of physical heterogeneity (Bigelow, Benda, Miller, & Burnett, 2007;Borga, Stoffel, Marchi, Marra, & Jakob, 2014;Gomi, Sidle, & Swanston, 2004), and in this regard, debris flows are considered important agents of habitat formation in aquatic ecosystems (Bisson, Dunham, & Reeves, 2009).
Despite the long-term aquatic habitat benefits of debris flows, the immediate consequences of these disturbances are usually catastrophic for biota in and adjacent to the impacted stream channels.
Natural resource managers often undertake restoration actions such as restocking fish or creating habitat where structural roughness elements including large wood and boulders have been scoured (Montgomery et al., 2002). While the effects of restoration projects have received extensive study (Roni, 2005), fewer investigations have been conducted of streams that are recovering passively without intervention from both human-caused and natural disturbances (Martens, Devine, Minkova, & Foster, 2019).
A powerful storm struck southwestern Washington State, USA, on 1-3 December 2007. Dubbed the "Great Coastal Gale," this storm brought a combination of snow, gale force winds, and heavy rains which caused many landslides, debris flows, and severe lowland flooding across the region (Mote, Mault, & Duliere, 2007;Read, 2007). In Capitol State Forest, an actively managed forest near Olympia, Washington, two adjacent streams experienced debris flows during this December 2007 storm. Coincidentally, these two streams had several years of prestorm aquatic biota and temperature data, providing a rare opportunity to examine the effects of these unusual disturbances. Because of the unpredictable nature of debris flows, recovery studies typically involve comparisons between affected and unaffected areas, or retrospectively comparing changes at different intervals after an event. For example, Cover et al. (2010) used a space-for-time substitution on 10 debris-flow-affected streams in northern California to infer disturbance effects on stream ecosystem structure. In the Olympic Mountains of western Washington, periphyton biomass, water temperature, chemistry, and aquatic macroinvertebrates were compared between a debris-flow-affected stream and nearby control stream (Kiffney et al., 2004). Snyder and Johnson (2006) employed a study design that combined channel stratification along a debris-flow track with a nearby stream that was not disturbed to assess differences in aquatic insect assemblages in Shenandoah National Park, Virginia.
To our knowledge, only three published debris-flow studies were fortunate enough to have predisturbance data for beforeafter comparisons. In a western Oregon Cascade Range stream partially affected by a large debris flow, Lamberti, Gregory, Ashkenas, Wildman, and Moore (1991) compared an upstream unaffected area and downstream disturbed area. Roghair, Dolloff, and Underwood (2002) stratified a debris flow in Shenandoah National Park, Virginia (the same event studied by Snyder & Johnson, 2006), to assess brook trout (Salvelinus fontinalis) colonization and movements, comparing affected and unaffected areas. As cited previously, Kiffney et al. (2004) had predisturbance data and were able to make before-after comparisons for several metrics. Results are forthcoming from a fourth study conducted in a nearby watershed affected by the same 2007 storm event as our study (Fransen, Walter, Reiter, & Tarosky, 2013). In this study, we use predisturbance data for aquatic biota and water temperature for several years prior to the two adjacent debris flows to make before-after comparisons. The data were amassed as part of the Riparian Ecosystem Management Study (REMS) on headwater stream riparian buffers from 2002 to 2006 (Bisson, Claeson, Wondzell, Foster, & Steel, 2013).
The objectives of this study of debris flows were to (a) evaluate the recovery of the streams by fishes (primarily trout and sculpin), aquatic insects, and riparian vegetation; (b) document post-debris flow changes in stream channel morphology and water temperatures; and (c) describe interactions among physical habitat, water temperature, and species colonization. Specifically, we wanted to determine whether the species patterns of temporal and spatial reoccupation were similar between the two disturbed streams, and how aquatic communities and habitat conditions compared before and after the debris flows and to a nearby unimpacted reference stream.
We use the results of our monitoring to infer the mechanisms that

Channel morphology
To assess changes to channel morphology in Camp Four and Potosi creeks before (2002) and after (2011) the debris flows, we used Light Detection and Ranging (LiDAR) data. We first normalized the LiDAR data so that the same bare earth point density was obtained between data sets. We performed before and after impact comparisons for channel slope and valley floodplain width along both debris-flow tracks. Channel-slope profiles used x, y, and z cell values from the LiDAR data, computing an overall slope % from the elevation (z) change from the toe of the initiating landslide to the Waddell Creek confluence. Floodplain width was the distance perpendicular to the channel between the bases of opposing valley walls (Grant & Swanson, 1995;May, 2002) taken at specific locations coinciding F I G U R E 1 Study area location in Washington State, USA with vegetation reference points (RPs) described below, as defined by high-resolution contour maps derived from the LiDAR data. At each cross-sectional location, we computed the difference and percent change in floodplain width before and after the debris flows.

Stream temperature
We collected stream temperatures in the summer for three years before (2003)(2004)(2005) and six years after (2008-2013) the debris flows.
In the predisturbance years, as part of the REMS study, we placed iButton ® (Maxim Integrated, precision of ±0.5°C) temperature loggers at three locations in Camp Four (CF2, CF4, and CF6), three in Potosi (P2, P5, and P6), and two in Sunbeam (SB1 and SB2) ( Figure 2  We calculated the daily maximum temperature difference for each debris-flow site as compared to a control site for each year of the study, and repeated this step for daily minimum temperature differences. The site SB1 was used as the control for sites CF2 and P2, as they are all near the mouth of their respective streams, whereas SB2 was used as the control for CF4, CF6, P5, and P6, as they are all farther upstream ( Figure 2). In the pre-disturbance years, temperature response differences were most often zero between the disturbed and the control sites. Given the small variation in temperature differences within each 3-month annual period of study, the temperature responses were modeled as the difference between disturbed site and control site temperature per day (i.e., ∆T = Ttrt i − Tref i on each study day i). We use repeated-measures ANOVA (R v.3.1.3) to test for significant differences in daily maximum and minimum temperature differences for each disturbed versus control site comparison between pre-and post years. If the ANOVA was significant, we used Dunnett's contrasts to compare each of the six postyear differences to a single control (all preyears combined). Dunnett's t test is designed to hold the familywise error rate at or below α when performing multiple comparisons of a number of treatments with a single control. Any pre-existing differences between the sites prior to the debris flows were incorporated in the ANOVA mean estimate results. The ANOVA was repeated for each debris-flow site and response metric (daily maximum, daily minimum). Note that water temperature data were not available in 2008 at the debris flow sites CF6 and P5. F I G U R E 2 Debris-flow sampling sites showing vegetation transects and reference points (RP), lower and upper index survey reaches, benthic invertebrate reaches, and temperature monitoring sites (SB, P, CF)

Riparian vegetation
Surveys commenced in 2009 to assess vegetation recolonization and growth along the longitudinal profiles of Camp Four and Potosi creeks. We established two 5-m-diameter circular plots (19.6 m 2 ) at equal distance along transects across the width of each debris flow.
Starting from the confluence with Waddell Creek, plot transects occurred every 200 m going upstream ( Figure 2). We GPS-mapped the right-bank terminus of each transect with differential correction.
These locations were used as reference points (RPs) for spatial verification of other study components (e.g., fish distribution). We collected riparian vegetation data on 11 transects (22 plots We estimated species percent cover using a modified Braun-Blanquet method (Wikum & Shanholtzer, 1978). Most plants were identified to species, except grasses (Poaceae) were identified to order, rushes (Juncus) and sedges (Carex) to genera, and mosses (Bryophyta) to division. Wood debris (minimum of 5 cm diameter), water, and bare ground were each recorded as separate categories.
Total cover may sum to over 100% owing to overlapping vegetation at different heights. For red alder (Alnus rubra) only, we recorded the maximum tree height at each plot. Because the number of plots and transects differed between Camp Four Creek and Potosi Creek, we present the data as the plot mean with one standard error (SE).

Aquatic insects
We collected benthic macroinvertebrates annually from 2008 to 2012 (late May or June) from one 50-m reach on each of the debris-flow streams, Camp Four and Potosi, and the reference stream, Sunbeam ( Figure 2). Each stream-per-year sample was a composite of ten spatially distributed Surber samples (500-μm mesh, 0.09-m 2 benthos) stored in 80% ethanol. All individuals were counted and identified to the lowest taxonomic level possible, usually genus, although Chironomidae (midge larvae) were identified to subfamily (Merritt et al., 2008). We did not include noninsects (e.g., mites, copepods, ostracods, and worms) in the results because of their ubiquitously low abundance. We assigned each insect taxon a functional feeding group (FFG) that describes the individual's primary feeding mechanism (Merritt et al., 2008). We present insect counts as density (number/m 2 ) or proportions (%) of total density. To examine insect community composition and abundance patterns over time, we used nonmetric multidimensional scaling (NMDS, PC-ORD v.7) with Sorenson distance to ordinate 15 benthic samples by 68 insect taxa with density data (McCune & Grace, 2002). We removed rare taxa that were present in only a single sample (13 rare taxa) and log10(x + 1) transformed the densities prior to performing the NMDS. Joint plots (line vectors) display sample-level taxa richness and density proportions associated with each axis (Pearson's cor-

Aquatic vertebrates and crayfish
Prior to the debris flows, aquatic organism information was avail-

| Channel morphology
On Potosi Creek before the debris flow in 2002, the elevation-based slope change from the Waddell Creek confluence to the toe of the initiating landslide (3,584 m) was 10.6%, but after the debris flow in 2011, the slope was 9.1% over the same distance-a 14% decrease.
Likewise, on Camp Four Creek before the debris flow, the average slope from Waddell Creek to the toe of the initiating landslide (2,535 m) was 9.3%, but after the debris flow, the slope was 8.1%-a 13% reduction.
The active channel width increased from before to after the debris flows at all measured cross sections taken at vegetation reference points (RP's) ( Figure 2). Along Potosi Creek, more than 100% widening occurred at three locations with one location at 98% (

| Riparian vegetation
The mature forest that formed the riparian vegetation was completely removed by the debris flows, as average floodplain width increased 65% along Camp Four Creek and 66% along Potosi Creek (Table 2). In summer 2008, prior to the start of our official surveys, only scattered plant seedlings were observed and the soil was very rocky and unconsolidated. Even in 2009, few plants with little cover were recorded, but colonization and growth increased each year. Taxa richness, on average, was slightly higher in Camp Four Creek than Potosi Creek, although both streams followed the same

| Aquatic insects
Initially, the debris flows in Camp Four and Potosi creeks reduced benthic insect density and taxa richness considerably, whereas insects exposed to flooding in Sunbeam Creek were less impacted.
Over time, the insect communities in all three streams increased in richness, abundance, and became more similar in composition. In total, 81 aquatic insect taxa were recorded from Camp Four, Potosi, and Sunbeam from 2008 to 2012 (Table A2). However, had chironomids been identified beyond subfamily, many more Dipteran species would have been counted. Mayflies (Ephemeroptera), stoneflies (Plecoptera), and caddisflies (Trichoptera), collectively termed EPT's, made up the majority of taxa richness (range of 70%-88% richness).
Benthic insect communities from all three streams were primarily influenced by time since disturbance (i.e., sample year) as depicted in the resulting 2-dimensional NMDS ordination of insect densities (stress = 7.68, total R 2 = .94; Figure 6). However, the Sunbeam com-

| Aquatic vertebrates and crayfish
Seasonal and spatial variation of tailed frog tadpole density was apparent in both Potosi and Camp Four creeks before the debris flows.

Ref. point (RP)
Distance upstream a (m)   Table   A3). In 2008, only one tailed frog was captured in Camp Four Creek.
In 2009, the density increased, but in 2010, the density decreased to a level significantly lower than before the debris flow. In 2011 and 2012, the density increased again in Camp Four Creek and was not significantly different than before the disturbance. Tailed frog tadpole densities after the debris flows showed no significant differences between the disturbed streams and Sunbeam Creek, the reference stream (Table A4).
Signal crayfish densities in both Potosi and Camp Four creeks dropped to nearly zero after the debris flows and did not recover to levels found before the debris flows over five years of mon-   (Table A4).
Cutthroat trout density in Potosi Creek in 2008, 2010, and 2011 was significantly higher than before the debris flow ( Figure 8; Table   A3). In comparison, Camp Four Creek trout density in [2008][2009] was somewhat lower than before although not significantly, then The proportional distribution of cutthroat trout age classes from September surveys on Potosi Creek were fairly consistent in the years before the debris flows ( Figure 9; Table A5). Camp Four Creek had more age-0+ fish than Potosi Creek before the debris flow, but lower proportions of age-1+ and 2+ fish and no age-3+  After that, sculpins were captured in both July and September (2011)(2012), and at both reaches (2012). The densities were not significantly different than before the debris flow in 2011 and 2012 F I G U R E 7 Before and after densities (no./m 2 ) of coastal tailed frog tadpoles and signal crayfish occurring on the debris flows. Asterisk mark those years significantly different (p ≤ .05) from the combined average of years before the debris flows (Table A3) F I G U R E 8 Before and after densities of fish occurring on the debris-flow streams. Asterisk mark those years significantly different (p ≤ .05) from the combined average of years before the debris flows (Table A3) (Table A4).  (Table A5) surmounted by the trout. In September 2008, trout were found in a pool below a bedrock chute ( Figure 10a) and then advanced 25 m upstream to a pool below a 4.9-m bedrock chute with 34% channel slope that prevented upstream movement from Average stream gradient measured at the waterfall was 11.5%
In Sunbeam Creek, the reference stream, a wood jam with a sediment terrace above it located 1,613 m upstream from Waddell Creek formed a barrier to cutthroat trout from 2008 through 2012. A complex, 2-m-high waterfall over the jam was present during July surveys; however, the stream became intermittent as it flowed into the sediment terrace during September surveys to re-emerge in a pool at the base of the jam where the farthest upstream trout was always found ( Figure 10f). Here, the average gradient was 11.0% (SD = 7.1; Table A6).  (Table A7). Sculpins were again found progressively upstream in Creek. Average channel slope at that location was 5.9% (SD = 5.2).

Sculpins in Camp Four
Creek were more sporadic and inconsistent in their movement upstream than in Potosi Creek. In Camp Four Creek, sculpins were first located 355 m upstream from Waddell Creek  Table A7). Sculpin distribution was not monitored in Sunbeam Creek.

| Channel morphology
Channel slope decreased along the longitudinal profiles of both Potosi and Camp Four creeks. Sediment terraces retained by large logs or accumulations of smaller pieces of wood often act as "steps" and can account for a significant fraction of the total relief along stair-stepped channels (Lancaster & Grant, 2006;MacFarlane & Wohl, 2003). Debris flows are an effective agent for removing accumulated wood in streams (Hassan, Hogan, et al., 2005). Both wood

| Stream temperature
Stream temperatures increased similarly in both debris flow streams.
After six years, temperatures were still higher than before, but less so than initially and more closely resembled the reference stream.
The primary cause of warming after the debris flows was increased solar energy from the complete removal of riparian vegetation that provided shade. Solar radiation has been shown to be the largest factor in the stream energy budget on summer days in unshaded streams (Janisch, Wondzell, & Ehinger, 2012;Webb & Zhang, 1997;Wondzell, Diabat, & Haggerty, 2019). Johnson and Jones (2000) found that maximum stream temperatures increased 7°C and high temperatures were seen earlier in the summer when riparian vegeta-

| Riparian vegetation
Riparian vegetation was completely obliterated by the debris flows in Camp Four and Potosi creeks but plant recolonization occurred quickly, as found in other case studies. Following a debris flow in the Oregon Coast Range, total vegetative cover increased twofold to sevenfold within three years (Pabst & Spies, 2001). Red alder and salmonberry (Rubus spectabilis) were the dominant pioneer species in their study, but by the tenth year they found that rapid growth of red alder allowed it to outcompete salmonberry and inhibit conifer establishment. Our findings were similar, although we found salmonberry to be a minor component of the riparian plant community from the beginning. In our study, the maximum height of red alder increased steadily and rapidly over time, achieving heights greater than 4 m in six years after the debris flows. Seedling conifers were observed along the debris flow tracks adjacent to the streams, but the red alder was clearly outcompeting and suppressing the seedlings. Red alder is an aggressive pioneer species due in part to its nitrogen fixation ability, but is a short-lived tree (Harrington, 2006); still, we anticipate it will be the dominant tree in the riparian areas for several decades while conifers slowly re-establish.

| Aquatic insects
Aquatic insect populations are often affected by disturbance-caused changes in food availability or physical condition of stream channels.
Most first-year changes appear to be caused by physical alterations in stream habitat, such as major sediment scouring or deposition, rather than from chemical or thermal changes (Minshall, Royer, & Robinson, 2001). In the first year after a debris flow, benthic insect diversity is typically reduced, with high relative dominance by a few generalist taxa, such as Chironomidae and Baetis (Kiffney et al., 2004;Lamberti et al., 1991;Minshall et al., 2001;Mundahl & Hunt, 2011). These insect taxa are considered to be well adapted to disturbance owing to their short generation times and high dispersal abilities via larval drift and as strong adult fliers (Anderson, 1992;Merritt et al., 2008). In Potosi and Camp Four creeks, the densities of Chironomidae and Baetis were relatively high 6 months after the debris flows, but other insects (e.g., large-bodied mayflies, caddisflies, and stoneflies) took multiple years to recover. Even in Sunbeam Creek, which did not have a debris flow, flooding during the storm affected the benthic insect community. Over a period of four years, all three streams steadily increased in richness, abundance, and community similarity.

Removal of riparian vegetation by debris flows increases solar
insolation and water temperatures, which support high levels of primary production by algae and large populations of grazing consumers (Cover et al., 2010;Kiffney et al., 2004;Lamberti et al., 1991;Snyder & Johnson, 2006). Less than a year after the debris flows on Potosi and Camp Four creeks, insect scrapers that feed on epilithic algae had extremely low relative abundances, but quickly became the second most abundant functional feeding group (FFG) a year later and maintained their abundances for another three years in both the debris-flow-affected streams and similarly in the reference stream. Although not measured, large patches of filamentous green algae and thick mats of diatoms were observed in both debris-flow streams over the study years, but were not observed in Sunbeam Creek. Elevated water temperatures in Camp Four and Potosi creeks were likely favorable to algal community development in the exposed channels.

Riparian vegetation along Sunbeam
Creek the reference stream remained intact and limited sunlight exposure, but stream discharge during the storm was strong enough to mobilize substrate.
Therefore, in all three streams, scrapers appeared to be initially limited by substrate stability, as opposed to light or temperature. Strict adherence to a patterned succession of FFG's has not been observed because physical factors, particularly turbidity, sedimentation, and scouring, have an overriding influence on invertebrate occurrence and most stream invertebrates are not narrow food specialists (Minshall, 2003). In addition, insects drifting into our study reaches from tributaries unaffected by the debris flows could have provided a source for immediate and repeated colonization. Changes in insect communities between our streams over time most likely reflected changes in physical conditions and not a lack of available colonists.
Stream food webs are expected to undergo a shift from autochthonous-to allochthonous-dominated energy sources as the riparian vegetation recovers and a forest canopy shades the stream; however, allochthonous energy pathways may take decades to recover to predisturbance levels (Minshall et al., 2001;Wootton, 2012). In the Klamath Mountains of northern California, large wood, benthic organic matter, and detritivorous stoneflies were all very sparse in streams disturbed by 10-year-old debris flows (Cover et al., 2010). In headwater streams in Japan, the reduced abundance of shredders was attributed to the loss of large wood and channel structure from debris flows more than ten years prior (Kobayashi, Gomi, Sidle, & Takemon, 2010). After five years, we found the riparian vegetation along the debris flow tracks had grown considerably, but based on the high abundances of scraping insects and low abundances of shredders, it appears that these streams were still primarily supported by autochthonous energy sources.

| Coastal tailed frogs
Some annual variation was apparent in both debris-flow streams, but overall tadpole densities were lower in both streams, sometimes significantly lower compared with what they were before the debris flows. Studies of coastal tailed frog tadpoles in streams impacted by scorched tree blowdown following the eruption of Mount St. Helens (Washington, USA) also found densities to vary between sites, but to be relatively abundant four to five years posteruption (Crisafulli, Trippe, Hawkins, & MacMahon, 2005).
The authors suggest that abundant periphyton, from increased light, and microhabitats in the streams provided survival and recolonization opportunities for tailed frogs, whose tadpoles scrape algae from substrate. Riparian areas along Potosi and Camp Four creeks that were not affected by the debris flows likely provided refuge for adult tailed frogs, which were seen in and near the streams the first summer following the debris flows. Coastal tailed frogs are strongly associated with late-seral forests that are typically environments of stable temperatures and substrates (Welsh & Lind, 2002)-both of which were highly altered after the debris flows. For example, the highest 7DADM temperature was 23.4°C in lower Potosi Creek. This temperature was within the incipient lethal temperature range (23.4-24.1°C) in which 50% mortality of larval tailed frogs can occur (Claussen, 1973). Nevertheless, both debris-flow streams experienced many daily maximum temperatures well above 16.0°C, which most likely stressed and affected the recovery of this cool-water stenotherm.

| Signal crayfish
Common in both Potosi and Camp Four creeks before the debris flows, crayfish were nearly absent five years after the debris flows.
Crayfish typically mate in October and eggs are carried under the female's tail until they hatch from late March through June. This species is known to burrow into substrates during unfavorable stream conditions. They mature at two years and usually exhibit 10% to 50% survival from egg to adult (Lewis, 2002). The timing of the December debris flows (potentially wiping out an entire cohort), late maturity, extended egg bearing, and poor dispersal ability may all contribute to this species' vulnerability to debris flows. Burrowing behavior would help keep the species from being dislodged during seasonal and high flows but would offer little protection in a debris flow, in which most stream substrate becomes mobilized. It is unknown whether some crayfish escaped the debris flow by occupying habitats in side tributaries thus potentially providing a source for recolonization. Cover et al. (2010) found that crayfish were common in 40+-year-old debris flows in northern California streams but not found in recent 10-yearold debris flows.

| Coastal cutthroat trout
Differences in the recolonization of cutthroat trout in terms of density and age-class composition between the two debris-flow streams were not expected. After the debris flows, we found trout density to Potosi and Camp Four creeks both exhibited an initial increase in age-0+ cutthroat trout, although the increase was delayed in Camp Four Creek. The temporary increase in young trout was similar to patterns of increased abundance found after a volcanic eruption (Bisson, Crisafulli, Fransen, Lucas, & Hawkins, 2005), forest clear-cutting (Bisson & Sedell, 1984;Hawkins, Murphy, Anderson, & Wilzbach, 1983), and wildfires (Bisson et al., 2003), fitting regional trends of recovery in which early postdisturbance conditions favor survival (Bisson et al., 2009). Similar to our study, the density of age-0+ brook trout exceeded predebris-flow levels within one year, and adult density exceeded predebris-flow levels within two and half years on a debris flow stream in Shenandoah National Park, Virginia (Roghair et al., 2002). However, as we observed on Camp Four Creek, an increase in trout abundance can be delayed until colonists arrive and establish in available habitats. For example, in a debris flow stream in the western Oregon Cascade Range, age-0+ cutthroat trout declined postdisturbance and remained low throughout the first year, yet in the following years densities increased to about double those found upstream from the debris flow (Lamberti et al., 1991). Similarly, in debris-flow streams in northern California, steelhead/rainbow trout (Oncorhynchus mykiss) biomass initially did not differ from nondebris flow streams, but in subsequent years, trout biomass exceeded that in the nondebris flow streams (White & Harvey, 2017). Following a debris flow in a stream on Washington's Olympic Peninsula, Scarlett and Cederholm (1996) recorded a 50% reduction in cutthroat trout density relative to predisturbance levels. They reasoned that in addition to direct mortality caused by the event, a reduction in total wetted stream area and reduced pool depths after the debris flow were influencing recovery. Colonists from side-tributary streams, in addition to fish moving upstream from Waddell Creek, could explain the initial abundance of age-0+ fish in Potosi Creek just 6 months after the disturbance. Some trout may have survived in one or more side tributaries along Potosi Creek, but recolonization of Camp Four Creek likely occurred exclusively upstream from Waddell Creek. Cover et al. (2010) found that trout abundance varied considerably after debris flows in northern California, and showed no consistent recovery pattern; our observations were similar. We found that, at least during initial trout colonization after the debris flows, recovery patterns were site-specific.

| Sculpins
Very little information exists describing sculpin colonization after a debris flow or flood large enough to mobilize the streambed. In Oregon, on two debris flows triggered by a large storm event in 1996, Danehy et al. (2012) found that sculpins repopulated previously occupied areas in about six years. Swanson, Johnson, Gregory, and Acker (1998) found that sculpin densities declined 70% on average from floods caused by the same 1996 storm event in the Cascade Range, Oregon. They also posited recovery times for sculpins to be longer than five years, based on weak dispersal rates and low fecundity. Hawkins and Sedell (1990) found that sculpin

| Upstream fish distribution
The farthest upstream habitat occupied by cutthroat trout in each study stream was a plunge pool below a waterfall caused by either a bedrock ledge, large boulders, or a wood jam. These steps along the longitudinal profile can persist for several decades or longer  depending on the longevity and stability of the boulders or logs, and the interval between large, mobilizing flows. Cover et al. (2010) found that streams with recent debris flows had fewer steps along their profiles. In streams with older debris flows, pools were formed by alluvial steps, whereas in streams with recent debris flows, pools were formed by very large boulders or exposed bedrock ledges. Our observations in this study were similar. In general, the debris flows appeared to even the gradient and eliminate rock steps and large wood jams, thus opening up new habitats for trout colonization.
We found the length of the last upstream trout was always greater than 130 mm, representative of an age-3+ fish. This may be due to size-structured dominance in salmonid populations (Ward, Webster, & Hart, 2006)  suspect that 6%-10% channel slope may be a threshold for sculpin movement in these small, headwater streams.

| APPLI C AB ILIT Y OF FIND ING S TO OTHER ARE A S
Many of the biological responses to debris flows in our study streams are explained by the interaction between the ecological adaptations of pioneering species and the conditions in the debris flow track.
We expect that the pioneer aquatic and riparian species that we observed, common to the Pacific Northwest region, will play analogous roles in the biological responses to debris flows in headwater streams in other areas. However, specific recovery patterns of these species will strongly depend on an array of site-specific conditions, including the characteristics and timing of the debris flow, refuge locations, weather patterns, and persistence of legacy habitats. To reoccupy a stream after a debris flow, pioneer species need source populations and access routes. Debris-flow volume, momentum, and the geometry will shape the longitudinal profile and cross section of the headwater valley, all of which we found can affect recolonization.
One of our key findings was how rapidly stream food webs recovered after the severe disturbance created by the debris flows.
A food web containing multiple trophic levels supported populations of pioneer species approaching predisturbance levels less than a year after most ecosystem structure and function had been almost completely obliterated. We suggest that pioneer species higher in the food web, like trout, are uniquely adapted for colonizing disturbed streams owing to their strong swimming ability, but to persist after arriving, these species need suitable habitat and food availability. Thus, a key to the success of trout and other large predators for re-occupying streams after a severe disturbance is the rapid recovery of the aquatic insects. For example, the rapid growth of Chironomidae and Baetis insect populations in the debris-flow streams in the early phase of recovery was likely due to their affinity to drift (from upstream or tributary source populations), fast reproduction (multivoltine), and abundant food sources (legacy organic matter or new periphyton growth). If these insect generalists and their food sources are available in streams shortly after disturbance, higher trophic species may also recover rapidly assuming local conditions permit accessibility.
Often referenced by such terms as "severe" or "catastrophic," natural disturbances such as debris flows in small watersheds are thought to be important to the long-term productivity and biological diversity of these and downstream ecosystems (Bisson et al., 2009).
These infrequent events are widespread across forested landscapes (Benda, 1990;Benda et al., 2005) and typically create habitat diversity (Bilby, Reeves, & Dolloff, 2003). Long-term impacts of large disturbance events in a given watershed are influenced by the time for recovery between episodes. Some biota and physical habitats take longer to recover than others, and the rate of recovery appears to be site specific. Given the uniqueness of recovery processes, an increase in the frequency of extreme disturbances may have significant long-term implications for stream ecosystems if source populations or refuge habitats are eliminated.
The role of disturbance in Pacific Northwest streams, whether caused by debris flows, floods, wildfires, or volcanism, is an important mechanism influencing the structure of aquatic ecosystems (Beechie & Imaki, 2014;Bisson et al., 2003). Species diversity, life history and spatial diversity, and phenotypic plasticity are mechanisms that allow communities and populations to adapt to variable and changing environments. Broadscale disturbances can regulate debris flow activity across landscapes (Benda, 1990;Benda et al., 2005;May & Gresswell, 2003). Given the management regime of Capitol State Forest where our study occurred, we do not know if the frequency of debris flows will be accelerated, remain at natural background levels, or actually decrease over time. We found that rapidly dispersing aquatic organisms quickly recolonized the debris-flow channels. Other organisms with limited dispersal ability have recovered more slowly; riparian plant communities and aquatic habitats along the streams will likely be altered for decades. Our study continued for only six years following the debris flows, but it is clear that future changes in Potosi and Camp Four creeks can be anticipated, even if we cannot predict exactly what they will be.
At present, the short-term impacts of large disturbances on stream organisms can be reasonably modeled, but long-term consequences of disturbance and subsequent biological recovery will depend on future climate and management activities.

ACK N OWLED G M ENTS
The field work over a period of six years was accomplished by the We appreciate the comments and suggestions of two anonymous reviewers.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
All authors contributed to the conception and design of the work, or the collection, analysis, and interpretation of data.

TA B L E A 1 (Continued)
TA B L E A 3 Mean densities (no./m 2 ) of cutthroat trout, tailed frog tadpoles, signal crayfish, and sculpins averaged for years before the debris flows (2002)(2003)(2004)(2005)(2006) and after the debris flows Note: Bolded t statistics and p-values denote significant before (combined years) versus after (each year separate) differences at p ≤ .05. Asterisk signifies zero captures or capture only occurring in one sample occasion that year; -= no data. July and September surveys, and upper and lower reach data were combined for each year (n = 4