Salmonid fish such as steelhead (anadromous Oncorhynchus mykiss) build spawning nests (redds) using gravel that falls within a narrow size range (Kondolf and Wolman, 1993). Complex interactions among climatic, geomorphic, biochemical, and human processes can either limit or increase the availability of that resource (Scheurer et al., 2009, Miller et al., 2008; Jacobs et al., 2021). The recent increase in wildfire incidents and intensity in western North America (Westerling et al., 2006) provides an opportunity to better understand the influence of fire on fluvial ecosystems (East and Sankey, 2020). In particular, there have been no studies of the role that fire-mediated colluvial processes play in the direct recruitment of spawning gravel from river-bounding hillslopes. Jacobs et al. (2021) and Flitcroft et al. (2016) argue that occasional post-fire debris flows introduce woody debris and sediment to a river system over time, ultimately improving potential redd occurrence and habitat availability. We directly studied how debris flows that were generated by intense rains following wildfire impacted steelhead habitat quality in a reach of riverbed that had been both starved by 95 years of dam impoundment and impacted by a fine bed material sediment pulse following dam removal.

Steelhead are a key part of California’s coastal watershed and marine ecosystems, but the Carmel River (Figure 1) population is threatened and continues to decline (Williams et al., 2016). Williams et al. (2016) reported that the steelhead population in the Carmel River had been inexplicably declining by about 50% per generation (4-years generation) between 1996 and 2015. Two obvious impediments to steelhead recovery on the Carmel River were the 32 m tall, concrete-arch San Clemente Dam and the Los Padres Dam, a large earthen dam located approximately 7 km upstream from San Clemente Dam (Figure 1; Boughton et al., 2016). San Clemente dam had trapped all bedload (>1.67 × 106 m³) during its operating life, including redd-building gravel, from reaching the lower 30 km of the river (Harrison et al., 2018; Smith et al., 2020). In concert with the waning steelhead population, the number of redds counted downstream from the dam had also been declining (Williams et al., 2016), pointing to a paucity of spawning-sized gravel as one potential limiting factor (Boughton et al., 2016).

San Clemente Dam was removed in 2015 with the anticipation that passive restoration of gravel and large wood transport past the dam site would gradually improve downstream spawning conditions (Boughton et al., 2016). A unique aspect of this dam removal was that most of the reservoir sediment was engineered to remain in place, so it was not available to replenish spawning gravel downstream (Harrison et al., 2018). In 2017, incision, widening, and avulsion in a short reach of river perched on unconsolidated reservoir sand and gravel generated a post-dam sediment pulse of sand and small gravel that covered spawning beds and locally filled pools and runs along more than 30 km of channel downstream to the mouth of the river (Harrison et al., 2018). Short-lived, fine-grained sediment pulses, supplied by erosion of relatively fine-grained reservoir sediment, are expected (Doyle et al., 2003; Major et al., 2012; Randle et al., 2015; Wang and Kuo 2016). The two-year gap between dam removal and a significant post-dam sediment pulse on the Carmel River was unusual. Harrison et al. (2018) ascribed the delay to climate variability typical of the regional Mediterranean climate. Their study ended two years after dam removal, before the life span of the sediment pulse could be determined.

Downstream sedimentary and geomorphic response to dam removal can vary considerably in magnitude, style, and duration (Kibler et al., 2011, Tullos et., al, 2014, East et al., 2015; Foley et al., 2017). Numerical modeling and small-scale flume experiments suggest that sediment pulses from a wide range of sources will evolve and decay through some combined proportion of translation and dispersion (Lisle et al., 2001; Sklar et al., 2009). In contrast, a recent landscape-scale study of post-dam sediment pulses hinted that the translation-dispersion wave-behavior ideals are simplistic (Wohl and Cenderelli 2000; East et al., 2018). East et al. (2018) report that the style and duration of sediment pulse behavior relies on a combination of climate, gradient, large wood, pulse magnitude, and grain size. Our work includes watershed-scale descriptions of a post-dam sediment pulse with a focus on impacts to steelhead spawning habitat.

Harrison et al. (2018) conducted a Before-After-Control-Impact (BACI) study to assess the magnitude and duration of steelhead habitat impacts forced by a sediment pulse following San Clemente Dam removal. Using a habitat suitability model modified from Dettman and Hanna (1991), they found that in addition to spawning-gravel impairment along most of the river, gravel deposition immediately downstream of the former dam had locally improved spawning habitat (Harrison et al., 2018). Our work reanalyzes the 2013 to 2017 Harrison et al. (2018) grain size data set (East et al., 2017), and adds an additional four years (2018–2021) to further assess the temporal and spatial substrate impacts generated by the post-San Clemente Dam sediment pulse.

A sustainable source of spawning sized gravel in the Carmel River is required to foster recovery of steelhead in the post-dam era. Gravel of the appropriate size for redd construction was abundant upstream of the relict reservoir material in 2017 (site “CR” of Figure 1; Harrison et al., 2018; East et al., 2017), but that material did not reach the lower river in significant quantity in 2017 despite the potential for bedload transport generated by two 10-year and one 25-year to 30-year flood event in that year (Harrison et al., 2018; Smith et al., 2020). The long-term re-establishment of significant gravel transport might therefore hinge on removal of Los Padres Dam, located 7 km higher in the watershed (Figure 1). It appears that near-term recovery of spawning gravel in the lower Carmel River will rely upon direct recruitment of colluvial material from canyon walls rather than reconnection with upstream sources. Our study focuses on direct recruitment as a potential gravel source.

The upper Carmel River watershed includes pervasively fractured and faulted granodiorite and metamorphic rocks (Smith et al., 2004) that are the ultimate source of bedload and spawning gravel. The colluvial slopes transport gravel to the valley bottom, and sometimes directly to the stream by dry ravel, landslides, and debris flows. Bathymetric data from the Los Padres Reservoir (Figure 1) provide insights into the conditions that foster bedload recruitment in the Carmel River watershed. Precise re-surveys of the bathymetry and surrounding topography document the episodic nature of reservoir capacity loss (Figure 2). The reservoir capacity dropped from 2,709 to 1,030 acre-feet (38%) between construction in 1941 and the most recent survey in 2017 (MEI, 2005; Smith et al., 2009; 2018). Most capacity loss occurred in years when wildfire was followed by a very wet rainfall season (Figure 2). The combined impact of just two fires, the Marble Cone and Soberanes, accounted for 66% of the long-term capacity loss (Hecht 1981; Smith et al., 2018). The need for both fire and intense rain for bedload recruitment is further supported by noting that the expansive Basin-Complex fire did not immediately lead to loss of reservoir capacity, while the similar Soberanes fire preceded a 7% drop in capacity. A wet winter following the Soberanes Fire was the only obvious difference between the two events (Richmond 2009). The bulk of reservoir-filling material came to rest in the reservoir delta top, and virtually no sediment reached the distal, deep basin floor, indicating that the post-Soberanes Fire sediment was chiefly bed load rather than suspended material. Given the highly fractured condition of local bedrock, the bedload recruited from fire-mediated slope failure likely contains abundant gravel appropriate for spawning habitat. Our current study quantifies the amount of spawning gravel present in several fire-related debris flows in the Carmel River watershed and explores their potential negative and positive impacts on spawning habitat.

Nine years of grain size data that span the removal of San Clemente Dam indicate that spawning-size gravel abundance diminished downstream of the dam site in 2017, relative to the control site, and has not yet recovered (Figure 5; Table 1). The downstream sites (impact) lost spawning gravel when the dam was removed (Figure 5; Wilcoxon: W = 12,482, p = 0.002), in contrast to the upstream site (control), which has remained stable (Figure 5; Wilcoxon: W = 221, p = 0.1).

High magnitude flows in 2017 generated a pulse of pool-filling sand and fine-gravel eroded from both unstable reservoir deposits (Harrison et al., 2018), and a river construction project (Smith et al., 2020). Existing spawning gravel was buried by a carpet of finer bed material that impacted all sites in the lower gradient reaches of the river (Figure 6; Harrison et al., 2018). The most recent grain size data (2021) show that the lowermost sites may be recovering as the 2017 sediment pulse washes away to expose the pre-dam removal gravel layer. In contrast, the remaining upstream sites had below average percent spawning gravel in 2021 (Figure 6). If the downstream sites were recovering spawning gravel through additions from upstream sources, we would expect the upstream sites to be showing the recovery first.

The 3 km long reach of river immediately downstream of the former dam (sites DM and SH; Figure 1) were armored by large cobble and boulder bed material at the time that the dam was removed (East et al., 2017). Following dam removal those sites remained generally too coarse for steelhead nests, except for pools and runs, which were filled by sand and fine gravel that was too fine (Chow et al., 2016; Harrison et al., 2018). These impaired conditions lasted through to the 2020 sampling season.

In 2021, we monitored the river valley between those two sites (DM and SH) to assess the impact of post-fire debris flows on spawning gravel abundance. The steep, granitic, chaparral-covered slopes were denuded by the Carmel Fire (Figure 7). At least 17 post-fire slope failure events occurred in the study reach (Table 2). Time-lapse cameras aimed at two basin mouths indicated that the slopes failed sometime between 0315h and 0330h PST on January 27th^, 2021. The debris flows occurred when rain intensity surpassed the 15-minute-duration debris flow threshold of 24 mm/hr, in keeping with forecasts modeled by the United States Geological Survey (Figure 8; USGS 2020; Staley et al., 2017; Gartner et al., 2014).

Rilling and gullying were pervasive on the slopes that generated debris flows. Erosional features are consistent with runoff-dominated flows rather than soil-slip related processes. Runoff-dominated flows are thought to be favored on intensely burned slopes where soil infiltration is impaired (Cannon et al., 2000).

Of the 13 debris flows in the study area, 11 were successfully predicted by the USGS model and 10 had volumes that were within the predicted range (Table 2). Flows 12 and 13 and all rockfall events were not predicted by the USGS models because the source basins fell below the minimum 0.2 km² minimum threshold for inclusion models. Site 11 (Table 2; Figure 4) was the only basin in the study area to fail to produce a flow despite having a published flow probability of 20–40%. The site 11 basin differed from the other sites in having oak woodland rather than chaparral habitat, lower burn intensity, and several abandoned road traces crossing the slope that likely reduced the runoff velocity and increased colluvium stability (Figure 4). Ten of the debris flows we studied entered the flood water of the Carmel River (Figure 9; Table 2), with the potential to impact the substrate. Visual inspection of the debris flow snouts where they had been eroded by the river indicated that the deposit matrices were virtually mud free.

The debris flow snouts had a median of 45% of spawning-sized gravel (Figure 10). Although the debris flows were rich in spawning-sized gravel, and several debris flows reached the river, both reconnaissance observations and a BACI-sampling design of pebble counts in the Carmel River channel show that no significant patches or bars of new gravel were present in the study area when the flood waters receded (Figure 10). Spawning gravel in the Carmel River channel before and after the debris flow event was approximately 12% of the bed, and it was too dispersed to be useful for redds. Particle counts near the debris flow sites (“channel” of Figure 10) included 9% that had been recruited directly from the recent debris flows. Just 3% of the particles located near the debris flows were new particles that also fell within the spawning-size range.

A paucity of spawning gravel may have contributed to the decline of the Carmel population of steelhead. Counter to anticipated results (Boughton et al., 2016), our BACI analysis of percent spawning gravel spanning 9 years indicates that availability of gravel suitable for nest building decreased downstream of the dam site following dam removal (Figure 5). A pulse of pool-filling sand and small gravel swept downstream to the river mouth in 2017. While downstream sites lost gravel to burial by sand, the upstream reach (within 3 km of the dam) has remained generally armored by large cobble and boulders. Apparently, the post-dam sand pulse passed by most of this upper reach. This observation is in keeping with flume experiments where small-scale sediment waves completely translate across a previously-armored substrate (Sklar et al., 2009). The lack of thick deposits in the armored upper reach (outside of pools) is also expected because the gradient is somewhat higher there (1%–0.8%) than at downstream reaches (∼0.2%) (Harrison et al., 2018). In 2021 the percent of spawning gravel was still diminished below pre-dam conditions, but the two sites closest to the mouth showed slight recovery (Figure 6). We are reluctant use “wave” behavior to describe the sediment pulse because wave geometry (thin at ends, thick in middle; Lisle et al., 2001) is not present. Instead, the pulse has a longitudinally-discontinuous geometry, with gaps in some of the riffles, as described by Harrison et al. (2018). As the sediment washes through, it is moving from pool to pool on the way to the sea, behavior noted in another dam related sediment pulse (e.g., Wohl and Cenderelli, 2000). If the downstream two sites are improving, rather than showing natural variability, it reflects the exhumation of spawning gravel through attenuation of the downstream terminus of the sediment pulse. This behavior is also not typical of previously-described post-dam sediment pulses (Cashman et al., 2021). We acknowledge the difficulty in interpreting complex bed dynamics using single annual measurements at widely-spaced, fixed locations on the river.

Although the dam came down in 2015, there was little change downstream of the dam until high flows of 2017 initiated the sand wave that covered or diluted existing spawning gravel. (Figure 5). This lagged response was the result of highly variable rainfall and runoff sequencing (Harrison et al., 2018). Previous studies have determined that post-dam sediment waves can last between 1 and 10 years (Jacobs et al., 2021). A disproportionate number of drought years, when bedload transport is very limited, has extended the San Clemente dam removal impact to six years, and there is little evidence that the sediment pulse is waning except perhaps near the river mouth (Figure 6).

Improvement in spawning gravel abundance was predicted to occur from the transport of new material from upstream sources. The lack of recovery may be the background condition imposed by gravel-trapping behind the Los Padres Dam. If so, spawning gravel will continue to be a limiting factor for steelhead until Los Padres dam is also removed. A temporary increase in spawning gravel noted throughout the study area in 2019 (Figure 5) was likely the result of bedload transport and winnowing of fine particles during two approximately Q₅ flow events (Figure 3).

Reservoir studies in the upper Carmel watershed show that abundant bedload can be recruited if a wet year immediately follows wildfire (Figure 2). This pattern was locally witnessed when the January 27, 2021 rains triggered general failure of colluvial slopes adjacent to the Carmel River immediately below the former San Clemente Dam site. The debris flows contained virtually no mud or ash, so they had little potential to negatively impact the channel benthic habitat (Minshall 2003; Flitcroft et al., 2016; Jacobs 2021). The debris flows were sourced in burned chaparral habitat, so there could be no habitat benefits of new large wood pieces (Flitcroft et al., 2016). Ultimately, the debris flows did not contribute enough new gravel to the 10 m-wide river channel to immediately replenish spawning gravels, although the cumulative volume of deposited material was 2,500 m³ and the grain sizes were suitable (Figures 9, 10; Table 2). The debris flows had to traverse a rural road (Figure 9) and, in some places, a relict 1.5 m diameter pipeline. These impediments slowed the debris flow runout lengths, leading to most material being stored in the canyon bottom rather than extending far into the river. In a more natural landscape, the cumulative benefit of the debris flows might have been much greater. Future high flows will continue to erode the flow snouts, so benefits to local gravel supply might be forthcoming (Jacobs et al., 2021).

The complex impacts of climate change on interdependent human, geomorphic, and ecological systems are being considered and inventoried (East and Sankey, 2020; Jacobs et al., 2021). Regional models of climate change point to two variables that will influence recruitment of spawning gravel from the slopes of the Carmel watershed. First, conditions are forecast to be dryer, leaving more fire fuel available early in the fire season (e.g., Ullrich et al., 2018). Second, annual rain will be delivered in fewer, but more intense, storms (e.g., Swain et al., 2017). Our debris-flow study period was a drought year with one intense storm that followed an intense fire, so it is a direct analog for future climate conditions at the site. Our data suggest that a combination of higher fire frequency and more intense precipitation will increase the rate of spawning gravel recruitment in the Carmel watershed in future decades.