Repeated high flows drive morphological change in rivers in recently deglaciated catchments

Climate change is decreasing glacier cover and increasing the frequency and magnitude of precipitation‐driven high flows and floods in many regions of the world. Precipitation may become the dominant water source for river systems in recently deglaciated catchments, with major rainfall events driving significant changes in river channel morphology. Few studies, however, have examined river channel response to repeated precipitation‐driven high flows. In this study, we measured the geomorphological condition of four low‐order rivers in recently deglaciated catchments (70–210 years ice free) before and after a series of repeated precipitation‐driven high flows during summer 2014. High flows drove substantial initial morphological change, with up to 75% change in baseflow channel planform position and active channel form change from pre‐ to post‐high flow. Post‐high flow years were associated with increased instream wood and geomorphological complexity at all but the youngest river. Channel changes were part of an active relaxation stage at all rivers, where channels continued to migrate, and complexity varied through time. Overall, these measurements permit us to propose a conceptual model of the role of geomorphologically effective high flows in the context of paraglacial adjustment theory. Specifically, we suggest that older rivers in recently deglaciated catchments can undergo a short‐term (<10 years) increase in the rate of geomorphological development as a result of the recruitment of instream wood and channel migration during and following repeated precipitation‐driven high flows. Enhancing our knowledge of these geomorphological and paraglacial processes in response to high flows is important for the effective management of riverine water and ecosystem resources in rapidly changing environments.


| INTRODUCTION
Arctic, sub-arctic and high mountain regions are presently experiencing more extreme and rapid responses to climate change than other parts of the world (Pepin et al., 2015;Box et al., 2019). In addition to increasing atmospheric temperature (Hansen et al., 2006), these regions are expected to encounter increases in the occurrence of extreme rainfall (Bintanja & Andry, 2017;Curtis, 2019), leading to increased frequency and magnitude of precipitation-driven floods (Berghuijs et al., 2017). The increasing frequency and magnitude of precipitation-driven floods and high flows are important because The data used in this study is available from the corresponding author upon reasonable request. these events can substantially alter river hydraulic and geomorphological conditions (Schindler & Smol, 2006). Extreme floods, whose peak discharges are characterised by recurrence intervals >50-100 years (Woodward et al., 2015), can drive major geomorphological change in rivers (Hauer & Habersack, 2009;Magilligan et al., 2015;Fryirs, 2017). Extreme floods have been shown to mobilise substantial volumes of sediment and drive changes to river planform (Thompson & Croke, 2013;Pasternack & Wyrick, 2017;Borga et al., 2019). Specific geomorphological responses depend upon the flood hydrograph and the catchment geology, topography including valley confinement, landcover and connectivity (Milan, 2012;Brogan et al., 2017;Lane et al., 2017;Righini et al., 2017;Tunnicliffe et al., 2017;Scorpio et al., 2018;Wohl et al., 2019).
In addition to mobilising sediment through bed and bank erosion, floods facilitate widespread erosion, transport and local accumulations of instream wood (Comiti et al., 2016;Steeb et al., 2017). Instream wood drives localised changes to velocity profiles, accelerating flow locally, increasing turbulence and enhancing erosion and deposition, driving downstream sediment fining and the development of lowvelocity geomorphological unit types (Naiman et al., 1999;Klaar et al., 2011), as well as altering river-wide fluvial processes (Montgomery et al., 1995;Abbe & Montgomery, 1996;Wenzel et al., 2014). As a roughness element, the form of a wood structure (Daniels & Rhoads, 2003;Wondzell & Bisson, 2003) influences the extent of hydraulic change and sediment erosion/deposition (Gippel et al., 1996;Hughes et al., 2007). Where increases in instream wood abundance and complexity occur, it can be expected to play an important role in the development of instream (baseflow and active channel) geomorphic complexity (Wohl et al., 2015;Wohl et al., 2016;Grabowski et al., 2019).
A post-flood geomorphological relaxation or response stage can be recognised immediately after extreme floods (Milan, 2012). Floodmobilised sediment alongside newly accumulated or exposed roughness elements, such as instream wood and boulders (Brummer et al., 2006), drive ongoing geomorphological change through increased local erosion and deposition (Lester & Wright, 2009). The stability or persistence of geomorphological features post-high flow (Dean & Schmidt, 2013) defines the duration of this period.
The capacity of a given flood or high flow to drive geomorphological change is related to the overall energy available to effect change (Wolman & Miller, 1960;Costa & O'Connor, 1995). Geomorphological change is associated with increased total energy available, rather than simply peak flow magnitude, flood frequency or total duration. This framework has since been supported in observational work (Chappell et al., 2003), although not in all cases (Amponsah, 2017). Yet much of the literature continues to focus on single event, extreme, high-magnitude floods (e.g. Haschenburger & Wilcock, 2003;Carrivick & Rushmer, 2006;Carrivick & Rushmer, 2009;Hauer & Habersack, 2009;Dean & Schmidt, 2013;Fryirs et al., 2015).
Sequences of large floods have been shown to drive substantial morphological change in observational and modelling studies (Warburton, 1994;Dunning et al., 2013;Guan et al., 2016). However, the aggregated effects of smaller high flows (including smaller floods), or those occurring in quick succession, have received less attention.
Repeated high flows are defined here as a high-frequency series of recurrent discharge peaks which individually may not be considered to represent significant flood disturbances (flow and flood pulses; Bertoldi et al., 2010) occurring over weeks to months. Understanding the effects of these smaller 'flow pulses' (< 1 to 1 year return intervals) and 'flood pulses' (on average 1 in 2 year return intervals; terms defined by Bertoldi et al., 2010) are necessary because they can mobilise sediment (e.g. Kavan et al., 2017), drive localised morphological change (Pickup & Warner, 1976;Chappell et al., 2003;Zonta et al., 2005;Brown et al., 2015;Surian et al., 2015;Stocker-Waldhuber et al., 2017) and, with water temperature (e.g. Carrivick et al., 2012), can govern water quality. Previous research has demonstrated repeated high flows drive significant geomorphological change in large, braided, gravel bed rivers (Bertoldi et al., 2010;Mao & Surian, 2010). The geomorphological response of low-order (first to third order) rivers to repeated large floods and to repeated high flows (Lenzi et al., 2006;Rickenmann et al., 2012;Rainato et al., 2018) has been explored yet still requires further advancement (Comiti & Mao, 2012).
Rivers in Arctic, sub-arctic and mountainous environments are intrinsically linked to paraglacial adjustment, as water sources shift following complete catchment deglaciation (e.g. Carrivick et al., 2013). Projected increases in precipitation have been identified as a significant contributor to future river flows following glacier decline (Immerzeel et al., 2013, Milner et al., 2017. Initially unstable and active morphological conditions are observed as numerous geomorphological and hydrological processes alter the landscape (Church & Ryder, 1972;Ballantyne, 2002). Young catchments (those that have been recently deglaciated), with abundant unconsolidated glacial sediment (McColl, 2012;Carrivick & Heckmann, 2017), demonstrate high terrestrial landform erosion rates and sediment availability (Klaar et al., 2015). Whilst a temporal trend for decreasing sediment supply is observed as vegetation and soil organic matter develop (Egli et al., 2006, Malone et al., 2018 with systems becoming less sensitive to change (Harvey, 2001). Physical disturbances such as rock slope failures, debris flows and fluvial incision can continue to locally mobilise sediment despite the trend for stabilisation (Curry et al., 2006;Ballantyne & Stone, 2013;Legg et al., 2014).
The aim of this study is to quantify the extent to which repeated high flows (eight flow and flood pulses with recurrence intervals of up to 5 years) drive morphological change from the analysis of four rivers in recently deglaciated catchments at different stages of paraglacial adjustment in Glacier Bay, Alaska (Malone et al., 2018). On the basis of previous work, three hypotheses were tested in this study: H1. Repeated high flows will change baseflow river channel position and active channel form due to significant bank erosion, sediment transport (Pasternack & Wyrick, 2017) and accumulation of instream wood (Ruiz-Villanueva et al., 2016); H2. Post-high flow response of baseflow channel position and active channel form will be greatest at older sites, due to greater change in relative sediment supply and instream wood recruitment and; H3. Repeated high flows will facilitate the ongoing development of river geomorphological complexity due to fluvial sediment transport which represents an important element of ongoing paraglacial adjustment in rivers in recently deglaciated catchments.

| Study site
Glacier Bay National Park and Preserve in south-east Alaska (58 10 0 -59 15 0 N; 135 15 0 -138 10 0 W) is dominated by a tidal fjord, 150 km long and 20 km wide (Figure 1). The park covers an area of 11,030 km 2 with a maritime climate and a mean annual precipitation of 1,400 mm. At least three Holocene glacier advances and retreats have occurred in Glacier Bay (Lawson et al., 2004;Wiles et al., 2011), the largest being the formation of a Neoglacial ice sheet during the Little Ice Age which covered the majority of the region at its maximum ($1700). Retreat of the ice sheets began at the end of the Little Ice Age, and its rate was identified as one of the most rapid in modern global conditions (Chapin et al., 1994). Glacier retreat has resulted in the development of proglacial landscapes and river systems undergoing rapid adjustment to meltwater and sediment supply regimes as part of paraglacial adjustment.
Landscape development has a number of key recognised stages in Glacier Bay (Chapin et al., 1994), beginning with the development of a crust of blue-green algae, lichens, liverworts, forbs, mountain avens (Dryas drummondii) and sparse willows (Salix spp.). Mountain avens are an effective fixer of nitrogen and as a result dominate early-stage successional landscapes, alongside individuals of willow.
By $50 years ice free, Sitka alder (Alnus crispa) and willow are the dominant plant species, mountain avens is typically lost and cottonwood (Populus trichocarpa) begins to occur. After $100 years, generally Sitka spruce (Picea sitchensis) dominates with an increasing contribution of Western hemlock (Tsuga heterophylla) through time.
As catchment glacial cover declines, the depletion and stabilisation of formerly deposited glacial sediment results in decreasing suspended sediment loads and increased channel stability in river systems (Sidle & Milner, 1989). The initial development of vegetation including trees in the riparian corridor acts to stabilise banks as their roots bind soils and ground cover decreases erosion, which in turn further increases river channel stability. As riparian wood is accumulated into the river channel it provides a roughness element, causing significant local and reach scale alterations to hydrogeomorphic conditions (Klaar et al., 2011;Wenzel et al., 2014). Instream wood facilitates the continued development of instream geomorphic and hydraulic habitat (Milner and Gloyne-Phillips, 2005;Klaar et al., 2011).
A peak in habitat complexity is observed in rivers at an intermediate age ($150 years), as instream wood begins to accumulate but channels still support high velocity habitat types .
Fluvial systems in the National Park have been aged based upon the time since glacier ice recession from the stream mouth using historic records and imagery (Milner et al., 2000). This allows for a spacefor-time substitution approach to be applied in studies of their physical and ecological succession (Milner et al., 2008;Klaar et al., 2009;Klaar et al., 2015).  Figure 2). These four catchments ($20-30 km 2 ) span a 140-year age range (from 70 to 210 years; Table 1). Consequently, these catchments can be considered to represent diverse stages of paraglacial adjustment with varied catchment sediment stability and riverine development. Other than their stage of paraglacial adjustment and vegetative succession, these catchments share broad morphological similarities. These similarities allow for comparison of their morphological response to disturbances, including repeated high flows to be made. The study sites at each of these rivers are alluvial and support predominantly pool-riffle sequences (Montgomery & Buffington, 1997).
During the summer of 2014, recurrent intense precipitation events occurred from June to September, which resulted in the wettest summer period in a 30 year record , causing repeated flooding and high flows across the region of south-east Alaska and within Glacier Bay. Additional information regarding weather patterns and regional river flows are reported in Supplementary Information 1 (Supplementary Figures S1, S2 and Table 1).
Through the summer, eight notable high flow events occurred, characterised by peak discharges exceeding twice the median daily F I G U R E 1 Map of Glacier Bay, Alaska, including location of study catchments (outlined in white; age recorded in parentheses, described time since ice recession from river mouths) and rainfall and river gauge locations (Bartlett Cove and Lemon Creek, respectively) [Color figure can be viewed at wileyonlinelibrary.com] T A B L E 1 Physical characteristics of study rivers. Adapted from Hill et al. (2009), Klaar et al. (2009) and Klaar et al. (2015). River age equals years since ice recession from mouth. Average discharge calculated from gauged flow in 2007. Reach gradient calculated over intensively studied representative reach. Estimated critical bed shear stress calculated using D 50 (Berenbrock & Tranmer, 2008  The absence of detailed river discharge data for our study rivers prevents the precise description of the form of each flow and flood pulse at each site and consequently precludes quantitative analyses of the relationship between river channel response and geomorphological effectiveness of floods. However, the data sets included in this study allow for valuable information to be gathered regarding the morphological response of rivers to a group of flow events.

| Geomorphological mapping of baseflow channels and analyses
Ground-based surveys of river planform were undertaken in two pre- Channel Geomorphic Unit (CGU) approach, first established in Glacier Bay National Park by Klaar et al. (2009). CGUs were mapped along the same survey stretches as planform surveys. During CGU surveys the thalweg length of each CGU was recorded moving downstream using a hierarchical visual approach based upon bed form, levels of turbulence and position within the channel (Hawkins et al., 1993 A percentage metric of baseflow wetted channel positional persistence (persistence) was also calculated (Equation 3). The metric is based on the persistence of occupied raster cells between two surveys either both before, one before and one after or both after the floods. GPS line maps were converted to raster layers, and the raster calculator tool was used to create a new raster layer containing all cells occupied in both surveys (Supplementary Figure S3). The number of cells occupied in this new raster layer was divided by the total number of unique cells occupied across both surveys (i.e. sum of both raster layers cell counts minus number of shared cells). This was multiplied by 100 to create a percentage score, in which, a score of 100 means that the wetted stream channel was identical in both surveys (i.e. no grid cells representing a wetted channel were lost or formed). A score of 0 means no part of a wetted stream channel was shared between the two surveys (no grid cells representing a wetted channel were shared). Comparisons were allocated to one of three relative high flow periods ( The relationship between relative high flow period, the sites and persistence of the river channel was tested using a Wald chi squared test for difference based upon a GLM which incorporated relative high flow period and site into a model with the form: pre-high flow, high flow and post-high flow periods) + β 2 × site (Fox & Weisberg, 2011).
Channel geomorphological complexity was calculated using Shannon's Diversity Index (SHDI) calculated from raster layers using Equation 5 ( McGarigal et al., 2012). SHDI increases from 0 to 1 as geomorphological unit type richness increases and/or as the proportional contribution of geomorphological unit types becomes more equitable. were identified metrics were removed from the dataset depending upon their perceived relevance in response to high flows (R-Core-Team, 2017).

| Sediment and active channel cross sections surveys and analyses
A modified Wolman walk (Wolman, 1954) moving in a zigzag walk upstream was used to collect sediment b axis length data (nearest mm) from a minimum of 300 (n = 300 to $2,700) sediment T A B L E 2 Year comparison allocations to relative high flow periods in persistence analyses.

Relative high flow period Year comparisons
Pre-high flow 2007-2010 High flow 2007-2015, 2007-2016, 2007-2017 2010-2015, 2010-2016, 2010-2017 Post-high flow 2015-2016,2015-2017 2016-2017 particles at each river in 2008 and 2017 only. A minimum of 30 measurements were taken from sediment in at least 5 CGUs of each type found at each river. No sediment data collection was possible at Wolf Point Creek in 2008. These data were used to calculate D 50 for each site and CGU type.
The Kolmogorov Smirnov test was used to test if the distribution of sediment b axis lengths were significantly different between years within and between rivers, p values were adjusted for multiple testing using a Bonferonni adjustment (R-Core-Team, 2017).
Cross sections established in the 1980s by Sidle & Milner (1989) in a reach representative of the wider river network were re-

| RESULTS
During the pre-high flow period (2007)(2008)(2009)(2010), persistence of the position of the baseflow wetted river channel was high (mean = 58.4 ± 7.2%, Table 3) across all sites. The highest pre-high flow persistence score was recorded at the oldest site (Rush Point Creek -65.8%), whilst the lowest score was recorded at the intermediate age (146 years Figure 6).
Significant sediment fining was identified in segment scale cumulative length frequency curves at Rush Point Creek (Table 5). In contrast, at Berg Bay South Stream, a loss of fine sediments was identified following flooding, resulting in statistically significant changes in cumulative length frequency curves (Table 5)   T A B L E 4 Cross section sediment areas, and percentage change in total area and persistence of form between years.

| River responses to repeat high flows
We demonstrate that repeated high flows are capable of causing sub-  (Downs & Gregory, 1993;Fryirs, 2017). A similar response has previously been reported for a series of repeated high flows (six flow and flood pulses) in a large gravel-bed river in which entire sections of the river channel network were reworked (Bertoldi et al., 2010;Surian et al., 2015). Our study extends these findings and indicates that flow and flood pulses are also capable of causing extensive in channel change in low-order rivers. An important future step in understanding the role of high flows in river channel change is the quantitative analyses of the relationship between river channel response and geomorphological effectiveness of floods represents an important future step. To date this topic has received limited attention, with research mainly addressing extreme floods (Leyland et al., 2017). This is in part a result of the unpredictable nature of such high flows which makes field data collection challenging.
Of the four rivers, the lowest response in channel morphology to the repeated summer high flows occurred at the youngest river (Wolf Point Creek -70 years since deglaciation). Active channel change was dominated by incision evidenced in the limited lateral migration and minor channel profile adjustment. This small response is consistent with that reported to similar floods in forested headwater rivers (Phillips, 2002). . This increasing contribution of low velocity CGUs is likely to be associated with the accumulation of instream wood during the high flow period , driving the formation of low velocity CGUs (Abbe & Montgomery, 1996).
Instream wood accumulation was potentially the result of bank erosion and positional change observed at each site within the riparian forests. This process has been shown to be the dominant driver of wood recruitment in small catchments (Martin & Benda, 2001).  (Abbe & Montgomery, 1996;Montgomery et al., 2003;Mao et al., 2008).

| Post-high flow morphological response
Extensive morphological change following the summer high flows has been observed in the three older rivers in this study, during which the channels remained extremely active supporting H2 that the post-high flow response of active river channel form and baseflow positional persistence will be greatest at older sites. Such mobile channels have been reported in response to extreme floods in rivers of varying sizes where substantial volumes of sediment are mobilised during floods (Milan, 2012;Dean & Schmidt, 2013;Tunnicliffe et al., 2017). To date, negligible evidence exists for similar responses following repeated high flows as observed in our study. Our findings indicate that high flows during the study period had the capacity to drive substantial baseflow and active river channel change, and therefore should be considered to represent important geomorphological events in loworder rivers.
In contrast to the substantial post-high flow responses of the three oldest rivers in this study, Wolf Point Creek demonstrated comparatively little change in the period following the high flows. There are a number of potential reasons for this limited response. The presence of an upstream lake may act to smooth the peaks of individual flow events (although it is unlikely to alter the total energy available for change) and capture sediment mobilised by the high flows reducing any increase to sediment supply (Arp et al., 2007). The loss of instream wood observed from pre-to post-high flow may further reduce the potential to maintain geomorphological complexity (Yarnell et al., 2006). Coarser bed sediment compared with other rivers may have acted as an armour layer preventing further bed degradation following change during the high flows (Wilcock & Southard, 1989).
Whilst reported channel incision and reach scale straightening during the high flows period may have acted to increase average flow velocities conveying mobilised sediment downstream of the survey section and preventing the accumulation of any new instream wood (Wilcock, 1993). Given the other similarities of catchment and channel form between Wolf Point Creek and the older rivers it is likely in

| Floods, sediment supply and paraglacial adjustment
Substantial wetted channel migration observed in both cross section and planform analyses, as well as turnover of sediment in cross sections, demonstrate that repeated precipitation-driven high flows have the potential to drive major morphological change mobilising sediment in recently deglaciated catchments. These catchments have previously been shown to become increasingly stable through time as riparian vegetation establishes, soils develop and bank cohesion increases (Sidle & Milner, 1989;Klaar et al., 2009;Klaar et al., 2015;Malone et al., 2018), consistent with well-established principles of landscape development (Gurnell et al., 2000). As banks are stabilised and sediments stored within alluvial channels, smaller flow events become limited in their capacity to mobilise sediments.
Through this process the relative importance of larger flow pulses and floods increases as they become the dominant driver of sediment transport (Gurnell et al., 2000). Our findings offer some support for H3 that repeated high flow driven fluvial sediment transport represent an important element of ongoing paraglacial adjustment (Miller, 1990;Darby et al., 2007;Guerra et al., 2017). The morphological response of the four study rivers following the high flows mirrored the temporal development of geomorphological complexity already reported for Glacier Bay National Park rivers ). Here, we conceptualise the role of geomorphologically effective high flows in the paraglacial adjustment model (Figure 9), based upon findings from this study within the necessary context of the wider paraglacial adjustment literature.
The sediment supply/capacity ratio can dictate the capacity for geomorphological complexity in rivers (Yarnell et al., 2006), with peaks observed at intermediate levels of sediment supply and where roughness elements are abundant. Patterns observed as geomorphological complexity develops in our study rivers in recently deglaciated catchments Klaar et al., 2015). Sediment supply in these rivers is closely coupled to their stage of paraglacial adjustment, declining through time as catchment geomorphology stabilises and sediments are sorted within paraglacial river channels (Carrivick & Heckmann, 2017;Lane et al., 2017). However, temporal variability in sediment supply, sediment transport and landscape connectivity has recently been reported demonstrating the complexity of these relationships (Micheletti & Lane, 2016;Lane et al., 2017;Comiti et al., 2019).
High flows and floods, including repeated flow and flood pulses as observed here, have the capacity to increase the relative sediment supply through bank and bar erosion (Kociuba & Janicki, 2014;Miller et al., 2014;Fox et al., 2016;Leyland et al., 2017). Additionally, instream roughness elements (wood) can interact with high flows further increasing sediment erosion (Abbe & Montgomery, 1996;Lester & Wright, 2009;Parker et al., 2017). Together, these high flow induced changes should act to elevate the potential for a river to support geomorphological complexity (Yarnell et al., 2006;Klaar et al., 2009), as observed following the high flows in the study rivers. Although not directly evidenced in this study, increased post-high flow sediment supply is supported by extensive baseflow channel migration and active channel change observed across rivers, and by significant sediment fining response observed at Rush Point Creek. A pattern linked to increased sediment supply in experimental and observational studies (Hassan and Church, 2000, Recking, 2012. Indeed, sediment supply has recently been shown to govern channel geometry in a dataset of over 300 rivers spanning North America (Pfeiffer et al., 2017).
Due to the unstable nature of freshly recruited sediments and the presence of newly recruited roughness elements flow pulses following main high flow events may be more geomorphologically effective than similar events pre-high flow. Increases in the potential for geomorphological complexity would be more significant still where stream power is insufficient to wash out instream wood allowing rapid local accumulation (Wohl & Goode, 2008;Wohl et al., 2016), as would be the case during smaller flood and flow pulses, enhancing the river channel's capacity for geomorphological complexity (Yarnell et al., 2006). These two factors combined appear to enable the rapid development of river geomorphological complexity under suitable conditions during and following high flows. This role of high flows in paraglacial adjustment theory may become increasingly significant and prevalent globally as glaciers continue to retreat F I G U R E 9 Model of paraglacial adjustment processes including relative geomorphological effectiveness of high flows [Color figure can be viewed at wileyonlinelibrary.com] (Marzeion et al., 2014;Milner et al., 2017), and as rainfall becomes a dominant water source and falls more intensely under climate change (Trenberth, 2011;Berg et al., 2013).

| CONCLUSION
This study has quantified and explored the capacity of repeated precipitation-driven high flows to drive geomorphological change within rivers in recently deglaciated catchments. Repeated high flows caused significant change in baseflow channel planform persistence, adjusted active channel form, recruited large amounts of instream wood at three rivers and removed instream wood from the youngest river, with years following the high flows generally being associated with increased geomorphological complexity. Post-high flow geomorphological responses of river channels were more extensive in older rivers where instream wood was recruited and continued turnover of sediment was observed in cross sections. Inclusion of the relative importance of geomorphologically effective high flows into the paraglacial adjustment model identifies an important short-term (<10 years) driver of continued river development. Short-term geomorphological activity due to high flow pulses is likely to become increasingly pronounced both as glacial contributions to river flows decline in the future and as precipitation inputs become greater. Knowledge of such geomorphological activity will be important for understanding sediment supply, water quality and rapidly evolving riverine habitats, as well as for water and ecosystem resource management.

ACKNOWLEDGMENTS
We acknowledge the contribution of research assistants who have through the duration of this study provided important contributions to the delivery of this research. These individuals include Sophie