Interacting geomorphic and ecological response of step-pool streams after wildfire

Steps and pools are characteristic bedforms that dominate the channel morphology of steep mountain streams, regulating flow and sediment dynamics feeding into lowland rivers and larger water bodies (Fig. 1). Step-pools are stable features in the landscape, typically mobilized by flows with recurrence intervals exceeding 30–50 years (Grant et al., 1990; Chin, 1998; Lenzi et al., 2006; Molnar et al., 2010). Step-pools are significant in providing energy dissipation in river channels (Chin, 2003). Energy dissipation occurs through spill and form resistance as water flows over and through the large roughness element that comprise steps and plunges into the pools below (Wilcox et al., 2006; Comiti et al., 2009; David et al., 2010; Zimmermann, 2010). By creating a vertical drop in water surface elevation as water flows from step to pool, steps also reduce potential energy that otherwise would convert to a longitudinal component of kinetic energy, used for erosion and sediment transport (Marston, 1982). In this respect, steps counteract the erosive power of steep slopes and prevent excessive erosion and channel degradation (Chin, 1989). Because step clasts offer habitats for sensitive and specialized organisms (Chin et al., 2009a), step-pool streams are also important from an ecological standpoint (Dupuis and Friele, 2006; Wang et al., 2009; O’Dowd and Chin, 2016).

Step-pool channels are increasingly vulnerable to a range of natural and human disturbances. Direct impacts occur as urban development progressively encroaches upon mountain fronts in response to population growth (Chin et al., 2009b; Chin et al., 2017). Possible increases in the intensity and frequency of extreme climatic events, including large floods, wildfire, and droughts, also threaten the health and viability of step-pool ecosystems (Robson et al., 2013). Yet, research has not addressed sufficiently the impact of many such disturbances, and data documenting the response of step-pool streams to such events are rare (but see, for example, Lenzi, 2001; Turowski et al., 2009; Molnar et al., 2010). In the summer of 2012, several catastrophic wildfires spread across the Rocky Mountains of Colorado (USA) and burned numerous watersheds containing step-pool channels. These wildfires were followed in the next summer by an active hydro-meteorological season (Scott, 2013), in which several extreme events occurred. These events provided timely opportunities to document the impacts of wildfire as a major disturbance on step-pool streams, and to develop understanding of the resilience and potential recovery of such systems. As the frequency and magnitude of wildfires increase along with global warming, such understanding becomes increasingly important.

Wildfire induces changes on Earth’s surface in ways that alter hydrologic and geomorphic processes (Moody et al., 2013). When fire occurs, the loss of vegetation decreases organic matter and the cohesion of soils (Woudt, 1959; Letey et al., 1962; Roberts and Carbon, 1971; Shakesby and Doerr, 2006), leading to an increase in soil water repellency (DeBano, 2000) and a reduction in infiltration (Robichaud, 2000). These alterations induce greater runoff and flooding potential (Ebel et al., 2012). The extent of impacts often increases with the severity of burn (Doerr et al., 2006; Moody et al., 2008; Ryan et al., 2011) and rainfall intensity (Moody and Martin, 2001a; Moody et al., 2013). Dry-season flow can remain elevated more than a decade following wildfire (Kinoshita and Hogue, 2015), with flooding potential dampening when vegetation re-grows (Kinoshita and Hogue, 2011). The loss of vegetation from hillslopes also enhances erosion from bare surfaces, with consequent input of sediment into river channels (Moody and Martin, 2009; Silins et al., 2009; Lamb et al., 2011; Florsheim et al., 2016) or debris flows (Cannon et al., 2001). Both bedload (Beaty, 1994) and suspended sediment concentrations (Malmon et al., 2007; Ryan et al., 2011) typically increase by several orders of magnitude after burns. Although sediment concentrations often return toward pre-fire levels within several years (Beaty, 1994), fire-related sediment can remain in the fluvial system for hundreds of years (Moody and Martin, 2001b; Moody, 2017). The increased quantities of sediment supply into and through mountain watersheds, coupled with greater runoff and potential for flash flooding, pose primary risks for human communities downstream (Chin et al., 2016).

With respect to step-pool systems, excessive input of sediment and elevated flow regimes after fire also threaten their hydraulic functioning and ecological health. An increase in sediment supply to a step-pool channel may destabilize the bed by decreasing form roughness (Koll and Dittrich, 2001; Rickenmann, 2001), resulting in channel beds that are more mobile (Recking et al., 2012). The stochastic introduction of bedload onto the channel surface (Saletti et al., 2016), including keystone particles (Church and Zimmermann, 2007), may also create irregularities for sediment to further accumulate (Curran, 2007). Sediment filling in pools additionally reduces the effectiveness of kinetic energy dissipation that occurs via turbulent mixing (Whittaker and Davies, 1982). Decreased depth in pools from sediment accumulation further reduces potential energy dissipation by step-pools, as the staircase-like profile of the step-pool streambed becomes less pronounced (Chin, 2003). Thus, early flume studies by Hayward (1980) documented higher velocity of water through pools filled with gravel than those devoid of gravel. Increased flooding can therefore dislodge the steps and natural dams that form pools, and remove high-quality aquatic habitats (Beylich and Sandberg, 2005; Fuller et al., 2011). Such alterations to channel morphology, coupled with excessive sediment, often contribute to post-fire devastation of aquatic communities (Gresswell, 1999). Typical changes include reductions in abundance, diversity, and sensitive organisms (Rinne, 1996; Minshall et al., 1997; Vieira et al., 2004; Bêche et al., 2005; Oliver et al., 2012), including benthic macroinvertebrates (Hall and Lombardozzi, 2008; Mellon et al., 2008), until vegetation re-grows and sediment sources decrease (Robichaud, 2000). Because clasts in steps are ecologically significant in attracting sensitive and specialized organisms (Scheuerlein, 1999; O’Dowd and Chin, 2016), step-pool streams are particularly vulnerable to diminishing ecological quality from the impacts of post-fire flooding and sedimentation.

This paper reports a field campaign of 2012–2014 to tackle fundamental questions regarding the initial biogeomorphic response of step-pool streams following the Waldo Canyon Fire in Colorado, USA. Several key questions guided this research. First, how susceptible is the step-pool morphology to changing under post-fire regimes, where elevated runoff and excessive sediment input are expected to occur? In other words, do step-pool channels tend to become unstable after wildfire, and thereby diminish their fundamental roles in the river system? Second, how do post-fire ecological conditions change in conjunction with morphological alterations in step-pool systems? Although previous studies have addressed post-fire ecological impacts, up to now, few have considered such responses together with geomorphic changes following post-fire storms, particularly in the context of step-pool systems. The analyses tested the central hypothesis that the extent of morphological change in step-pool systems varies with the severity of burn, which in turn influences ecological responses. Exploration of this hypothesis therefore led to a third question: What primary factors influence the interacting morphological and ecological responses?

The Waldo Canyon Fire burned a portion of the Pike National Forest along the eastern slope of the Rocky Mountains in Colorado. The Pikes Peak batholith dominates the geology of the study area, along with some sedimentary and metamorphic rocks (Stoeser et al., 2005). Because the Pikes Peak granite is highly conducive to weathering, a layer of coarse scree commonly overlies the rock. The shallow depth of soils (U.S. Geological Survey [USGS] Hydrologic Soil Group D; Natural Resources Conservation Service, U.S. Department of Agriculture, Web Soil Survey, Hydrologic Soil Group—Summary by Map Unit—El Paso County Area, Colorado; and Pike National Forest, Eastern Part, Colorado, Parts of Douglas, El Paso, Jefferson, and Teller Counties: https://websoilsurvey.sc.egov.usda.gov/ [accessed April 2018]) leads to low infiltration characteristics with high runoff potential. Streamflow peaks seasonally with snowmelt in spring and thunderstorms in summer, which also causes flash floods in the area (National Water Information System, USGS surface-water daily data for the nation, https://nwis.waterdata.usgs.gov/nwis/dv [accessed April 2018]). Precipitation, temperature, and vegetation vary along elevational gradients. With an average range of 500–650 mm (∼20–25 inches) of precipitation annually, the area supports a ponderosa forest, along with mixed conifer and montane shrubland and lower montane-foothills shrubland (Colorado Natural Heritage Program, 2011).

The Waldo Canyon Fire began on 23 June 2012 and burned 74 km² of land over 17 days. Approximately 19% of the burn was classified as high severity, 40% moderate severity, and 41% low severity (Young et al., 2012). The size and severity of the Waldo Canyon Fire were modest compared to other fires in the history of the state of Colorado. Yet, the Waldo Canyon Fire was among the most expensive fires in Colorado, primarily because it caused damages to homes in close proximity (∼8 km) to the major city of Colorado Springs and other communities. Dry conditions and erratic winds exceeding 100 km/h caused the fire to spread across a topographic ridge toward the Mountain Shadow neighborhood in the northeast portion of Colorado Springs (Chin et al., 2016). The fire destroyed 346 homes and killed an elderly couple. It also forced the evacuation of more than 22,000 residents on 26 June 2012 during a two-hour period (City of Colorado Springs, 2013). Due to its diverse and significant costs—in firefighting, damages from homes, loss of businesses, as well as costs in restoring utilities and mitigating post-fire effects—the Waldo Canyon Fire was the most expensive fire in the history of the state at the time (Wineke, 2012). More details of the case of the Waldo Canyon Fire are found in Chin et al. (2016) and Kinoshita et al. (2016).

The general approach to the study was to compare, over the study period, the character and behavior of a range of step-pool channel reaches burned by the Waldo Canyon Fire with reference reaches outside the burn (Fig. 2). Concomitantly, we evaluated the ecological conditions of the channel reaches along with post-fire morphological change. Thus, after the fire, before significant changes occurred, we measured morphological and ecological conditions at selected study sites. Then, at successive intervals after storm events, when we expect potentially substantial changes in the study reaches, we repeated the measurements to record post-fire responses in the fluvial system. We used ground surveys and light detection and ranging (LiDAR) terrestrial laser scanning (TLS) to record topographical and geomorphological characteristics, and the collection and analysis of benthic macroinvertebrates to document stream ecological conditions. We detail these methods in subsequent sections.

We focused on the burned watersheds of Camp Creek and Williams Canyon (Fig. 2), where access to some of the burned areas was possible through permission by the U.S. Forest Service and private landowners. We selected study reaches for morphological comparison according to similarity in drainage area, slope, geology, soils, and vegetation. For documenting the response of benthic macroinvertebrates in a subset of these study reaches, comparable elevation was also important. We selected mainly small, steep channels where well developed step-pool sequences commonly occur (Table 1), and where changes in morphological and ecological functioning would greatly affect the channels downstream. The study and reference reaches were greater than 0.03 in slope, and all but two reaches (one study reach matched with one reference reach) had drainage areas less than 4 km². Initially, we selected an additional study reach in the main channel of Camp Creek with slightly larger drainage area (matched with the larger reference channel). Construction of a tall fence to contain debris near the mouth of Camp Creek by the landowners (Chin et al., 2016), however, disturbed this study reach, and thus it is not included in this paper (Table 1).

This article therefore focuses on ten reaches: seven study reaches in the burned area representing various severities of burn, and three reference unburned channel reaches (Table 1 and Fig. 2). Each reach was 30–50 m in length (∼20 times channel width) and contained well-developed step-pool sequences before post-fire changes. For comparative purposes, we grouped study channels into three broad categories according to the extent affected by high, moderate, and low severities of burn (Table 1). We assigned study channels into a category of high burn if >80% of upstream drainage areas were burned with high and moderate severity. The moderate burn group represents study reaches with 60%–80% upstream areas burned with moderate and high severity. The low burn category comprises study reaches with <60% upstream areas burned with moderate and high severity. Areas burned largely by moderate and high severity suffer long-term damage to soils that include destruction of roots up to 10 cm (Young et al., 2012), thereby increasing soil water repellency and risk of flooding and erosion (e.g., Scott and Van Wyk, 1990; Scott and Schulze, 1992; Scott, 1993; Soto and Diaz-Fierros, 1998; Shakesby and Doerr, 2006). Therefore, we expected that the magnitude of post-fire impacts and responses would vary according to the groupings of burn severity. Access to severely burned areas limited the number of study sites in the highly affected category (Table 1). Yet, the distribution of study reaches across degrees of burn mirrored the overall character of the burn. In other words, over 80% of the burn was low or moderate severity (Young et al., 2012). Additionally, each type of measurement (ground survey, benthic macroinvertebrate sampling, and LiDAR terrestrial laser scanning; Table 1) was not always possible at every study reach and at each time step due to the large areal spread of the study sites and the labor-intensive nature of our data collection. Nevertheless, the intensive field campaign produced a large data set for analysis, interpretation, and integration using methods described in the following sections.

To track changes in the morphological character of step-pool channels, standard surveying techniques recorded in detail the thalweg longitudinal profiles of study and reference reaches in the field. A level and rod provided data for bed elevation as a function of distance along the stream channel. We recorded elevation at every break in slope, which included the step crest, the base of each step, and the deepest point in the pool in between steps, identified in the field. Additionally, measurement of the long b-axis of the five largest rocks at each step provided representative particle sizes (Chin, 1999). Averaging the b-axis of these rocks provided an estimate of D₈₄ or D₉₀ (the 84th or 90th percentile, respectively) for the coarse step particles (Costa, 1983), which must be mobilized for de-stabilization of the step-pool channel. Finally, we established at least two cross sections at each reach for repeat surveys, with rebar anchoring each end of the cross section.

To characterize quantitatively the step-pool sequences within each reach, we first identified steps during the field survey. Steps are cobbles, boulders, and rock that span the channel width and cause a drop in water surface elevation during low flow, congruent with Comiti and Mao (2012). Because quantification of changes between one time step to the next was most important, a single observer identified step-pool units through the study period. This approach has yielded reliable results (Wooldridge and Hickin, 2002; Weichert et al., 2008) and allows consistency with previous field studies by the authors (e.g., Chin, 1989; Chin, 1999; Chin and Phillips, 2007; Chin et al., 2009a; Florsheim et al., 2017). Because algorithms for extracting steps from longitudinal profiles are also available (Milzow et al., 2006; Zimmermann et al., 2008), we compared the step-pools identified in the field in two study reaches with those computed with a method according to Zimmermann et al. (2008). This comparison yielded strong correspondence—in Willis Reach and Eagle Reach, only one step (out of 26) identified in our field survey did not qualify as a step according to the algorithm.

From the plotted longitudinal profiles, we then extracted values to represent the main properties of the step-pool channel: number of steps, average step length (L), and average step height (H). In this study, step length is the distance from crest to crest parallel to the channel bed slope; height is the change in elevation from the crest of the step to the bottom of the downstream pool, representing a loss of head (Abrahams et al., 1995). Values of L and H enabled calculation of the average steepness of step-pools (H/L) within study reaches. To test for statistical differences in these morphological characteristics, we applied the Kolmogorov-Smirnov two sample-test (0.1 significance level). This test assessed changes in the distributions of H, L, and H/L between time steps at each reach (e.g., Molnar et al., 2010; Keller et al., 2015).

We used the relative steepness parameter (; H/L normalized by channel slope) as a simple expression of the degree of step-pool development in a given channel reach. In other words, it represents the extent to which the observed step-pool morphology approximates an idealized sequence. A perfect staircase-like structure with regularly spaced steps and no scouring in pools has a value of one for . Over time, as flow scours pools to create exaggerated reverse gradients for step treads (the length component of the channel between steps), the steepness of the step-pool bedform (H/L) becomes large relative to the channel slope, resulting in higher values of that exceed one. Abrahams et al. (1995) first proposed that an ideal step-pool geometry has values of between one and two, thought to offer maximum flow resistance and therefore stability for the channel reach. Many field measurements have yielded the range of values for in natural step-pool channels, with higher values exceeding two commonly reported for channel reaches with gradients below 5–7 percent (e.g., Wohl et al., 1997; Chartrand and Whiting, 2000; Duckson and Duckson, 2001; Zimmermann and Church, 2001; Lenzi, 2002; MacFarlane and Wohl, 2003; Wohl and Wilcox, 2005; Chin and Phillips, 2007; Chin et al., 2009b). Although recent studies have questioned the physical significance of the relative steepness parameter (e.g., Church and Zimmermann, 2007; Molnar et al., 2010), demonstrating its limitations in representing the grain-size distribution and other features of the step-pool channel, it offers a single descriptive index useful to distinguish the overall morphological character of the study channels at reach scales.

Light detection and ranging (LiDAR) terrestrial laser scanning (TLS) provided a novel tool to capture detailed topographic changes in the channel character of five of the study reaches. These channel reaches range from high to low burn severity (high: Tributary; moderate: Aussie and Willis; low: Eagle and Gowe). TLS is particularly useful to record in detail the topography of post-fire landscapes because of the absence of vegetation (e.g., Staley et al., 2014; Wester et al., 2014; Orem and Pelletier, 2015; Rengers et al., 2016; Florsheim et al., 2017). We used the approach to quantify volumetric changes in erosion and deposition of sediment in step-pool channels caused by post-wildfire alterations to sediment sources and hydrology in the burned watersheds.

A Riegl VZ-400 scanner, operating in a near-infrared wavelength (900–1000 nm), generated scans at a rate of up to 122,000 measurements per second. Three to six images (or scan positions) captured the topographic details of each selected study reach. A control survey of tie points in Universal Transverse Mercator zone 13N registered the point cloud data within each reach from each scan position. Tie points are fixed control points identified by retro-reflective fixed targets; they provide accurate local measurements confirmed with global positioning system coordinates. After registration, the scanned images were then combined to represent the study reach.

To process the raw TLS data, an octree method filtered the point clouds. This method recursively subdivided the point cloud into an octree (cell size 0.1 m) and averaged data within each cell, resulting in a homogenous point cloud. We then filtered the data with removal of vegetation from the point clouds to estimate a bare ground surface. Finally, triangulation with the point clouds created a bare Earth digital elevation model (DEM) for the active channel reach. We compared successive DEMs for study reaches using a DEM of difference method (DOD; Lane et al., 2003; Milan et al., 2007; Heritage et al., 2009; Wheaton et al., 2010) in RiSCAN Pro 2.0.3 to calculate volumetric changes. Using the cloud to mesh method to create a distance difference image (Lague et al., 2013) of the active channel, the DOD images provided estimates of differences in elevation between TLS intervals, and thereby reach-scale changes quantified as volumes of post-fire erosion and deposition. We quantified uncertainty inherent in TLS analyses (Wheaton et al., 2010) following methods described in detail in Florsheim et al. (2017). Error (reported in meters) represented a combination of LiDAR instrument inaccuracy, registration inaccuracy, point coverage gaps, the octree method of filtering, topographic filtering that removes irregular topography, and failure to remove all non-bare Earth points.

Documenting ecological responses to geomorphic changes in the study channels focused on benthic macroinvertebrates. Present in almost all stream environments, these organisms indicate stream health related to known tolerances to pollution and disturbance (Cairns and Pratt 1993; Merritt et al., 2008). Following standard protocols (Barbour et al., 1999; Merritt et al., 2008), we took samples in the steps and pools of two cross sections in each study and reference reach at each time step, except in one instance where flow was insufficient for sampling. A 500 µm mesh D-frame net collected samples in pools, whereas a flexible-frame 500 µm mesh net provided a seal between the net and rocks of the step (Chin et al., 2009a). After collection over one-minute periods while disturbing the pool bed and agitating steps to dislodge organisms, we rinsed each sample in a 500 µm sieve before storing in 90% ethanol. We also recorded data for environmental conditions during sampling, including the temperature, pH, and dissolved oxygen of the water. In the laboratory, using a dissecting microscope, we sorted and identified macroinvertebrates to the family level following Merritt et al. (2008), and non-insects to higher taxonomic levels (subclass, class, or phylum).

The benthic macroinvertebrate analysis focused on the key metrics of taxa richness and percentage of organisms belonging to the insect orders Ephemeroptera, Plecoptera, and Trichoptera (EPT). Taxa richness, or the number of different taxa present in a sample, is an indicator of biological diversity in a macroinvertebrate community (Boulton, 2003; Cole et al., 2003; Merritt et al., 2008). High taxa richness indicates a stable community with a broad range of available niches. We calculated richness based on subsamples of 300 organisms if more than 300 organisms were present. Percentage EPT indicates the proportion of organisms belonging to three insect orders that are generally considered sensitive to pollution (Barbour et al., 1999). Higher percentage EPT therefore indicates generally healthier ecological conditions. Because the families Baetidae and Hydropsychidae (in the orders Ephemeroptera and Trichoptera, respectively) contain tolerant species (Barbour et al., 1999), we also included a metric of percentage EPT excluding Baetidae and Hydropsychidae. Solverson (2015) contains detailed analyses of additional metrics for benthic macroinvertebrates following the Waldo Canyon Fire.

To test for statistical differences in these ecological metrics, we applied repeated measures analysis of variance, using logarithmically-transformed data where appropriate. This test assessed changes in each metric at each reach over multiple time steps, treating samples collected from each step or pool as related samples over time (Park et al., 2009). In other words, we compared the characteristics of benthic macroinvertebrates at each site over time. For cases where the data did not meet the assumptions of normality or equal variance, the nonparametric Skillings-Mack test with Monte Carlo simulation was used (Chatfield and Mander, 2009). Where these procedures yielded significant differences, we further applied post-hoc multiple comparison tests (Holm-Bonferroni and Friedman, respectively) to determine which time steps differed.

We used ordination to explore how ecological and morphological responses varied with a range of possible factors. Ordination allowed exploration of multiple parameters simultaneously by representing taxonomic similarity within ecological communities in two or three dimensions (Zuur et al., 2007; Sueyoshi et al., 2014). Moreover, ordination allowed identification of correlations among variables that may have influenced the resulting benthic macroinvertebrate taxonomic composition. Specifically, we used non-metric multidimensional scaling (NMS) to assess how relative percentages of benthic macroinvertebrate taxa varied with four general factors: (1) changes in step-pool sequences; (2) characteristics of the post-fire channel; (3) severity of burn; and (4) rainfall characteristics. To represent these factors, we included the following variables in the ordination analysis: step-pool sequences (height H, length L, H/L, , step particle size); post-fire channel (drainage area, water temperature, pH, dissolved oxygen, cumulative change in thalweg elevation); severity of burn (percentage of upstream drainage area burned with high severity); and rainfall (total depth, total duration, maximum recurrence interval, maximum five minute peak intensity, maximum average intensity). We also included data totaling cumulative change in all of the morphological and channel characteristics over the period of study. To account for different numbers of days in between measurements, we normalized these data with the number of days since the first measurement. All together, the ordination procedure analyzed 22 parameters, which were all derived from direct field measurements as described in Sections 2.3, 2.5, and 2.7. In general, as we hypothesized the step-pool morphology to change after fire with the severity of burn, affecting ecological conditions, we would expect ordination analysis to correlate benthic macroinvertebrate samples with severity of burn and the character of the post-fire step-pool channel, thereby revealing the control factors for the interacting responses.

Existing USGS rain gauges provided precipitation data for the major storm events following the Waldo Canyon Fire (Fig. 2). The rain gauges at Upper Queens Met Station NR Ormes Peak, CO in Camp Creek and Upper Williams Canyon Met Abv Manitou, CO in Williams Canyon, respectively, recorded rainfall near the burned study reaches. Similarly, the rain gauges at Bear Creek near Colorado Springs, CO and West Monument Creek at Air Force Academy, CO measured precipitation for the areas near the reference reaches. Additionally, to fill in gaps, we installed four tipping bucket rain gauges along the crest line between Camp Creek and Williams Canyon, in conjunction with the USGS National Research Program. Table 2 includes data from the gauge named “UCD-USGS-2,” closest to the study reaches reported (Fig. 2). In the absence of actual field data for runoff and sediment transport in study channels, data from these gauges provided a proxy for the exceedance threshold discharges that mobilize coarse sediment in step-pool channels (Wohl and Jaeger, 2009).

During the first post-fire summer storm season of 2013, four major storm events occurred (Table 2), with the storm of 1 July representing the first geomorphologically significant event after the fire. The storms of 1 July, 10 July, and 9 August 2013 had recurrence intervals largely under two years (see shaded columns in Table 2), though the gauge in Upper Williams recorded recurrence intervals of 5–10 years for the 1 July and 9 August events. The storm of 11–12 September 2013, however, was an extreme event with duration exceeding two days (Scott, 2013). The gauges near the burned study reaches recorded rainfall with recurrence intervals of 100–200 years, whereas those near the unburned reference reaches registered rainfall with recurrence intervals ranging from 50 to 1000 years.

In the summer of 2014, the second post-fire storm season, five storms occurred: 16 July, 24 July, 29 July, 14 August, and 9 October (Table 2). These storms generated precipitation with recurrence intervals largely under two years near the burned study reaches. The gauge at Bear Creek near an unburned reference reach recorded rainfall for the 9 October event with a recurrence interval of 10 years. This moderate event was the largest of the summer, unlike the previous year with the extreme rainfall.

The sections below outline the response of step-pool channels to the post-fire storms (Table 2; Section 2.7), grouped by severity of burn. Following results for unburned reference reaches and study reaches burned by low severity (Sections 3.1 and 3.2), for which we documented minor or no response, we present data for highly burned channels for contrast (Section 3.3), for which we recorded major responses. We then follow with results for moderately burned channels (Section 3.4), in which the responses were variable in comparison. We augment these results with data from TLS (Section 3.5), which quantified volumetric changes associated with erosion and deposition of sediment in burned study reaches. Section 4 presents results of the ordination analysis that identified the correlating factors (Section 4.1) influencing the morphological and ecological responses. A series of graphs (Section 4.2) synthesize the relationships among post-fire precipitation, morphological change, ecological responses, and severity of burn.

The three unburned reference reaches (Table 1)—Academy Reach, Hunter Reach, Gage Reach—did not change substantially in their step-pool morphology through the first two post-fire storm seasons that included the extreme event of 11–12 September 2013. The longitudinal profiles and cross sections (Fig. 3) of Academy Reach and Hunter Reach, in particular, were nearly identical through the three surveys. Minor shifting in the step-pool profile occurred within Gage Reach, including erosion of 0.22 m documented in a cross section measured after the first season. Nevertheless, despite the extreme event of September 2013 that recorded rainfall reaching 1000 years in recurrence interval (Table 2), the repeat surveys showed that the unburned step-pool sequences remained stable, true to their form.

A summary of the morphological characteristics of unburned reference reaches highlights this stability (Table 3; see also Supplemental Fig. 1 for presentation in figure format¹). Shortly after the fire (2012), and before significant post-fire geomorphological changes occurred, the three reference reaches exhibited a step-pool morphology that approximates ideal sequences (Abrahams et al., 1995; Section 2.3). Scouring in pools and reverse-gradient step treads resulted in ratios of 1.27, 1.56, and 2.41, respectively, for Academy Reach, Hunter Reach, and Gage Reach. Because the step-pool profiles remained intact through the two post-fire storm seasons in 2013 and 2014 (Fig. 3), the characteristics of the step-pool sequences—including the number of steps, average step height H, average step length L, and ratios—remain largely unchanged (Table 3; Supplemental Fig. 1; see footnote 1). The Kolmogorov-Smirnov two-sample test identified no significant difference in these characteristics through the post-burn storm seasons.

For the benthic macroinvertebrates found in steps and pools, characteristics of stability were also largely evident in the unburned reaches over the two post-fire storm seasons. The only statistical difference in the key ecological metrics after the fire was percentage EPT excluding Baetidae and Hydropsychidae in Hunter Reach between December 2013 and December 2014 (Fig. 3; p = 0.0112). Taxa richness remained similar, for example, in the three reaches, although percentage EPT and percentage EPT excluding Baetidae and Hydropsychidae showed minor increases in Academy Reach and Hunter Reach.

For reaches affected largely by low-severity burn (Table 1), the step-pool channel also remained intact with similar biophysical characteristics through the post-fire study storm seasons. Eagle Reach in Camp Creek, in particular, showed little change in its longitudinal profile and cross sections (Fig. 4). Meadow Reach in Camp Creek also exhibited minor changes, with accretion of 0.23 m in the cross section at 7.1 m (Fig. 4). Several plunge pools of up to 0.58 m developed in Gowe Reach (Camp Creek) downstream of large rocks in 2014, accompanied by minor changes in bed elevation along the profile. The overall morphological character of Gowe Reach, nevertheless, remained intact during the two post-fire seasons. The Kolmogorov-Smirnox two-sample test of the characteristics of the individual steps and pools within the reaches affected by low-severity burn detected significant differences only in step length L in Gowe Reach between 2012 and 2014 (Table 4; Supplemental Fig. 2 [footnote 1]; p = 0.0489), though minor changes in the number and characteristics of individual steps are evident throughout.

Similarly, the benthic macroinvertebrate assemblages exhibited only minor changes in the key metrics following post-fire storms. Statistical analysis detected no differences in these characteristics through the time steps studied (Fig. 4). For example, in Eagle Reach and Meadow Reach, for which we have samples of benthic macroinvertebrates (Table 1), taxa richness increased incrementally through the 2014 storm season, reaching a mean of 18 taxa in the steps and pools of Eagle Reach. Percentage EPT also increased incrementally in Eagle Reach and Meadow Reach (mean percentage EPT rose from 20.4% to 54.2% in Meadow Reach from 2012 to 2014), though Baetidae and Hydropsychidae (tolerant taxa) accounted for some of this increase. The mean percent EPT excluding these two tolerant taxa rose slightly from a mean of 20.2% to 25.9% of organisms in Meadow Reach from 2012 to 2014.

At the other end of the spectrum of burn severity, the first comparatively minor post-fire storms of 1 July and 10 July 2013 (Table 2) began to transform the channel morphology of highly burned Tributary Reach in Williams Canyon (Fig. 2). These storms, which occurred before the extreme event of 11–12 September, had recurrence intervals up to 5–10 years near the study reaches (Table 2). The number of steps decreased by 41.7% following these initial storms and continued to decrease through our final field campaign in 2014 (Fig. 5), as erosion of the channel banks and bed exposed bedrock along the channel and gave way to rock steps. Consequently, the morphological characteristics of the step-pool sequences in Tributary Reach varied greatly after the fire. The standard deviation of L and H/L increased by 183.4% and 100%, respectively, after the first post fire storm (Table 5; Supplemental Fig. 3; footnote 1). The Kolmogorov-Smirnov two-sample test detected differences over the time steps in step height H (p = 0.0741, p = 0.0695), length L (p = 0.0916, p = 0.0352, p = 0.0949, p = 0.0996), and height/length H/L (p = 0.0537, p = 0.0634, p = 0.0057) (Table 5; Supplemental Fig. 3), unlike those that remained similar through the same storms in the unburned and low-severity burned channels (Tables 3 and 4).

After the large 11–12 September 2013 storm, which had a recurrence interval of 100–200 years in upper Williams Canyon (Table 2), Tributary Reach had largely transformed into a bedrock channel (Figs. 5 and 6). The progressive downcutting evident in the repeated surveys reached a maximum of more than one meter (Fig. 6; see cross sections at 17.1 m and 26.1 m). Importantly, the average size of the step clasts in Tributary Reach before post-fire changes were not significantly different than those in the unburned reference reaches (for example, p = 0.6696 for Hunter Reach), despite the different channel responses.

Ecological conditions in Tributary Reach affected by high-severity burn reflect the highly altered post-fire channel morphology (Fig. 6). Although data for benthic macroinvertebrates were not available from 2012 (due to flow that was too low for sampling), the mean taxa richness of 8.25 in 2013 was low compared to those of reference and low-severity burned study channels (Figs. 3 and 4), increasing slightly to 10 taxa in 2014. The similarly low % EPT (zero in 2013 and 2.6% in 2014) over the post-fire storm seasons reflected the highly impacted character of the ecosystem in Tributary Reach. Statistical tests did not detect differences in these characteristics between 2013 and 2014, suggesting that the ecosystem had not begun to recover.

In the intermediate category of burn severity (Table 1), channels exhibited variable geomorphic responses. On the one hand, Aussie Reach in Camp Creek showed only minor changes in its longitudinal profile and cross sections after the fire (Fig. 7). The step-pool channel remained intact, including the number and characteristics of individual step-pool sequences, which exhibited less than 30% change in the parameters studied (Table 6; Supplemental Fig. 4; footnote 1). The Kolmogorov-Smirnov two-sample test showed no differences in the distributions of the height H, length L, and H/L of steps (Table 6; Supplemental Fig. 4) through the time steps. New Upper Reach and Willis Reach, in contrast, experienced greater impacts characterized by initial sedimentation in the channels (Fig. 7). Fine sediment quickly buried the individual step-pool sequences as the channel reaches transformed into plane beds. The step-pools re-emerged with subsequent flows, which was particularly evident in Willis Reach. The morphological characteristics of the step-pool sequences in these reaches therefore changed substantially, with over 200% change documented in mean L, mean H/L, and ratios for some time steps (Table 6; Supplemental Fig. 4, see footnote 1). Statistical tests also detected significant differences in H in New Upper Reach (p = 0.0942) and in L (p = 0.0684) and H/L (p = 0.0684) in Willis Reach (Table 6; Supplemental Fig. 4, see footnote 1), as variability in L increased by over 600%.

The response of benthic macroinvertebrates in Aussie Reach and Willis Reach, for which data are available (Table 1), also varied where channels were affected by moderate burn severity. Results of statistical analysis showed differences in the percentage EPT in both Aussie Reach and Willis Reach (Fig. 7). These differences resulted from large increases in percentage EPT in 2014. For percentage EPT excluding Baetidae and Hydropsychidae, Aussie Reach exhibited significant differences across the time steps from 2013 to 2014 (p = 0.064, p = 0.083), reflecting increases in these sensitive organisms over time. For Willis Reach, statistical tests did not detect differences across the time steps, similar to the situation at Tributary Reach affected by high-severity burn. The absence of EPT excluding Baetidae and Hydropsychidae suggests that the impacted ecosystem had not yet begun to recover. Neither study reach exhibited significant differences in taxa richness over time.

TLS data corroborate the sequence of events and changing conditions quantified through ground surveys and analysis of benthic macroinvertebrates (Figs. 3–7). In severely burned Tributary Reach (Fig. 8A), the progressive downcutting resulted in 193.0 m³ of sediment eroded from the study reach after the major storm of September 2013, compared to 12.5 m³ of deposited sediment. The repeat scan the following year (16 September 2014) documented an additional loss of 158.8 m³ of sediment from the study reach along with 80.0 m³ of deposition. Comparing the first scan (21 April 2013) with the third scan (16 September 2014) yielded a net erosion of 265.0 m³ compared to 9.8 m³ of net deposition over two post-fire years.

These results contrast with those of channels burned with low severity (Fig. 8B) that experienced minor net changes in channel bed topography. In Eagle Reach, for example, the minor net erosion and deposition through two post-fire storm seasons (5 May 2013–17 Sep 2014) were 1.4 m³ and 1.9 m³ of sediment, respectively. In Gowe Reach, net erosion of 16.1 m³ of sediment over two post-fire years (3 May 2013–3 Oct 2014) occurred primarily at a few plunge pools, evident in the longitudinal profiles and cross sections (Fig. 4), along with negligible net deposition of 3.1 m³.

Results of the TLS analysis are also consistent with the variable findings for moderately burned channels (Fig. 8C). Whereas Aussie Reach showed minor topographic changes over two post-fire storm seasons (5.0 m³ erosion and 6.9 m³ deposition from 4 May 2013–3 Oct 2014), TLS documented variable quantities of erosion and deposition in Willis Reach over the same time period. After the first post-fire storm season (Sep 2013), deposition of 91.5 m³ of sediment occurred along with 34.0 m³ of erosion in Willis Reach. In the second post-fire year (14 Sep 2013–16 Sep 2014), 59.3 m³ of sediment was eroded along with 10.7 m³ of deposition.

The spatial arrangement of data representing the study sites plotted in the NMS ordination displays patterns regarding the effects of fire severity on benthic macroinvertebrate communities as related to geomorphic characteristics (Figs. 9A and 9B). First, the cluster of low-severity and unburned sites (blue and green symbols in Fig. 9A) on the right side of the ordination plot indicates that similar benthic macroinvertebrate taxa were present in unburned and low-severity burned sites. Second, a clear spatial separation of plotted data was apparent between high-severity burned sites (Tributary Reach, indicated by red symbols in Fig. 9A) on the left side of the ordination plot and low-severity/unburned sites on the right side of the ordination plot, which indicates benthic macroinvertebrate compositional dissimilarity. That is, the taxonomic composition of benthic macroinvertebrates found in severely burned sites is different from those in unburned and low-severity burned reaches. This separation of data from high-severity and low-severity burned/unburned sites indicates that the horizontal Axis 1 in the ordination plot may represent a gradient from high to low burn severity from left to right. For example, the highly tolerant family Ephidridae of the order Diptera was among the strongest correlations with Axis 1 and occurred more often at sites that were burned with high and moderate severities, clustered on the left side of the plot. Lastly, the wide spatial spread of moderately burned sites (indicated by orange symbols in Fig. 9A) across Axis 1 indicates that the moderate burn severity category included sites with benthic macroinvertebrate composition similar to both high- and low-severity burned sites, as well as unburned sites.

Ordination analysis also revealed correlations among benthic macroinvertebrate composition and each of the four general factors explored (described in Section 2.6): changes in step-pool sequences, characteristics of the post-fire channel, severity of burn, and rainfall characteristics. Horizontal vectors (purple lines) in Figure 9A, for example, indicate that biological communities sampled at sites burned by high severity were associated with higher values of step length (L; r = 0.56), cumulative change in step height/days (Ch_H/d; r = 0.72), cumulative change in thalweg elevation/days (Ch_Elev/d; r = 0.75), percentage of upstream area burned with high severity (%_Hi; r = 0.60), and the maximum average rainfall intensity (Max_Avg_I; r = 0.59). In other words, of the 22 parameters analyzed, these variables resulted in the strongest correlations with benthic macroinvertebrates. These correlations suggest that benthic macroinvertebrate composition at a site varies according to the extent to which post-fire effects induced morphologic changes in the step-pool channel, particularly step height and thalweg elevation with the highest correlations. The vertical Axis 2—aligned with the vector representing drainage area (Dr_Area; r = 0.59)—suggests that differences in the composition of macroinvertebrate taxa in unburned channels (blue symbols; which vary vertically in Fig. 9A) are likely a result of differences in upstream drainage area.

The second ordination plot (Fig. 9B) shows that sites with well-developed step-pool characters (represented by high values of ; shown as larger symbols) contain similar benthic macroinvertebrate taxa. Figure 9B includes the same plot of sites as in Figure 9A, but the size of the symbols scale according to the value of the relative steepness parameter . The sites shown with larger symbols (which indicates high values of ) cluster on the right side of Figure 9B and consist primarily of unburned and low-severity burned channels, with some that are also moderately burned. In contrast, study sites that exhibited degraded step-pool morphologies (lower values of ) are found in the center and left side of the ordination plot of Figure 9B, and correspond to highly and moderately burned areas. It is worth noting that the lowest values of (smallest symbols bounded by the oval in Fig. 9B) belong to both severely burned and moderately burned channels.

Figure 10 provides a synthesis of the interacting relationships among post-fire rainfall characteristics, morphological change, ecological responses, and the severity of burn. The graphs show the variables that correlated most strongly in the ordination analysis (Fig. 9A) with benthic macroinvertebrate compositions: Maximum average rainfall intensity, average step length, cumulative change in step height (normalized by days), cumulative change in thalweg elevation (normalized by days), and severity of burn. In other words, Figure 10 shows in bivariate plots how benthic macroinvertebrates varied with each of the rainfall, burn severity, and morphological variables that correlated strongly from the ordination analysis (Fig. 9A). The correlated rainfall variable was maximum average rainfall intensity (x-axis in Fig. 10); on the y-axes are the correlated morphological variables of average step length, cumulative change in step height (normalized by days), and cumulative change in thalweg elevation (normalized by days). The variables of benthic macroinvertebrates are displayed in the three columns of plots: taxa richness, percentage EPT, and percentage EPT excluding Baetidae and Hydropsychidae, using the sizes of data points within the plots to represent their values. Finally, the severity of burn, the remaining correlated variable from the ordination analysis (Fig. 9A), is shown as colors as defined in Figure 9 (where blue is unburned; green is low-severity burn; orange is moderate-severity burn; red is high-severity burn).

The graphs highlight several key points. First, as shown by the regression trends of Figure 10, as rainfall intensity (x-axis) increases, so do cumulative change in step height, cumulative change in thalweg elevation (the morphological variables on the y-axis), and, to a lesser extent, average step length. These relationships suggest that the intensity of rainfall influences morphological change in ways that include erosion and deposition (i.e., change in thalweg elevation) and the movement and destruction of step-pool sequences (i.e., increasing step length and changing step height). Second, the plots indicate that, as increasing rainfall intensity leads to greater morphological change, the ecological character of the study channels tends to degrade. In other words, the dots representing benthic macroinvertebrates get smaller (especially % EPT and % EPT excluding Baetidae and Hydropsychidae) at high rainfall intensities, large step lengths, large cumulative changes in step height, and large cumulative changes in thalweg elevation. Although the relation with taxa richness is less evident, the lowest percent EPT and lowest percent EPT excluding Baetidae and Hydropsychidae (smallest dots; Fig. 10) occurred at sites experiencing greatest rainfall intensities and morphological change. Third, the plots additionally show that the sites experiencing large rainfall intensities (x-axis), morphological change (y-axes), and ecological responses (smallest dots) were also burned with highest severity (orange and red colors). Therefore, the circumstance of large rainfall intensities coupled with high burn severity produced the largest morphological and ecological impacts after the Waldo Canyon Fire.

The natural variability inherent in fluvial systems is important to consider in this field study. The findings—whereby the highest rainfall intensities fell on the most severely burned sites and produced the greatest biophysical impacts—highlighted a key challenge of disentangling the effects of burn severity and rainfall intensity on the observed responses. Both control factors of burn severity and rainfall intensity are clearly important (Doerr et al., 2006). The intensity of rainfall during storms influences runoff and erosional processes after wildfire (Cannon et al., 2008; Moody and Martin, 2009; Kampf et al., 2016). The intensity of burn also influences soil properties and vegetation character, which in turn affect runoff responses (Booker et al., 1993; Robichaud, 2000; Benavides-Solorio and MacDonald, 2001). The spatial variability of burn severity and rainfall intensity often leads to complex erosional responses during storms (Moody et al., 2008).

The case of the Waldo Canyon Fire did not present a perfect field experiment to decipher which factor was more important—although the selected study reaches ranged widely over degrees of burn severity, the study period did not generate a case whereby low rainfall intensities fell on highly burned sites, for direct comparison with observed responses at these sites under high rainfall intensities. Neither did study sites in the low burn severity category experience a storm with high rainfall intensities, to match the observed cases of lower rainfall intensities that fell on these sites. The extreme storm event of 11–12 September 2013 (Table 2), however, did produce the greatest depths of rainfall in the study period over unburned reference areas (244.6 mm, Bear Creek gauge; Table 2; Fig. 2), with recurrence intervals of 500–1000 years. Yet, the unburned reference reaches remained virtually unchanged through the study period. These data suggest that rainfall alone likely does not fully explain the impacts of the Waldo Canyon Fire, especially as expressed through depths and recurrence intervals of rainfall.

The idea that the severity of burn offers a singular explanation for the observed responses may also seem incomplete. The category of burn severity represents an average condition of an upstream area that experienced heterogeneous burn intensity, making it difficult to capture the real controls in a sampling strategy. Although the contrast seems clear between the substantial geomorphological and ecological impacts of the severely burned site and the negligible/minor changes in unburned and low-severity burned sites, the variable responses at sites in the moderate burn category suggest other possible influences. Local factors that affect post-fire responses include fine-scale physiographic properties such as topography, lithology, soil characteristics, and land cover (Staley et al., 2015). The position of a site within the fluvial network relative to tributaries or other highly disturbed reaches may also play a role. One might attribute the aggradation at Willis Reach (Section 3.5; Fig. 7), for example, to its lower slope and its position in the main channel downstream of the steeper eroded Tributary Reach, rather than to the moderate severity of burn. Indeed, the lower slope might explain the difference between the aggradation at Willis Reach and the minor geomorphic response registered at the steeper Aussie Reach, burned similarly with moderate intensity. Yet, New Upper Reach, also burned with moderate intensity and with comparable slope as that for Willis Reach, registered a similar response of sedimentation despite its position upstream of the eroded Tributary Reach. Moreover, although the topographic position and slope of Aussie Reach (moderate burn category) is similar to that of Tributary Reach (severe burn category) within the fluvial network, it sustained negligible impacts compared to the major responses at Tributary Reach. These variations, therefore, still point to the important role of burn severity in influencing the ultimate extent of biophysical impacts.

The data reported in this study suggest that the most plausible interpretation is that both rainfall intensity and burn severity contributed to the observed post-fire bio-geomorphological responses in complementary and interacting ways. Results of the ordination analysis support this interpretation, as the strengths of the correlations between benthic macroinvertebrates and precipitation intensity and burn severity are nearly equal (r = 0.59 and r = 0.60, respectively; Fig. 9). Morphological change in the step-pool channels also correlated strongly with benthic macroinvertebrates, as demonstrated through the variables of cumulative change in thalweg elevation (r = 0.75), cumulate change in step height H (r = 0.72), and average step length L (r = 0.56). The higher correlation between benthic macroinvertebrates and step height (relative to step length) may reflect the tendency of large step clasts to attract sensitive organisms (Chin et al., 2009a). Moreover, because step height also correlated more strongly with rainfall intensity (Fig. 10; r² = 0.75) compared to step length, the changing height in step-pool sequences was apparently both more responsive to post-fire effects and influential in the ecological outcomes after fire. The results therefore suggest that rainfall intensity and burn severity together influenced geomorphological processes that altered the step-pool morphology, particularly through changes in step height. Both these factors in turn resulted in different types of organisms present following the Waldo Canyon Fire.

Analyses of field data over three years following the Waldo Canyon Fire of Colorado permit answers to the research questions posed in this paper. First, wildfire increased the susceptibility of step-pool channels to de-stabilize according to the severity of burn. Whereas the step-pool morphology remained relatively unchanged in unburned channels and at sites burned by low severity—even through the extreme event of September 2013 with rainfall recurrence intervals reaching 500–1000 years—the channel reach affected by high-severity burn (Tributary Reach in Williams Canyon) changed substantially following the first comparatively small magnitude though intense storm events (July 2013) after fire, with rainfall recurrence intervals of 5–10 years or less at all study sites. Extensive net erosion, in fact, transformed the step-pool reach into a bedrock channel after the large storm of September 2013. In between the two ends of the spectrum, for channels burned with moderate severity, morphological changes during the same events were variable. Whereas some channels (Willis Reach) changed substantially in their step-pool morphology due to extensive net deposition of sediment, others remained relatively unaffected (Aussie Reach). Other interacting factors, including local topography, channel slope, and position within the fluvial network, may account for some of this variability and serve as additional controls at these sites burned with moderate intensity during the Waldo Canyon Fire.

The morphologic responses recorded in the study channels after the Waldo Canyon Fire accord with changes observed in step-pool systems following large floods. Lenzi (2001), Turowski et al. (2009), and Molnar et al. (2010), for example, recorded how exceptional floods can obliterate step-pool sequences, in ways similar to the post-fire response in highly burned Tributary Reach. In other cases, Gintz et al. (1996) documented burial of step-pool morphology in the Bavarian Alps during a moderate flood, with rapid recovery during subsequent smaller events. In southern California, USA, post-fire influx of fine sediment similarly buried step clasts, but the step-pool structure quickly recovered as subsequent flows transported the fine material downstream and re-exposed the step-pool morphology (Florsheim et al., 1991; Keller et al., 2015). After the Waldo Canyon Fire, a similar response was also recorded in Willis Reach, which experienced a large influx of sediment and buried the step-pool structure temporarily.

Second, as the post-fire step-pool morphology altered according to the severity of burn, so did ecological responses. In the unburned and low-severity burned channels where the step-pool morphology remained intact through post-fire storms, ecological stability was largely evident by only minor changes in benthic macroinvertebrate metrics. In contrast, in high-severity burned Tributary Reach where the first post-fire storms quickly obliterated the step-pool structure, benthic macroinvertebrate metrics reflected degraded ecological conditions and did not show recovery. In study channels within the moderate burn category, the ecological responses also varied according to the extent of geomorphic changes. Of the moderate burn sites, sensitive organisms (percent EPT excluding Baetidae and Hydropsychidae) increased during the post-fire years only in Aussie Reach, where the step-pool features remained intact. In Willis Reach (a site burned by moderate severity) in which post-fire sedimentation buried step-pools, ecological conditions had not begun to recover three years after fire, as evidenced in the absence of EPT excluding Baetidae and Hydropsychidae.

The post-fire ecological degradation observed is also consistent with channels affected by geomorphic impacts of erosion and deposition. TLS showed substantial erosion and sedimentation in Tributary Reach and Willis Reach, respectively, where benthic macroinvertebrate communities also declined, particularly in EPT excluding Baetidae and Hydropsychidae. In seven Appalachian streams, Kaller and Hartman (2004) found lower EPT taxa richness and higher percentage of Baetidae organisms within the order Ephemeroptera, which were correlated with fine sediment accumulation. Elsewhere, erosion and channel bed instability were associated with lower percentages of EPT taxa in Vermont streams (Sullivan et al., 2004). The interacting hydrologic, geomorphic, and sedimentologic impacts on aquatic ecology after wildfire are difficult to generalize (Oliver et al., 2012), however, with recovery to reference conditions often requiring many years (Gresswell, 1999). Vieira et al. (2004), for example, documented depressed macroinvertebrate richness up to six years after a fire and subsequent flooding events in New Mexico, USA. The reduced percentages of sensitive taxa in the study streams (e.g., absence of EPT organisms in Tributary Reach and Willis Reach), therefore, reflect the initial devastating impacts of, and early recovery from, the Waldo Canyon Fire. Overall, the post-fire flooding causing erosion and sedimentation impair the hydraulic functioning and ecological health of step-pool systems.

Third, precipitation intensity and the severity of burn together influence the interacting morphologic and ecological responses after the Waldo Canyon Fire. Disentangling the roles of rainfall intensity and burn severity proved difficult in this study, as discussed above. Results show, however, that they served interacting and complementary influences on geomorphological and ecological impacts. Because the highest rainfall intensities coincidentally occurred over areas of high burn severity, these sites experienced the greatest impacts.

In conclusion, this paper reports initial changes in step-pool channels in the Colorado Front Range, USA, following the 2012 Waldo Canyon Fire. The findings suggest that (1) wildfire increased the susceptibility of step-pool channels to de-stabilize according to the severity of burn; (2) ecological responses of benthic macroinvertebrates mirrored the geomorphic changes following the post-fire storms; and (3) precipitation intensity and the severity of burn together influenced the interacting morphologic and ecological responses after the Waldo Canyon Fire.

This study is significant in generating a rich and detailed set of field data which expands our knowledge regarding the interacting biogeomorphic responses of step-pools streams after wildfire. Even though step-pools are among the most stable features in the fluvial system, this study demonstrated how wildfire can lead to destabilization of these bedforms quickly through post-fire storms, and thereby eliminate their important roles of hydraulic resistance and energy dissipation in the fluvial system. In addition to increasing understanding of the post-fire response of step-pool streams, in greater detail than before, this study is also significant in providing specific cases for the destabilization of step-pool sequences. When step-pools become unstable, the risk of flooding, erosion, and sedimentation increases downstream, further degrading ecosystems and impacting lives and property. Finally, the findings of this study offer insights for the post-fire management of step-pool channels. The data showed that step-pool systems are inherently resilient to the impacts of low-severity burns, supporting the concept that low-severity fire is, in fact, beneficial for ecosystems with little need for management. On the other hand, high-severity burn, especially when coupled with high-intensity rainfall, makes step-pool systems susceptible to complete destruction; thus management is not likely productive or practical. Therefore, it is the areas burned by moderate intensities, with variable responses that are difficult to predict and likely influenced by a range of factors, that might benefit from further examination and possible management. In these cases, the installation of step-pools to increase or maintain the integrity of such channels (e.g., Chin et al., 2009b) may be beneficial on a case-by-case basis. As wildfires grow in frequency and magnitude along with global warming, these management strategies become increasingly important for anticipating future changes in burned river landscapes, mitigating potential floods and hazards, and promoting sustainable river ecosystems.

This project would not have been possible without the financial, logistical, and moral support and assistance from numerous organizations and individuals. The National Science Foundation (EAR1254989) and the University of Colorado (CU) Denver (Center for Faculty Development and College of Liberal Arts and Sciences) provided grants for this study. UNAVCO’s Keith Williams collected the data with terrestrial LiDAR scanning. The U.S. Forest Service provided access (Permit PPK577), guidance, and logistical support, in particular Steven Sanchez, Dana Butler, Lisa Gowe, Jeffrey Hovermale, and District Ranger Allan Hahn. The City of Colorado Springs Parks, Recreation and Cultural Services Department, Regional Parks Division, granted access to the references reaches in the Bear Creek watershed. The Navigators, especially Derek Strickler, offered background information, access to the study reaches in Camp Creek, and lodging for the field crew over many nights. The U.S. Air Force Academy, with assistance from Stanley Rader, facilitated access to unburned Academy Reach. John Moody and Deborah Martin of the U.S. Geological Survey provided tipping-bucket rain gauges and insightful discussions through field visits to the burned areas. Laura Laurencio, Francis Rengers, and Rune Storesund offered expertise with data collection and analysis. Many students at CU Denver served as field and laboratory assistants: Rhonda Barton, Daniel Ben-Horin, Brian Bencivengo, Dat Bui, Kim Conway, Tera Del Priore, Sam Epperly, Jonathan Key, Corine Roberts-Niemann, and Kara Utter. The Western Center for Monitoring and Assessment of Freshwater Ecosystems of Utah State University provided the electronic subsample program created by Dave Roberts. Aquatic Biology Associates, Inc. identified samples of benthic macroinvertebrates in 2012; those from 2013 and 2014 were processed in the Environmental Hydrology Laboratory at CU Denver. We are grateful to two anonymous reviewers and the editors whose insightful comments greatly improved the quality of this paper.