Geomorphology 293 (2017) 227–239

Contents lists available at ScienceDirect

Geomorphology

journal homepage: www.elsevier.com/locate/geomorph

A 184-year record of river meander migration from tree rings, aerial MARK imagery, and cross sections

⁎ Derek M. Schooka, , Sara L. Rathburna, Jonathan M. Friedmanb, J. Marshall Wolfc a Department of Geosciences, Colorado State University, 1482 Campus Delivery, Fort Collins, CO 80523, USA b U.S. Geological Survey, Fort Collins Science Center, 2150 Centre Avenue, Building C, Fort Collins, CO 80526, USA c Department of Ecosystem Science and Sustainability, Colorado State University, 1476 Campus Delivery, Fort Collins, CO 80523, USA

ARTICLE INFO ABSTRACT

Keywords: Channel migration is the primary mechanism of floodplain turnover in meandering rivers and is essential to the Channel migration persistence of riparian ecosystems. Channel migration is driven by river flows, but short-term records cannot Meandering river disentangle the effects of land use, flow diversion, past floods, and climate change. We used three data sets to Aerial photography quantify nearly two centuries of channel migration on the Powder River in Montana. The most precise data set Dendrochronology came from channel cross sections measured an average of 21 times from 1975 to 2014. We then extended spatial Channel cross section and temporal scales of analysis using aerial photographs (1939–2013) and by aging plains cottonwoods along transects (1830–2014). Migration rates calculated from overlapping periods across data sets mostly revealed cross-method consistency. Data set integration revealed that migration rates have declined since peaking at 5 m/ year in the two decades after the extreme 1923 flood (3000 m3/s). Averaged over the duration of each data set, cross section channel migration occurred at 0.81 m/year, compared to 1.52 m/year for the medium-length air photo record and 1.62 m/year for the lengthy cottonwood record. Powder River peak annual flows decreased by 48% (201 vs. 104 m3/s) after the largest flood of the post-1930 gaged record (930 m3/s in 1978). Declining peak discharges led to a 53% reduction in channel width and a 29% increase in sinuosity over the 1939–2013 air photo record. Changes in planform geometry and reductions in channel migration make calculations of flood- plain turnover rates dependent on the period of analysis. We found that the intensively studied last four decades do not represent the past two centuries.

1. Introduction from more subtle changes in land cover (Costa et al., 2003), irrigation extraction (Johnson, 1994), and climate (Favaro and Lamoureux, Flooding, sediment redistribution, and recycling of land are re- 2015). Climate change has modified the timing and magnitude of flood quired for the persistence of healthy river ecosystems (Gurnell et al., peaks, especially in snowmelt-driven rivers (Stewart et al., 2005; Clow, 2012; Meitzen et al., 2013). River flows not only transport in-channel 2010). Land use and land cover changes can increase (Costa et al., sediment, but their lateral channel migration also destroys and creates 2003) or decrease (Li et al., 2009) flood magnitudes. Channel migration land (Hickin and Nanson, 1975; Wohl et al., 2015). Channel migration can be directly suppressed through bank stabilization (Florsheim et al., rates vary between 0 and > 100 m/year across reaches, rivers, and 2008) or indirectly through species invasion (Cadol et al., 2011), flow hydroclimatic periods (Hooke, 1980; Lawler, 1993; Hudson and Kesel, regulation, and water extraction (Fremier et al., 2014). Any of these 2000). Modification to fluvial disturbance regimes can transition modifications can alter channel migration produced by point bar floodplains into upland communities with lower biodiversity (Merritt growth and cutbank recession (Parker et al., 2011; van de Lageweg and Cooper, 2000; Johnson et al., 2012). Even minor changes to water et al., 2014). and sediment regimes may have significant effects, which makes free- Channel migration is somewhat predictable (Howard and Knutson, flowing meandering rivers excellent locations to detect environmental 1984; Luchi et al., 2007), but ultimately it is sufficiently complex to changes (Erskine et al., 1992; Palmer et al., 2008). prevent accurate prediction of changes in a given meander (Nanson and Several conditions can alter the physical processes responsible for Hickin, 1983; Nanson, 1986; Hooke, 2007; Güneralp et al., 2012). channel migration. Damming dramatically alters fluvial processes and Channel gradient, meander curvature, bank resistance, and flow history landforms (Brandt, 2000; Nilsson et al., 2005), but shifts also result all affect the sensitivity of river meanders. Flow and geomorphic

⁎ Corresponding author. E-mail address: [email protected] (D.M. Schook). http://dx.doi.org/10.1016/j.geomorph.2017.06.001 Received 22 December 2016; Received in revised form 2 June 2017; Accepted 2 June 2017 Available online 06 June 2017 0169-555X/ © 2017 Elsevier B.V. All rights reserved. D.M. Schook et al. Geomorphology 293 (2017) 227–239 histories affect channel response to subsequent flows (Schumm, 1973; flows have been removed for uses such as irrigation (Chase, 2013), the Moody et al., 1999). For example, bank erosion may be affected by decrease in mean annual flow from 1931 to 2010 was not significant deposits from previous floods (Nanson and Hickin, 1983; Nanson, (Schook et al., 2016). Water development has had little impact on peak 1986), and extreme floods may incite channel responses that last sev- flows (Chase, 2013). eral decades. Because characterizing different meanders is difficult, Powder River floods are produced in four main ways (Moody et al., assessing meander migration at the reach scale is useful for character- 2002; Moody and Meade, 2014). First, low-elevation ice jam floods izing channel migration. Additionally, because of the time lags affecting occur between late February and early April when the southern, up- channel responses to dynamic flows and large floods, channel migration stream reaches of the river thaw before downstream reaches. These should be considered over a range of timescales. floods are spatially limited. Second, Bighorn Mountains high elevation Diverse research approaches have been used to study channel mi- snowmelt in May–June creates the peak annual flow in 60% of years gration, including repeat planimetric and cross sectional surveys, (Schook et al., 2016). These snowmelt floods can be augmented by structure from motion (Fonstad et al., 2013), terrestrial laser scanning spring rain, as occurred in 1978, to produce the largest peak discharge (Milan et al., 2007), photogrammetry (Lawler, 1993), and numerical (930 m3/s) in the continuous gage record. Third, flash floods from modeling (Larsen et al., 2006). The data sets generated from these convective rainstorms occasionally produce short-duration high flows approaches are usually collected over short timescales (years), leaving in spring, summer, and fall. Fourth, frontal storms can cause autumn high uncertainty in characterization of migration at the century-to- floods. The largest documented flood was in 1923 and preceded the millennial timescales (Hughes, 1997). Approaches addressing multi- gage record. It peaked at an estimated 3000 m3/s and was caused by a century timescales include radio carbon dating (Knox, 2000; Erskine slow-moving September cold front that stalled over the region for et al., 2012) and optically stimulated luminescence (Hobo et al., 2014), several days (Grover, 1925). but both techniques generally suffer from low spatial extent and low Since 1975, Moody and Meade (1990) have monitored 24 perma- temporal resolution. Intermediate length (101–102 years) data sets nent channel cross sections along the Powder River between Moorhead might be best for characterizing channel migration. These include and Broadus, MT, just north of the Wyoming-Montana border. Our multidecade channel cross section measurements, which have revealed study reach is a subset of the Moody and Meade (1990) study area mechanisms of floodplain formation (Pizzuto, 1994; Moody and Meade, between PR130 and PR194, a valley distance of 37 km and a 2013 river 2008) and meander evolution (Hooke, 2007). Aerial photographs pro- distance of 75 km (Fig. 1). The channel bed is composed of sand and vide data sets with longer timescales that extend up to 80 years. gravel (Hembree et al., 1952), while the point bars and floodplain are Dendrochronology (tree ring science) can offer an approach cov- ~80% silt-fine sand (Moody and Meade, 2008). ering longer timescales than channel cross sections and air photos, which is possible when floodplain trees establish predictably in re- 2.2. Floodplain forests sponse to channel migration (Merigliano et al., 2013). Cottonwood trees (Populus spp.) grow in floodplains and can provide a living record Plains cottonwood (Populus deltoides ssp. monilifera) is the most of centuries of channel changes (Hickin and Nanson, 1975; Edmondson common riparian tree across the western USA (Friedman et al., 2005), et al., 2014). Cottonwoods are widely distributed on floodplains in the including along the lower Powder River. Additional woody vegetation western USA (Friedman et al., 2005). Their propensity to establish soon includes native sandbar willow (Salix exigua) and exotics Russian olive after a point bar surface is created means that channel migration can be (Elaeagnus angustifolia) and tamarisk (Tamarix ramosissima). Cotton- traced using their ages (Cooper et al., 2003; Stella et al., 2011; wood seedlings require abundant moisture and light, which is available Merigliano et al., 2013; Wilcox and Shafroth, 2013). on point bars after floodwaters subside. Cottonwood establishment Combining information from multiple sources diversifies the spatial pathways often produce tree ages similar to those of underlying point and temporal scales of channel change investigations (Friedman et al., bars (Everitt, 1968; Merigliano et al., 2013), although tree establish- 2015). We combined repeated channel cross sections, historical aerial ment may lag surface creation (Scott et al., 1997; Friedman and Griffin, photographs, and aged cottonwood transects to document channel 2017). Establishment near the channel leaves trees vulnerable to sub- migration across three timescales. This research builds on four decades sequent flood disturbance, but their survival is increased when the river of ongoing cross section measurements from the Powder River, Mon- migrates away during point bar growth (Auble and Scott, 1998). River tana, USA (Moody and Meade, 1990). The lack of direct flow regulation migration creates scroll bars, each populated by bands of cottonwoods on the Powder River makes it an ideal system to quantify the effects of orthogonal to lateral channel migration with generally increasing ages flow on channel migration. Our objectives were to (i) determine if farther from the channel. Although vegetative reproduction by root channel migration rates have changed through time on the free-flowing sprouting is common in some Populus species (Rood et al., 2007), it is Powder River, and (ii) investigate potential causes for any changes uncommon in P. deltoides. This produces relatively even-aged stands of identified. trees along the Powder River.

2. Study site 3. Methods

2.1. Powder River 3.1. The three approaches

The Powder River is one of the few remaining medium- to large- 3.1.1. Overview sized rivers in the USA that is relatively unaffected by major en- We combined repeated channel cross sections (XSs; spanning gineering structures inhibiting fluvial processes such as channel mi- 1975–2014), historical aerial photographs (APs; 1939–2013), and cot- gration (Gay et al., 1998). The river is 600 km long and drains tonwood transects (CTs; 1830–2014) to analyze channel migration at 34,700 km2 of the eastern Bighorn Mountains in Wyoming and the different timescales (Figs. 2, 3). The XS surveys had been repeated at western Great Plains in Wyoming and Montana. At USGS streamgage approximately annual intervals from 1975 to 1998 and less regularly #06324500, located 20 km upstream from the study reach, 1930–2010 from 1999 to 2014 (Moody and Meade, 1990, Supp. mat. Table A.1). mean annual flow was 12.5 m3/s and peak annual flow was 158.9 m3/s Ten AP sets were acquired to identify channel migration distances at 5- (Schook et al., 2016). Lake DeSmet is the large off-channel storage re- to 13-year intervals. We established 13 CTs at river meanders identified servoir, and 6% of its storage is consumed by irrigation annually, which as actively migrating during the AP record. The CTs were the least translates to 0.8% of Powder River mean annual flow at the study gage precise data set but provided the longest time perspective. Each ap- (HKM Engineering Inc., 2003). Although an estimated 23% of annual proach contained uncertainties that we discuss. Because the APs had

228 D.M. Schook et al. Geomorphology 293 (2017) 227–239

Fig. 1. The Powder River study reach in southeast Montana (black line in inset, 45.2° N, 105.7° W). The topographic cross sections (black lines) and cottonwood transects (green lines) are shown. The 2013 channel (blue) overlays channels in the air photo record (pink) and the floodplain (gray). Flow in the four river panels is from bottom to top and left to right. Pictures (A), (B), and (C) show the study reach. continuous spatial coverage throughout the reach, they were used to the previous year. The XSs were originally perpendicular to the standardize the place-based XS and CT measurements to reach-aver- channel, but channel migration altered orientations. We explored ad- aged rates. Air photo coverage limited the upstream extent of the justing the XS measurements so that migration distances were perpen- channel migration analysis to a subreach for which migration rates dicular to the channel, but changes through time were difficult to were calculated (PR130-PR194, Fig. 1). Two of 13 cottonwood transects characterize. Furthermore, 18 of the 20 XSs remained within 11° of and 4 of 20 cross sections were located in the full study reach but up to perpendicular to the channel, which would produce distance adjust- 10 km upvalley from the subreach. ments of < 2%. Therefore, such adjustments were omitted.

3.1.2. Cross sections (XSs) 3.1.3. Air photos (APs) Twenty-four XSs were established to monitor channel changes and Active channel boundaries were visually identified in APs from sediment budgets beginning in 1975 (Moody and Meade, 1990; Moody 10 years (1939, 1944, 1954, 1967, 1973, 1978 (post-flood), 1991, et al., 2002). They include straight reaches, crossovers, cutoffs, and 1998, 2005, and 2013; Supp. mat. Table A.2), and channel centerlines different positions along meander bends. The XSs were surveyed along were determined to calculate migration rates. Air photos through 1978 a tag line stretched between permanent reference points using a stadia were georeferenced using boundaries identified by Martinson and rod and level (Moody and Meade, 1990; Moody et al., 2002). We Meade (1983). We acquired 1978–2013 APs from public sources and analyzed 20 of the XSs located between PR116 and PR194 and calcu- georectified them to the 2013 NAIP imagery in ArcMap using lated migration rates by averaging changes in centerline location from a second-degree polynomial created from 12 to 18 control points

Fig. 2. All years used in analysis of channel migration rates from channel cross sections (1975–2014), air photos (1939–2013), and cottonwood transects (1830–2014).

229 D.M. Schook et al. Geomorphology 293 (2017) 227–239

Fig. 3. An example Powder River meander sequence. Polygons represent the channel location in all air photos from 1939 (light blue, bottom) to 2013 (purple, top). The polygons constrain maximum floodplain for a given loca- tion. A transect showing cottonwoods (green triangles) with their establishment dates extends from the channel edge, and cross section PR125 is the black line at left. Background image is the 1978 air photo, taken two months after a 50-year flood.

(RMSE ≤ 1.6 m) (Hughes et al., 2006). digging out each tree and taking multiple cores or slabs to find the We calculated channel planform statistics for each AP year. We germination point, but excavations would have been very time con- found mean channel width by dividing channel area by centerline suming considering that > 1.5 m of aggradation on portions of the length. We calculated slope by dividing channel length by the differ- Powder River floodplain in just 18 years (Moody et al., 1999). Instead ence in water surface elevations between PR130 and PR194, as iden- of excavating trees, we aged them using cores collected 1.3 m above tified in the 1978 XS surveys (Moody and Meade, 1990). The 1978 ground (Edmondson et al., 2014; Schook et al., 2016). This strategy elevation difference was suitable for calculating all AP channel slopes increased the number of trees that could be sampled by at least a factor because XS measurements showed it to vary by < 0.4 m through time of 10, but the ages identified usually lagged from the year of true es- over the 37.1-km study reach. Sinuosity was calculated by dividing tablishment. To quantify uncertainty associated with the lag, we iso- channel length by valley length. lated the 83 trees sampled from floodplain locations that had been To assess the consistency of two research teams delineating the channel at some point in the AP record (e.g., the top seven trees in channel for APs from 1939 to 1978 (Martinson and Meade, 1983) and Fig. 3). The surface beneath each tree was assumed to have been formed from 1978 to 2013 (our group), we acquired the 1978 APs and repeated in the middle year between the last AP in which the location was channel delineation. Consistency was high. Channel length, slope, and channel and the first AP in which it was floodplain. We estimated the sinuosity all differed by only 0.11%. Channel width differed by 4.2%. lag by subtracting tree age from surface age. An uncertainty analysis of the two 1978 channel centerline locations Tree ages that represent point bar growth by lateral migration revealed a median difference of 4.5 m (interquartile range of should increase or stay the same as distance from the channel increases 2.5–10.0 m). The 1978 APs were taken two months after a 50-year because surfaces farther from a laterally extending point bar cannot be flood that caused channel migration and avulsion, destroyed vegeta- younger than locations closer to the channel. For determining channel tion, and deposited fresh sediment on the floodplain. These factors migration rates, we adjusted the age of any tree that dated younger than made the 1978 channel boundary harder to identify compared to any the next tree closer to the channel on a CT by assigning it the same age others analyzed. Thus, we conclude that the delineations produced by as that tree. This was reasonable because APs showed all CTs to be the different researchers were compatible. Our 1978 delineation was located on laterally extending point bars. used in analyses.

3.1.4. Cottonwood transects (CTs) 3.2. Channel migration analysis From a randomly selected start location 1–40 m from the water's edge, we cored the closest tree (n = 189) at 40-m intervals on transects 3.2.1. Channel identification extending to the end of a scroll bar sequence. Two cores were collected We measured channel migration by identifying the Powder River's from each tree 1.3 m above ground using 4.5–12 mm diameter Haglӧf active channel boundary (Osterkamp and Hedman, 1982) for all three increment borers. Cores were mounted and sanded to 600 grit. We used data sets. The XS channel boundary was identified as a break in slope skeleton plots (Stokes and Smiley, 1968) and the program COFECHA located above a steeper bank. In years with APs and XSs, we cross- (Grissino-Mayer, 2001) for cross-dating. Not all trees had a core con- referenced them to calibrate delineations and to confirm intermethod taining pith (biological center) because of rot or imperfect aim in col- consistency. For APs we identified the left and right banks and used the lection, so we estimated the number of additional years to pith using Stream Restoration Toolbox to calculate channel centerlines (Lauer, established methods (Meko et al., 2015). All CTs were oriented per- 2012). The centerline kept the river as a contiguous unit throughout the pendicular to historic channel migration, so they were often curved reach instead of alternating sides as is necessary to measure cutbank or (Fig. 3). We selected meander bends with high migration rates in the AP point bar migration. We used the Stream Restoration Toolbox to cal- period because these locations had the most abundant and widest areas culate the lateral migration distance between centerlines at 25-m in- of cottonwoods establishing on extending point bars. crements. After measuring XS centerline migration, we analyzed cut- The CT migration rates were determined from tree locations and bank migration for the 14 XSs where a cutbank was clearly identifiable. timing of establishment, which generally occurred soon after bare se- This enabled analysis of channel migration by lateral erosion, whereas diment was deposited on point bars near the river. In order to precisely centerline migration conflated cutbank erosion and point-bar deposi- date establishment, trees must be aged at the height of germination tion (Miller and Friedman, 2009). (Scott et al., 1997; Holloway et al., 2017). This would have involved

230 D.M. Schook et al. Geomorphology 293 (2017) 227–239

3.2.2. Comparing migration across data sets 3.2.5. Floodplain turnover We expected biased channel migration rates from XS and CT surveys We used the 1978 flood delineation (Meade and Moody, 2013)to because sample locations were not randomly selected. Therefore, the define the 50-year floodplain. The extent of inundation throughout the spatially continuous APs were used to standardize XS and CT data for study reach was identified by (i) field mapping of deposited sediment cross-method comparisons and to enable all migration rates to describe and flood debris, (ii) topographic map analysis to interpolate field data, the PR130-PR194 subreach. We first created polygons between every and (iii) consultation with resident ranchers (Meade and Moody, 2013). pair (i.e., not just consecutive pairs) of AP channel centerlines and We georeferenced and digitized the 1978 flood map and used it to found a cumulative polygon area between each. Dividing this area by define the floodplain, which covered 36.44 km2 over 37.1 valley km. average centerline length of the two AP channels yielded a reach- Because of dikes, the floodplain area identified is likely 10–15% smaller averaged lateral migration distance between the two years. This ana- than its presettlement extent (R. Meade, pers. comm., 2016). lysis produced 45 migration distances for 5- to 74-year time intervals. A Floodplain turnover occurs when channel migration converts linear regression model was fit to the data (Supp. mat. Fig. A.1), and the channel to floodplain, and it is related to other metrics of floodplain slope characterized the reach-averaged AP migration rate for sediment storage including sediment age distributions, transit times, 1939–2013. We standardized migration rates identified in XSs and CTs and the erosion hazard (Bolin and Rodhe, 1973; Bradley and Tucker, by finding the ratio of AP migration at locations of each XS and CT 2013; Pizzuto et al., 2017). To investigate floodplain turnover rate, we compared to the whole reach AP migration rate. The AP migration was calculated the proportion of the floodplain that was overtaken by the faster at the XS (1.11×, 95% CI = 0.98–1.24×) and CT (2.20×, 95% channel during the AP record. This involved calculating what percen- CI = 2.08–2.32×) locations compared to the reach average. Therefore, tage of the floodplain had turned to river from 1944 (time = 5 years) all XS and CT migration distances were divided by 1.11 and 2.20, re- through 2013 (time = 74 years) by overlaying AP channels. spectively, to standardize their migration rates to the analysis reach. These adjustments corrected for nonrepresentative locations of XS and CT surveys but did not normalize the XS and CT migration rates to those 4. Results of the APs. We present channel migration rates in two ways. The first describes 4.1. Channel migration the average migration rate over the duration of each data set, which was identified as the slope of a linear model fit to all data in each data Linear models were created to describe channel migration rates fi set. The second migration rates characterize changes through time because they t the data well and provided easily interpretable com- within each data set. For this, AP migration rates were determined from parisons across methods and periods. Prominently, the slope of each successive photos. The XS data represent channel migration between model represented the average migration rate for each data set (Fig. 4). fi consecutively measured years. The CT data produced migration rates The analysis identi ed the XS record to have the slowest migration rate fi through time by finding differences in tree ages related to their distance (0.81 ± 0.04 m/year, 95% con dence interval, p < 0.001). In con- apart (methodological details in Supp. mat.). trast, the AP (1.52 ± 0.24 m/year) and CT (1.62 ± 0.06 m/year; Table 1) migration rates were faster and not significantly different from each other (z = 1.18, p = 0.24). 3.2.3. Flow and channel migration Cross-method comparisons were supported by migration rates being fl To investigate what parts of the ow regime cause channel migra- similar for XSs (0.81 ± 0.04 m/year for 1975–2014, 95% confidence fl tion, we compared ow exceedance probabilities for each period be- intervals included) and APs for their most overlapping period tween consecutive XS and AP measurements (Miller and Friedman, (0.63 ± 0.33 m/year for 1978–2013; Table 1). The difference is not fl 2009). We ranked all daily ow values between each pair of APs and statistically significant (z = 1.06, p = 0.36) and is small compared to fi fl identi ed the ow corresponding to each of the following exceedance the 66-m mean channel width during the period. The slightly higher XS- … … probabilities: 0.001, 0.002 0.009 and 0.01, 0.02 1. The same pro- calculated migration rate may have been inflated by XS surveys not fl cedure was carried out for daily ows between each pair of XS mea- being evenly distributed through time, as relatively few measurements surements that were taken in consecutive years. This generated a series after 1998 underrepresented this slower-moving period. We compared of discharges for various exceedance probabilities between measure- the migration rate for the 1939–2013 AP record to that from the 101 ments. A series of regressions were run to relate the flow for a given exceedance probability to the migration distance within the corre- sponding period. 400 Trees + Air Photos ff Cross−sections 3.2.4. E ects of measurement interval on migration rates 300 The frequently repeated and long duration XS surveys provided a unique opportunity to analyze channel migration as a function of in- terval between measurements. This exercise addressed a common 200 methodological issue in channel migration analyses that occurs when Migration (m) + + + either (i) data are collected at irregular intervals, or (ii) measurements + ++ + 100 + + + taken at longer intervals are scaled to annual rates. Issues arise because + + + + + + dividing multiyear measurements by the time between them ignores + + + + +++++ + +++ back-and-forth channel migration (O'Connor et al., 2003; Konrad, +++++++ 0 + 2012). Longer time intervals miss more back-and-forth migration, 0 50 100 150 200 which should produce an increasing underestimate of annual migration Interval (years) rate as interval increases. To address this, we used the XSs to calculate migration distance as a function of the interval between measurements. Fig. 4. Channel migration rates over the measurement periods for cottonwood transects – – – We differenced centerline locations in each year for each XS, thus (1830 2014), air photos (1939 2013), and cross sections (1975 2014). Each point dis- plays the average migration distance for paired measurements over a given time interval creating an all-possible-pairs data set where interval between mea- (x-axis). Lines are linear models. Slopes of lines for cottonwood tree and air photo data surements was largely divorced from calendar years. This provided a are not statistically different but they are both significantly greater than that for channel 4856-point data set to calculate channel migration over 1- to 39-year cross sections. Time interval for cottonwoods is the difference between tree ages, which is measurement intervals. rounded to decade-intervals here for display.

231 D.M. Schook et al. Geomorphology 293 (2017) 227–239

Table 1 Values and 95% confidence intervals (CI) of the slope (migration rate) from linear models of channel migration from different data sets and periods (AP = air photo, XS = cross section, CT = cottonwood transect).

Data set Location Period Applied to whole reach? Migration rate ± 95% CI (m/year)

AP Continuous 1939–2013 Yes 1.52 ± 0.24 AP Continuous 1978–2013 Yes 0.63 ± 0.33 AP At XSs 1939–2013 No 1.68 ± 0.24 AP At CTs 1939–2013 No 3.34 ± 0.28 XS Centerline 1975–2014 Yes 0.81 ± 0.04 XS Cutbank 1975–2014 Yes 0.88 ± 0.06 CT All 1830–2014 Yes 1.62 ± 0.06 CT All 1939–2013 Yes 1.72 ± 0.16

Fig. 5. Channel migration rates through time for the Powder River study reach. Continuous discharge measurements (blue) began in 1930, but the 1923 peak flow is documented as 3000 m3/s (value surpasses axis extent). Circles ± standard error are channel centerline migration rates from cross sections. Air photo line (black) is the centerline migration rate between consecutive photos. Cottonwood transect migration (green line) is a migration rate calculated from tree ages, and it is smoothed to depict uncertainty in lags between floodplain surface age and tree age. The decreasing number of transects with data back in time led to a truncation of results at 1860, and the increase after 2000 (dashed) is an artifact of the sampling technique. cottonwoods that established between these years. Dividing the dis- the XS measurements from the past four decades represent a period of tance between trees by their age difference produced a regression relatively slow channel migration (Fig. 5). equation with a slope of 1.72 ± 0.16 m/year, similar to the The highly repeated and precise XS surveys enabled comparison of 1.52 ± 0.24 m/year rate from the APs (z = 1.36, p = 0.17). Similar migration rates at variable time intervals, and we found longer intervals channel migration rates in both cross-method comparisons supported to increasingly underestimate annual migration rate (Fig. 6). The mean the three-method comparisons. 1-year channel migration distance calculated from all 337 measure- Combining the AP, XS, and CT data sets shows that Powder River ments (i.e., all consecutive-year measurements within each XS) was channel migration has slowed over the last century (Fig. 5). The CT data 1.34 m. Underestimates of this rate were caused by increasing amounts had maximum migration following the 1923 flood, peaking in the 1920s and 1930s at ~5 m/year (Supp. mat. Table A.5). Given the ex- treme magnitude of the 1923 flood (3000 m3/s), the known planform effects of the large 1978 flood (930 m3/s) (Moody and Meade, 1990; Gay et al., 1998), and a newspaper report from the time (Powder River County Examiner, 1923), we assume that the flood widened the channel and deposited sediment on the floodplain. These processes would fa- cilitate subsequent channel narrowing and cottonwood establishment. Note that the increase in CT-calculated migration rate since the year 2000 is an artifact of the methodology caused by herbivory that sup- pressed cottonwood growth and caused 10 trees to date to 2010–2014. The maximum migration rate between APs was 5.10 m/year for 1939–1944, the period closest to the 1923 flood. Similar migration rates were found for 1967–1973, a period in the middle of the record. This migration was largely caused by channel narrowing from 117 to 77 m that occurred after four ≥14-year floods in 1962–1967. Over the entire AP period, a decline in channel migration rate through time occurred. The AP-calculated channel migration was < 1.1 m/year in all three interphoto periods after 1991, reaching a minimum at 0.58 m/ year from 2005 to 2013 (Fig. 5; Supp. mat. Table A.6). Fig. 6. Channel migration distance (m) as a function of the number of years between Within the XS record, 1978 channel migration was highest, and measurements, calculated from all pairwise comparisons within each cross section. Point size is scaled to sample size, ranging from 337 samples at a 1-year interval to 5 samples at each of the three most-recent years (2012–2014) had among the lowest 39 years. Error bars are standard error. Dashed line is the expected migration distance migration. As a whole, the long-term AP and CT data sets indicate that extended from the 1-year interval rate.

232 D.M. Schook et al. Geomorphology 293 (2017) 227–239

probability were dominated by 1978, and this large flood strongly in- fluenced the correlations. The AP analysis characterized centerline migration over a longer period and had weaker correlations to discharge compared to XSs (Fig. 8). Correlation peaked at r = 0.59 for a flow exceedance of 0.09, which corresponded to the 33rd highest flow day in a year. The AP analysis showed that above-average flows were more closely related to channel migration compared to below-average flows, but it did not identify the highest < 0.01 proportion of flows as most influential, unlike the XS analysis and Miller and Friedman's (2009) AP analysis on a neighboring river. The longer time interval and more even distribu- tion of flood peaks in the AP record compared to the XS record reduced the influence of the 1978 flood. Any correlations below 0.4 were con- sidered weak, and all below-average flows had weak correlations to migration.

4.2. Channel profile and planform evolution

Fig. 7. Percent of 1939 floodplain remaining in each photo year through 2013. The black line is an exponential decay function that suggests the floodplain half-life is 178 years, but Flows on the Powder River changed between the pre- and post-1978 the biased fit illustrates that floodplain turnover did not conform to this null model. flood years (1930–1977 vs. 1979–2015). Although mean annual dis- charge decreased by only 9% (12.9 vs. 11.7 m3/s), the median peak 3 of back-and-forth migration at longer intervals. For example, a 2-year annual flow decreased by 48% (201 vs. 104 m /s). The planform geo- interval revealed migration at 1.15 m/year, a 10-year interval at metry of the Powder River changed over the AP record. Between 1939 1.01 m/year, and a 30-year interval at 0.86 m/year (Supp. mat. Table and 2013, a gradual increase in sinuosity from 1.55 to 2.01 corre- A.3). Small sample sizes and unequal contributions from different XSs sponded to a decrease in slope and width (Fig. 9; Supp. mat. Table A.4). caused a decline in migration distance at the longest intervals. All 4856 The channel length increased from 57 to 75 km within the reach, and measurements contributed to the 0.81 m/year migration rate for the slope declined proportionally from 0.0012 to 0.0009. Channel length, complete XS data set reported above. slope, and sinuosity maintained a consistent trajectory throughout the Over the 1939–2014 AP period, channel migration destroyed 24.8% photo record except for a minor reversal after the 1978 flood. Peak of the 50-year floodplain, leaving 75.2% of the floodplain intact. We fit discharge of the 1978 flood was 1.43 times greater than any other an exponential decay function to the data to identify a floodplain half- during the AP record and 2.60 times greater than any occurring after. life (i.e., the period when half of the initial floodplain is lost to the The flood temporarily shortened and widened the channel. river) and to assess the null model that all areas of the floodplain have an equal chance of being destroyed in each year. The calculated half-life 4.3. Cottonwood population dynamics was 178 years, but the model had a biased fit to the data (Fig. 7). The area of floodplain remaining decreased rapidly in the first 39 years of Tree ages ranged from 0 (tree present, but shorter than sampling the AP record, but it levelled off over the next 35 years, when peak height) to 184 years and increased with distance from the channel flows and channel migration were low. This suggests that the time to (Fig. 10). Age distribution suggested that establishment conditions erode 50% of the floodplain should be > 178 years if peak flows re- differed after the two largest floods of the last century, in 1923 and main low. 1978. While tree establishment remained high for two decades after the Channel migration rate was most highly correlated to flows with 1923 flood, the pulse of establishment after the smaller 1978 flood was low exceedance probabilities (i.e., high flows), and relationships limited to the four years after the flood (Fig. 11). Few were trees dated slightly differed for XSs and APs. For XSs, migration rate was best to soon before each flood, possibly because the floods killed or knocked correlated to the discharge having 0.005 exceedance, or the second down young trees near the channel. highest flow day of a year. Cutbank migration was more strongly pre- Eighty-three of the 189 sampled cottonwoods were located within dicted by this discharge compared to centerline migration (r = 0.98 vs. the zone that transitioned from channel to floodplain during the AP 0.87; Fig. 8). The XS calculations for flows of low exceedance record (Fig. 3). If tree age was the same as that of the underlying sur- face, all trees would have aged to between the two bounding photo

Fig. 8. Correlation between channel migration and flow exceedance probabilities within each measurement period. Cross section data are shown for both cutbank (black solid line) and centerline (red dotted line) migration. Air photo (turquoise dashed line) data Fig. 9. Channel width (pink, triangles) and length (blue, circles) throughout the air photo describe centerline migration between photos. All correlations below r = 0.4 (dashed record. Peak annual flow (gray) is included. Channel sinuosity increased and slope de- gray line) are considered weak. creased proportionally to channel length.

233 D.M. Schook et al. Geomorphology 293 (2017) 227–239

Fig. 10. Distance from the 2014 channel as a function of first year of tree growth. Wider point distributions after 1923 and 1978 represent widespread post-flood recruitment. years. However, median tree age lagged 8.5 years behind the middle year between the photos, with the interquartile range being a 5–13 year Fig. 12. Tree ages (n = 83 trees) displayed as a lag from the middle year between the two bounding air photos (i.e., the last photo where the tree's location was within the channel lag (Fig. 12). This small lag and relatively narrow range support the and the first photo where the location was floodplain). Black line at time = 0 is where hypothesis that tree establishment was generally limited to point bars. tree age equals the middle year between photos, and positive numbers mean trees were These results closely resemble the 7.3-year lag for cottonwood at the younger. nearby Little Missouri River in North Dakota (Friedman and Griffin, 2017) and the 5-year lag between germination point and ground sur- 5.1. Extrinsic controls on channel migration face at the Missouri River in Montana (Scott et al., 1997). Several factors may contribute to this lag. Tree seedling establishment may be Climatically induced changes to the hydrologic cycle likely con- delayed for years by flow fluctuations around the time of germination tributed to decreasing channel migration rates at the Powder River from (Benjankar et al., 2014); it takes years for seedlings to grow up to coring 1978 to 2014. A climate shift has decreased snowpack in the northern height; and damage from ice, beavers, deer, or livestock can suppress Rocky Mountains (Pederson et al., 2010), causing earlier snowmelt growth. discharges (Stewart et al., 2005). Peak flow has also decreased at the nearby Redwater (Merigliano et al., 2013) and Little Missouri (Griffin and Friedman, 2017), whose watersheds are low in elevation and are 5. Discussion unaffected by mountain snowmelt. Both of these rivers have experi- enced little flow regulation, and changes in flows were linked to in- We integrated three research approaches to document variable creases in temperature and evapotranspiration and decreased intensity channel migration rates on the Powder River since 1830. Combining of ice jams. The Powder River shares characteristics of mountain and multiple data sets overcame two problems common to fluvial geo- plains rivers because the flow regime is snowmelt-dominated, but most morphic research. The first is that it is difficult to characterize the of the watershed is in the Great Plains. processes driving channel migration because of the long and variable Beyond climate change, channel narrowing and decreased migra- timescales at which migration occurs. Combining XS, AP, and CT data tion can result from damming (Shafroth et al., 2002), stabilization of sets enabled assessment of channel migration over three nested and floodplain water table (Nadler and Schumm, 1981), exotic plant inva- complementary scales. For example, the high frequency and precision sion (Birken and Cooper, 2006), gravel mining (Dufour et al., 2015), of XS measurements captured responses such as the large migration in and decreases in bed sediment load (Williams and Wolman, 1984). the 1978 flood. The XS data also enabled quantification of the migra- However, none of these mechanisms appears to dominate on the free- tion missed for multiyear measurement intervals. On the other hand, flowing Powder River. The Powder River's trend of decreasing channel although the XS record was exceptionally long for such a record migration and width is consistent with other rivers in the region (39 years), XS data alone were not long enough to document long-term (Schumm and Lichty, 1963; Nadler and Schumm, 1981; Miller and changes in migration rate. In spite of their lower resolution, the AP and Friedman, 2009; Merigliano et al., 2013). This may be a response to the CT data sets were long enough in duration to document the strong largest flood in 1923, which straightened and widened the river decline in migration rate since the 1920s. Second, we also overcame the (Powder River County Examiner, 1923), and provided abundant bare common problem of applying data collected from discrete locations to a sediment on the floodplain for plants to colonize. Four large floods (1- larger reach. The spatially continuous APs were used to identify the in-14 to 1-in-33 year) in the 1960s may have prolonged the straighter relative magnitude of channel migration over the AP period at XSs and channel, but recent small floods have caused a transition of the channel CTs, allowing us to compare channel migration across data sets. planform (Fig. 13). People of European descent began to colonize the Powder River watershed in the late nineteenth century, and changing land use likely

Fig. 11. Unadjusted establishment dates for all trees sampled on the cottonwood transects. Groups of trees established after the 1923 and 1978 floods.

234 D.M. Schook et al. Geomorphology 293 (2017) 227–239

Fig. 13. Conceptual model of the evolution of the Powder River. Changes in channel width (w), depth (d), and slope (s) are relative increases (+) or decreases (−) from that of the baseline sinuous river (Schumm, 1969). Big floods play a critical role in changing river planform, reworking floodplain surfaces, and promoting heterogeneous floodplain forests. Without big floods to rejuvenate the fluvial system, channel migration slows, channel complexity decreases, cottonwood recruitment decreases, and shade-tolerant and upland species replace cottonwood forests. has affected river planform geometry and channel migration. The peak flows through time (Clayton, 2009). What is clear is that XS, AP, Powder River headwaters are in the Bighorn Mountains, which have and CT records show patterns consistent with those from rivers re- experienced periods of heavy livestock use (peaking in the 1890s), tie sponding to large floods. drives (1890s–1910s), clearcutting (peaking in the 1960s), and forest Although the concept of a floodplain half-life has been effectively fires (peaking in the 1920s) (Winters et al., 2004). Each of these pro- used as a null model to describe floodplain turnover (Gottesfeld and cesses could increase the river's sediment load, pushing it toward the Gottesfeld, 1990; O'Connor et al., 2003), two assumptions this ex- wider and less sinuous geometry characteristic of the earlier parts of the ponential decay model holds are that all locations are equally likely to AP period. Land uses within the floodplain itself have changed through be destroyed in a given year and floodplain destruction rate is constant time as well. Homesteaders colonized the study reach's floodplain at the through time. Although meanders move away from point bars and start of the twentieth century. Our CT results suggested very little preferentially destroy old floodplain (Hickin and Nanson, 1975), em- channel migration in the twentieth century before the 1923 flood, but pirical and modeling efforts have shown fluctuations in channel width the arrival of the homesteaders and their cottonwood-foraging livestock that cause preferential destruction of younger floodplain areas (Miller could have coupled with the destructive flood to leave few surviving and Friedman, 2009; Konrad, 2012), and numerical modeling has and datable young trees. shown the active meander belt to occupy a limited width close to the modern channel (Xu et al., 2011; Bradley and Tucker, 2013). Some floodplain areas may be protected from erosion by large clasts, log 5.2. Channel adjustments jams, cohesive sediment, or valley position (O'Connor et al., 2003; Wohl, 2013). The floodplain half-life we calculated over the AP record fl Because geomorphic processes can respond to large oods over is time-period dependent because of hydroclimatic changes, human fl fl decades, river measurements re ect ow history. The chronology of impacts, and floods with prolonged effects. The decrease in Powder fl channel migration through time we documented was heavily in uenced River peak flows has reduced erosion rates causing deviation from ex- fl – by large oods in 1923, 1962 1967, and 1978, and cottonwood es- ponential (Everitt, 1968) or square root (Bradley and Tucker, 2013) fl fl tablishment dates re ected this (Fig. 11). The 1978 ood has been re- decay. Because such nonstationarity of flow is widespread across rivers, cognized as the catalyst for many Powder River changes over the past floodplain turnover rates calculated for any river should be cautiously four decades (Gay et al., 1998; Moody et al., 1999; Moody and Meade, interpreted. fl 2008, 2014; Meade et al., 2013). The ood widened the channel, which Channel centerline migration is a function of erosion and deposi- facilitated bench formation, vertical and lateral accretion, and channel tion, but these two processes are driven by different flows. Although narrowing (Pizzuto, 1994; Moody et al., 1999). These processes dee- erosion is commonly caused by high flows, deposition has been shown fl fl pened the channel and raised the oodplain, which increased the ow to be inversely correlated to the magnitude of low flows (Miller and fl magnitude required for overbank ows by a factor of 3.7 between 1978 Friedman, 2009). Our XS and AP analyses were based on movement of fl and 1996 (Moody et al., 1999). The 1923 ood was three times larger the channel centerline, but centerlines are calculated from both banks than that of 1978 and may have catalyzed responses that persisted and therefore affected by erosion and deposition. This combination of throughout the AP period. It is unclear exactly how the channel mi- processes lowered correlations between migration and high flows – ff gration patterns documented in the 1939 2013 AP record were a ected compared to AP analyses focused on erosion alone (Miller and by one extreme flood in 1923 or a general tendency toward decreasing

235 D.M. Schook et al. Geomorphology 293 (2017) 227–239

Friedman, 2009). We extracted the subset of XSs having a clear cutbank varies predictably across settings (Scott et al., 1996). For example, most to quantify differences between migration of centerlines and cutbanks, cottonwood recruitment on the Missouri River floodplain constrained and as expected cutbank migration was more strongly correlated to by the Missouri Breaks, Montana, was associated with overbank de- high flows than centerline migration. We found high between-photo position from large floods (Scott et al., 1997). However, farther migration for 1967–1973, a period when channel width reduced by downstream and at other Great Plains rivers, cottonwoods established 34% following four of the five largest floods in the gage record. Pre- from channel narrowing caused by flow regulation (Friedman et al., ferential narrowing from one bank was likely responsible for much of 1998; Dixon et al., 2012). Spatial patterns of tree ages at the Powder the documented channel migration. River suggested establishment through channel migration and nar- rowing. The high tree recruitment in the 1920s and 1930s occurred 5.3. Methodological uncertainties during a period of moderate floods on a channel that had been straightened and widened by the 1923 flood (Powder River County Each method revealed channel migration through time (Fig. 5), but Examiner, 1923). This combination of geomorphic and hydrologic various factors affected the precision of migration rates. For example, conditions facilitated channel migration and tree establishment, which the linear models in Fig. 4 provide easily interpretable migration rates, would in turn stabilize point bars, narrow and deepen the channel, and but each model revealed error by having a positive y-intercept, sug- focus flow energy on erodible cutbanks (Nadler and Schumm, 1981). gesting that migration occurred with no time passing. This was partly a Cottonwood ages indicate that a period of low channel migration function of back-and-forth channel migration being missed as mea- was followed by high migration after the 1923 flood. However, the surement interval increases. However, under the scenario in which no 1923 flood could have killed young cottonwoods (Friedman and Lee, migration occurs, the APs would likely indicate some migration caused 2002) that established in the early 1900s, precluding surface dating by imprecise georectification and bank identification that is compli- with them. Alternatively, young trees may have been knocked down in cated by variable river stages, interphoto intervals, and vegetation the flood before resprouting (Sigafoos, 1964). The resprout scenario patterns. This was illustrated by the 1978 APs, where we found a 4.5-m was not widely captured in the sample, however, because only six trees difference in channel locations between the two research teams. Other dated to within the two years after the 1923 flood. The flood likely researchers have quantified portions of AP uncertainty (Hughes et al., aggraded the channel and lowered the flow magnitude required for 2006; Lea and Legleiter, 2016), but a general framework for evaluating floodplain inundation, as occurred after the 1978 flood (Moody et al., uncertainty remains elusive. Recent advancements in frequently col- 1999). An even larger flood was rumored to have happened in 1887 lected high resolution satellite imagery create opportunity for such (Powder River County Examiner, 1923). We did not detect a pulse in investigations. cottonwood establishment after 1887, but a Powder River flow re- Calculating annual migration rates from data collected over longer construction that included ring widths from the same cottonwoods intervals is susceptible to systematic errors, and researchers have ac- analyzed here identified 1887 as the third wettest year since 1830 knowledged potential underestimates of annual channel migration rates (Schook et al., 2016). when using APs with multiyear intervals (O'Connor et al., 2003; Miller and Friedman, 2009). Underestimation with increasing interval also occurs for sedimentation (Sadler, 1981), channel incision (Mills, 2000), 5.5. Management implications tectonic uplift (Gardner et al., 1987), and bedrock erosion (Finnegan et al., 2014). Our XS measurements revealed decreasing migration rates In demonstrating that Powder River cottonwoods establish from at increasing intervals (Fig. 6). We could have used the XS data to channel migration and point bar formation, results suggest a future correct AP migration rates based on photo intervals, but we lacked a with reduced regeneration of native floodplain forests if dynamic concomitant CT adjustment because of their longer intervals. Therefore, flooding flows do not continue. Across western rivers, a general tran- the AP data retain an underestimate from longer than annual time in- sition from cottonwood forests to invasive, shade-tolerant, and upland tervals between photos that counters overestimates from georectifica- species is occurring (Merritt and Cooper, 2000; Friedman et al., 2005; tion, but relative magnitudes are unknown. Andersen et al., 2007; Merritt and Poff, 2010), and preserving flow The CT data covered the longest duration, but tree recruitment regimes on free-flowing rivers has one of the best chances to resist a processes led to this method having the greatest uncertainty. Instead of region-wide habitat loss. We found an 8.5-year lag between floodplain trees establishing immediately and reliably in narrow bands after formation and cottonwood establishment, but evolving conditions will decimeter-to-meter-scale annual migration, successful recruitment modify this relationship. Along one sampled CT, the first 70 m from the might only occur every several years (Scott et al., 1997; Friedman and river had few cottonwoods and no trees that had grown above herbi- Lee, 2002). We found an 8.5-year delay between surface creation and vore browsing height, even though APs showed 50 m of the stretch had tree age. An additional source of error is that tree ages might be up to a been floodplain since at least 1991. A possible mechanism explaining couple years off if the core did not contain pith. Pulses of recruitment decreased cottonwood recruitment is that slow channel migration en- occur, which can lead to even-aged bands tens of meters wide abled establishment of a relatively small number of cottonwoods, (Merigliano et al., 2013). This pattern was evident at the Powder River whose growth has been suppressed by ice, competition with Russian where sampled trees ≥40 m apart sometimes belonged to the same olive, and herbivory from deer, beaver, sheep, and cattle. Similar sup- cohort. Cottonwood recruitment depends on light availability, soil pression has occurred in Wyoming's Yellowstone National Park, where moisture, rate of water table decline, subsequent floods, and herbivory extreme floods with high cottonwood recruitment are required for (Mahoney and Rood, 1998; Benjankar et al., 2014; Rose and Cooper, seedlings to escape bison and elk herbivory (Rose and Cooper, 2016). 2016). At the Powder River, variable browse pressures through space Many rivers are dammed, and managers often remove the highest and time likely left a signal in the CT data that influenced channel flows to replenish reservoirs and protect downstream property. migration rates. However, a consequence of this action is a decrease in channel migra- tion and floodplain sedimentation and a decrease in regeneration of 5.4. Channel migration and floodplain forests disturbance-dependent riparian species including cottonwoods. If not holistically managed, climatic and anthropogenic changes to flow may We sampled cottonwoods from healthy, multiaged forests on induce cascading effects, including habitat loss from vegetation en- meander bends of the Powder River. Cottonwoods can establish from croachment, replacement of cottonwood forests, and decreases in ha- channel narrowing, meandering, and flood deposition within a single bitat complexity and biodiversity. site (Cooper et al., 2003), and the relative influence of these processes

236 D.M. Schook et al. Geomorphology 293 (2017) 227–239

6. Conclusion Benjankar, R., Burke, M., Yager, E., Tonina, D., Egger, G., Rood, S.B., Merz, N., 2014. Development of a spatially-distributed hydroecological model to simulate cotton- wood seedling recruitment along rivers. J. Environ. Manag. 145, 277–288. http://dx. Three data sets suggest that Powder River channel migration has doi.org/10.1016/j.jenvman.2014.06.027. decreased through time, and the recent history (1978–2014) is not re- Birken, A.S., Cooper, D.J., 2006. Processes of Tamarix invasion and floodplain develop- fl ment along the lower Green River, Utah. Ecol. Appl. 16, 1103–1120. presentative of uvial geomorphic processes from the past two cen- Bolin, B., Rodhe, H., 1973. A note on the concepts of age distribution and transit time in turies. Even though the Powder River is predominantly a snowmelt- natural reservoirs. Tellus 25, 58–62. http://dx.doi.org/10.1111/j.2153-3490.1973. dominated river where variability of interannual floods is relatively tb01594.x. low, the largest two floods were caused by a frontal storm (1923) and a Bradley, D.N., Tucker, G.E., 2013. The storage time, age, and erosion hazard of laterally accreted sediment on the floodplain of a simulated meandering river. J. Geophys. rain-on-snow event (1978). Both floods induced years-to-decades of Res. Earth Surf. 118, 1308–1319. http://dx.doi.org/10.1002/jgrf.20083. channel responses. The periods of highest channel migration were in Brandt, S.A., 2000. Classification of geomorphological effects downstream of dams. – the two decades following the 1923 flood and in the decade following Catena 40, 375 401. http://dx.doi.org/10.1016/S0341-8162(00)00093-X. fl fl Cadol, D., Rathburn, S.L., Cooper, D.J., 2011. Aerial photographic analysis of channel oods in the early 1960s. The relatively small peak annual ows from narrowing and vegetation expansion in Canyon De Chelly National Monument, the past three decades have contributed to a gradual trend of decreasing Arizona, USA, 1935–2004. River Res. Appl. 27, 841–856. http://dx.doi.org/10.1002/ channel migration since the 1920s. rra.1399. Chase, K.J., 2013. Streamflow statistics for unregulated and regulated conditions for se- This study can inform the selection and interpretation of research lected locations on the Yellowstone, Tongue, and Powder Rivers, Montana, methods in fluvial geomorphology, and findings from the Powder River 1928–2002. In: Scientific Investigations Report No. 2013–5173. U.S. Geological can aid in interpretation of channel migration across geographic and Survey. Clayton, J.A., 2009. Geomorphic adjustment, geographic context and disturbances. In: hydroclimatic settings. Direct comparison of the three approaches re- Geomorphology and Plate Tectonics. Nova Science Publishers, Inc., New York, pp. vealed that even four decades of highly repeated and precise cross- 97–116. sectional surveys may not well characterize the century-scales over Clow, D.W., 2010. Changes in the timing of snowmelt and streamflow in Colorado: a response to recent warming. J. Clim. 23, 2293–2306. http://dx.doi.org/10.1175/ which channel migration occurs. Additionally, aerial photos are often 2009JCLI2951.1. used alone to describe channel migration rates, but they may bias true Cooper, D.J., Andersen, D.C., Chimner, R.A., 2003. Multiple pathways for woody plant annual migration rates because of errors in georectification, difficulty in establishment on floodplains at local to regional scales. J. Ecol. 91, 182–196. ff consistent identification of channel boundaries, and the inability of Costa, M.H., Botta, A., Cardille, J.A., 2003. E ects of large-scale changes in land cover on the discharge of the Tocantins River, Southeastern Amazonia. J. Hydrol. 283, multiyear intervals to detect back-and-forth channel migration. Annual 206–217. http://dx.doi.org/10.1016/S0022-1694(03)00267-1. migration rates commonly are the standard when assessing geomorphic Dixon, M.D., Johnson, W.C., Scott, M.L., Bowen, D.E., Rabbe, L.A., 2012. Dynamics of processes, but converting measurements taken at multiyear intervals to plains cottonwood (Populus deltoides) forests and historical landscape change along unchannelized segments of the Missouri River, USA. Environ. Manag. 49, 990–1008. an annual rate should be conducted with consideration of the spatial http://dx.doi.org/10.1007/s00267-012-9842-5. and temporal scales at which the process varies. Ultimately, fluvial Dufour, S., Rinaldi, M., Piégay, H., Michalon, A., 2015. How do river dynamics and ff human influences affect the landscape pattern of fluvial corridors? Lessons from the geomorphic processes are a ected by multilayered, complex, and de- – – fl Magra River, Central Northern Italy. Landsc. Urban Plan. 134, 107 118. http://dx. layed responses induced by ows, climate, and land uses. Therefore, doi.org/10.1016/j.landurbplan.2014.10.007. integrating multiple research approaches can build a stronger under- Edmondson, J., Friedman, J., Meko, D., Touchan, R., Scott, J., Edmondson, A., 2014. standing of the processes causing channel migration. Dendroclimatic potential of plains cottonwood (Populus deltoides subsp. monilifera) from the Northern Great Plains, USA. Tree-Ring Res. 70, 21–30. http://dx.doi.org/10. 3959/1536-1098-70.1.21. Acknowledgements Erskine, W., McFadden, C., Bishop, P., 1992. Alluvial cutoffs as indicators of former channel conditions. Earth Surf. Process. Landf. 17, 23–37. http://dx.doi.org/10. 1002/esp.3290170103. We thank John Moody and Bob Meade for their essential role in Erskine, W., Keene, A., Bush, R., Cheetham, M., Chalmers, A., 2012. Influence of riparian introducing us to the Powder River study site, providing us with cross vegetation on channel widening and subsequent contraction on a sand-bed stream sectional data, and for conversations about the Powder River study since European settlement: Widden Brook, Australia. Geomorphology 147–148, – system. John Moody, Jim Pizzuto, Jordan Clayton, Richard Marston 102 114. http://dx.doi.org/10.1016/j.geomorph.2011.07.030. Everitt, B.L., 1968. Use of the cottonwood in an investigation of the recent history of a and four anonymous reviewers provided insightful and detailed reviews flood plain. Am. J. Sci. 266, 417–439. http://dx.doi.org/10.2475/ajs.266.6.417. that greatly improved the manuscript. Brendan Elba helped collect Favaro, E.A., Lamoureux, S.F., 2015. Downstream patterns of suspended sediment fl cottonwood cores, and Fisher Ankney and Brendan Elba helped with transport in a High Arctic river in uenced by permafrost disturbance and recent climate change. Geomorphology 246, 359–369. http://dx.doi.org/10.1016/j. tree ring lab analyses. Thanks to the many ranchers who granted us geomorph.2015.06.038. access to their land, including the Bowers, Gay, Held, Huckins, Randall, Finnegan, N.J., Schumer, R., Finnegan, S., 2014. A signature of transience in bedrock – – and Wilson families. Research was supported by the U.S. Geological river incision rates over timescales of 104 107 years. Nature 505, 391 394. http:// fi dx.doi.org/10.1038/nature12913. Survey, Colorado Scienti c Society, Colorado State University Florsheim, J.L., Mount, J.F., Chin, A., 2008. Bank erosion as a desirable attribute of rivers. Department of Geosciences, and National Science Foundation IGERT Bioscience 58, 519–529. Grant No. DGE-0966364 ‘I-WATER: Integrated Water, Atmosphere, Fonstad, M.A., Dietrich, J.T., Courville, B.C., Jensen, J.L., Carbonneau, P.E., 2013. ’ Topographic structure from motion: a new development in photogrammetric mea- Ecosystem Education and Research Program at Colorado State surement. Earth Surf. Process. Landf. 38, 421–430. http://dx.doi.org/10.1002/esp. University. Any use of trade, firm, or product names is for descriptive 3366. purposes only and does not imply endorsement by the U.S. Fremier, A.K., Girvetz, E.H., Greco, S.E., Larsen, E.W., 2014. Quantifying process-based mitigation strategies in historical context: separating multiple cumulative effects on Government. river meander migration. PLoS ONE 9. http://dx.doi.org/10.1371/journal.pone. 0099736. Appendix A. Supplementary data Friedman, J.M., Griffin, E.R., 2017. Management of Plains Cottonwood at Theodore Roosevelt National Park, North Dakota (No. NPS/THRO/NRR-2017/1395). U.S. Department of the Interior, National Park Service, Fort Collins, CO. Supplementary data to this article can be found online at http://dx. Friedman, J.M., Lee, V.J., 2002. Extreme floods, channel change, and riparian forests doi.org/10.1016/j.geomorph.2017.06.001. along ephemeral streams. Ecol. Monogr. 72, 409–425. Friedman, J.M., Osterkamp, W.R., Scott, M.L., Auble, G.T., 1998. Downstream effects of dams on channel geometry and bottomland vegetation: regional patterns in the Great References Plains. Wetlands 18, 619–633. Friedman, J.M., Auble, G.T., Shafroth, P.B., Scott, M.L., Merigliano, M.F., Freehling, M.D., ffi Andersen, D.C., Cooper, D.J., Northcott, K., 2007. Dams, floodplain land use, and riparian Gri n, E.R., 2005. Dominance of non-native riparian trees in western USA. Biol. – forest conservation in the semiarid Upper Colorado River Basin, USA. Environ. Invasions 7, 747 751. http://dx.doi.org/10.1007/s10530-004-5849-z. ffi Manag. 40, 453–475. http://dx.doi.org/10.1007/s00267-006-0294-7. Friedman, J.M., Vincent, K.R., Gri n, E.R., Scott, M.L., Shafroth, P.B., Auble, G.T., 2015. fi Auble, G.T., Scott, M.L., 1998. Fluvial disturbance patches and cottonwood recruitment Processes of arroyo lling in northern New Mexico, USA. Geol. Soc. Am. Bull. 127, – along the upper Missouri River, Montana. Wetlands 18, 546–556. http://dx.doi.org/ 621 640. 10.1007/BF03161671. Gardner, T.W., Jorgensen, D.W., Shuman, C., Lemieux, C.R., 1987. Geomorphic and

237 D.M. Schook et al. Geomorphology 293 (2017) 227–239

tectonic process rates: effects of measured time interval. Geology 15, 259–261. change in a free-evolving, small Alpine river: Ridanna Creek (North East Italy). Earth http://dx.doi.org/10.1130/0091-7613(1987)15<259:GATPRE>2.0.CO;2. Surf. Process. Landf. 32, 2104–2119. http://dx.doi.org/10.1002/esp.1511. Gay, G.R., Gay, H.H., Gay, W.H., Martinson, H.A., Meade, R.H., Moody, J.A., 1998. Mahoney, J.M., Rood, S.B., 1998. Streamflow requirements for cottonwood seedling re- Evolution of cutoffs across meander necks in Powder River, Montana, USA. Earth cruitment—an integrative model. Wetlands 18, 634–645. http://dx.doi.org/10.1007/ Surf. Process. Landf. 23, 651–662. http://dx.doi.org/10.1002/(SICI)1096- BF03161678. 9837(199807)23:7<651::AID-ESP891>3.0.CO;2-V. Martinson, H.A., Meade, R.H., 1983. Channel changes of Powder River, 1938–1978. In: Gottesfeld, A.S., Gottesfeld, L.M.J., 1990. Floodplain dynamics of a wandering river, No. HA-661. U.S. Geological Survey Hydrologic Investigations Atlas, Powder River dendrochronology of the Morice River, British Columbia, Canada. Geomorphology 3, County, Montana. 159–179. Meade, R.H., Moody, J.A., 2013. Erosional and depositional changes wrought by the flood Griffin, E.R., Friedman, J.M., 2017. Decreased runoff response to precipitation, Little of May 1978 in the channels of Powder River, southeastern Montana. In: USGS Missouri River Basin, northern Great Plains, USA. J. Am. Water Resour. Assoc. http:// Numbered Series No. 2013–5035. Scientific Investigations Report U.S. Geological dx.doi.org/10.1111/1752-1688.12517. Survey, Reston, VA. Grissino-Mayer, H.D., 2001. Evaluating crossdating accuracy: a manual and tutorial for Meade, R.H., Moody, J.A., Martinson, H.A., 2013. Inundated areas and channel changes the computer program COFECHA. Tree-Ring Res. 57, 205–221. wrought by the flood of May 1978 in Powder River between Moorhead and Broadus, Grover, N.C., 1925. Contributions to the hydrology of the United States, 1923–1924. In: Southeastern Montana. In: Scientific Investigations Report No. 2013–5035. U.S. No. Water Supply Paper 520. USGS, Washington, D.C.. Geological Survey. Güneralp, İ., Abad, J.D., Zolezzi, G., Hooke, J., 2012. Advances and challenges in Meitzen, K.M., Doyle, M.W., Thoms, M.C., Burns, C.E., 2013. Geomorphology within the meandering channels research. Geomorphology 163–164, 1–9. http://dx.doi.org/10. interdisciplinary science of environmental flows. Geomorphology 200, 143–154. 1016/j.geomorph.2012.04.011. (Meandering Channels). http://dx.doi.org/10.1016/j.geomorph.2013.03.013. (The Field Tradition in Gurnell, A.M., Bertoldi, W., Corenblit, D., 2012. Changing river channels: the roles of Geomorphology 43rd Annual Binghamton Geomorphology Symposium, held 21–23 hydrological processes, plants and pioneer fluvial landforms in humid temperate, September 2012 in Jackson, Wyoming USA). mixed load, gravel bed rivers. Earth-Sci. Rev. 111, 129–141. http://dx.doi.org/10. Meko, D.M., Friedman, J.M., Touchan, R., Edmondson, J.R., Griffin, E.R., Scott, J.A., 1016/j.earscirev.2011.11.005. 2015. Alternative standardization approaches to improving streamflow reconstruc- Hembree, C.H., Colby, B.R., Swenson, H.A., Davis, J.R., 1952. Sedimentation and che- tions with ring-width indices of riparian trees. The Holocene 25, 1093–1101. http:// mical quality of water in the Powder River Drainage Basin, Wyoming and Montana. dx.doi.org/10.1177/0959683615580181. In: USGS Numbered Series No. 170. Circular U.S. Geological Survey. Merigliano, M.F., Friedman, J.M., Scott, M.L., 2013. 12.10 tree-ring records of variation Hickin, E.J., Nanson, G.C., 1975. The character of channel migration on the Beatton in flow and channel geometry. In: Shroder, J.F. (Ed.), Treatise on Geomorphology. River, northeast British Columbia, Canada. Geol. Soc. Am. Bull. 86, 487–494. Academic Press, San Diego, pp. 145–164. HKM Engineering Inc, 2003. Lake DeSmet Level II Master Plan and Reservoir Merritt, D.M., Cooper, D.J., 2000. Riparian vegetation and channel change in response to Rehabilitation Plan (Final Report). (Sheridan, WY). river regulation: a comparative study of regulated and unregulated streams in the Hobo, N., Makaske, B., Wallinga, J., Middelkoop, H., 2014. Reconstruction of eroded and Green River Basin, USA. Regul. Rivers Res. Manag. 16, 543–564. http://dx.doi.org/ deposited sediment volumes of the embanked River Waal, the Netherlands, for the 10.1002/1099-1646(200011/12)16:6<543::AID-RRR590>3.0.CO;2-N. period AD 1631–present. Earth Surf. Process. Landf. 39, 1301–1318. http://dx.doi. Merritt, D.M., Poff, N.L.R., 2010. Shifting dominance of riparian Populus and Tamarix org/10.1002/esp.3525. along gradients of flow alteration in western North American rivers. Ecol. Appl. 20, Holloway, J.V., Rillig, M.C., Gurnell, A.M., 2017. Underground riparian wood: buried 135–152. http://dx.doi.org/10.1890/08-2251.1. stem and coarse root structures of Black Poplar (Populus nigra L.). Geomorphology Milan, D.J., Heritage, G.L., Hetherington, D., 2007. Application of a 3D laser scanner in 279, 188–198. http://dx.doi.org/10.1016/j.geomorph.2016.08.002. (Dynamics and the assessment of erosion and deposition volumes and channel change in a proglacial ecology of Wood in World Rivers). river. Earth Surf. Process. Landf. 32, 1657–1674. http://dx.doi.org/10.1002/esp. Hooke, J., 1980. Magnitude and distribution of rates of river bank erosion. Earth Surf. 1592. Process. Landf. 5, 143–157. http://dx.doi.org/10.1002/esp.3760050205. Miller, J.R., Friedman, J.M., 2009. Influence of flow variability on floodplain formation Hooke, J.M., 2007. Spatial variability, mechanisms and propagation of change in an and destruction, Little Missouri River, North Dakota. Geol. Soc. Am. Bull. 121, active meandering river. Geomorphology 84, 277–296. http://dx.doi.org/10.1016/j. 752–759. http://dx.doi.org/10.1130/B26355.1. geomorph.2006.06.005. (Geomorphic Instability and Change - Introduction: Mills, H.H., 2000. Apparent increasing rates of stream incision in the eastern United Implications of temporal and spatial scales Joint International Geomorphology States during the late Cenozoic. Geology 28, 955–957. http://dx.doi.org/10.1130/ Conference (JIGC)). 0091-7613(2000)28<955:AIROSI>2.0.CO;2. Howard, A.D., Knutson, T.R., 1984. Sufficient conditions for river meandering: a simu- Moody, J.A., Meade, R.H., 1990. Channel changes at cross sections of the Powder River lation approach. Water Resour. Res. 20, 1659–1667. http://dx.doi.org/10.1029/ between Moorhead and Broadus, Montana, 1975–88. In: USGS Numbered Series No. WR020i011p01659. 89–407. Open-File Report U.S. Geological Survey; Books and Open-File Reports Hudson, P.F., Kesel, R.H., 2000. Channel migration and meander-bend curvature in the Sections (distributor). lower Mississippi River prior to major human modification. Geology 28, 531–534. Moody, J.A., Meade, R.H., 2008. Terrace aggradation during the 1978 flood on Powder http://dx.doi.org/10.1130/0091-7613(2000)28<531:CMAMCI>2.0.CO;2. River, Montana, USA. Geomorphology 99, 387–403. http://dx.doi.org/10.1016/j. Hughes, F.M.R., 1997. Floodplain biogeomorphology. Prog. Phys. Geogr. 21, 501–529. geomorph.2007.12.002. http://dx.doi.org/10.1177/030913339702100402. Moody, J.A., Meade, R.H., 2014. Ontogeny of point bars on a river in a cold semi-arid Hughes, M.L., McDowell, P.F., Marcus, W.A., 2006. Accuracy assessment of georectified climate. Geol. Soc. Am. Bull. 126, 1301–1316. http://dx.doi.org/10.1130/B30992.1. aerial photographs: implications for measuring lateral channel movement in a GIS. Moody, J.A., Pizzuto, J.E., Meade, R.H., 1999. Ontogeny of a flood plain. Geol. Soc. Am. Geomorphology 74, 1–16. http://dx.doi.org/10.1016/j.geomorph.2005.07.001. Bull. 111, 291–303. http://dx.doi.org/10.1130/0016-7606(1999) Johnson, W.C., 1994. Woodland expansions in the Platte River, Nebraska: patterns and 111<0291:OOAFP>2.3.CO;2. causes. Ecol. Monogr. 64, 45–84. http://dx.doi.org/10.2307/2937055. Moody, J.A., Meade, R.H., Martinson, H.A., 2002. Erosion and deposition of sediment at Johnson, W.C., Dixon, M.D., Scott, M.L., Rabbe, L., Larsen, G., Volke, M., Werner, B., channel cross sections on Powder River between Moorhead and Broadus, Montana, 2012. Forty years of vegetation change on the Missouri River floodplain. Bioscience 1980–98. In: USGS Numbered Series No. 2002–4219. Water-Resources Investigations 62, 123–135. http://dx.doi.org/10.1525/bio.2012.62.2.6. Report. Knox, J.C., 2000. Sensitivity of modern and Holocene floods to climate change. Quat. Sci. Nadler, C.T., Schumm, S.A., 1981. Metamorphosis of South Platte and Arkansas Rivers, Rev. 19, 439–457. Eastern Colorado. Konrad, C.P., 2012. Reoccupation of floodplains by rivers and its relation to the age Nanson, G.C., 1986. Episodes of vertical accretion and catastrophic stripping: a model of structure of floodplain vegetation. J. Geophys. Res. Biogeosci. 117, G00N13. http:// disequilibrium flood-plain development. Geol. Soc. Am. Bull. 97, 1467–1475. http:// dx.doi.org/10.1029/2011JG001906. dx.doi.org/10.1130/0016-7606(1986)97<1467:EOVAAC>2.0.CO;2. van de Lageweg, W.I., van Dijk, W.M., Baar, A.W., Rutten, J., Kleinhans, M.G., 2014. Bank Nanson, G., Hickin, E., 1983. Channel migration and incision on the Beatton River. J. pull or bar push: what drives scroll-bar formation in meandering rivers? Geology 42, Hydraul. Eng. 109, 327–337. http://dx.doi.org/10.1061/(ASCE)0733-9429(1983) 319–322. http://dx.doi.org/10.1130/G35192.1. 109:3(327). Larsen, E.W., Fremier, A.K., Girvetz, E.H., 2006. Modeling the effects of variable annual Nilsson, C., Reidy, C.A., Dynesius, M., Revenga, C., 2005. Fragmentation and flow reg- flow on river channel meander migration patterns, Sacramento River, California, ulation of the world's large river systems. Science 308, 405–408. http://dx.doi.org/ USA. J. Am. Water Resour. Assoc. 42, 1063–1075. http://dx.doi.org/10.1111/j.1752- 10.1126/science.1107887. 1688.2006.tb04514.x. O'Connor, J.E., Jones, M.A., Haluska, T.L., 2003. Flood plain and channel dynamics of the Lauer, J.W., 2012. Channel Planform Statistics Toolbox, v 2.0. National Center for Earth- Quinault and Queets Rivers, Washington, USA. Geomorphology 51, 31–59. surface Dynamics. https://repository.nced.umn.edu/browser.php?current= Osterkamp, W.R., Hedman, E.R., 1982. Perennial-streamflow characteristics related to keyword&keyword=5&dataset_id=15. channel geometry and sediment in Missouri River basin. In: USGS Numbered Series Lawler, D.M., 1993. The measurement of river bank erosion and lateral channel change: a No. 1242. Professional Paper. review. Earth Surf. Process. Landf. 18, 777–821. http://dx.doi.org/10.1002/esp. Palmer, M.A., Liermann, C.A.R., Nilsson, C., Flörke, M., Alcamo, J., Lake, P.S., Bond, N., 3290180905. 2008. Climate change and the world's river basins: anticipating management options. Lea, D.M., Legleiter, C.J., 2016. Refining measurements of lateral channel movement Front. Ecol. Environ. 6, 81–89. from image time series by quantifying spatial variations in registration error. Parker, G., Shimizu, Y., Wilkerson, G.V., Eke, E.C., Abad, J.D., Lauer, J.W., Paola, C., Geomorphology 258, 11–20. http://dx.doi.org/10.1016/j.geomorph.2016.01.009. Dietrich, W.E., Voller, V.R., 2011. A new framework for modeling the migration of Li, Z., Liu, W., Zhang, X., Zheng, F., 2009. Impacts of land use change and climate meandering rivers. Earth Surf. Process. Landf. 36, 70–86. http://dx.doi.org/10.1002/ variability on hydrology in an agricultural catchment on the Loess Plateau of China. esp.2113. J. Hydrol. 377, 35–42. http://dx.doi.org/10.1016/j.jhydrol.2009.08.007. Pederson, G.T., Gray, S.T., Ault, T., Marsh, W., Fagre, D.B., Bunn, A.G., Woodhouse, C.A., Luchi, R., Bertoldi, W., Zolezzi, G., Tubino, M., 2007. Monitoring and predicting channel Graumlich, L.J., 2010. Climatic controls on the snowmelt hydrology of the northern

238 D.M. Schook et al. Geomorphology 293 (2017) 227–239

Rocky Mountains. J. Clim. 24, 1666–1687. http://dx.doi.org/10.1175/ doi.org/10.1890/1051-0761(1997)007[0677:FDOCEA]2.0.CO;2. 2010JCLI3729.1. Shafroth, P.B., Stromberg, J.C., Patten, D.T., 2002. Riparian vegetation response to al- Pizzuto, J.E., 1994. Channel adjustments to changing discharges, Powder River, Montana. tered disturbance and stress regimes. Ecol. Appl. 12, 107–123. http://dx.doi.org/10. Geol. Soc. Am. Bull. 106, 1494–1501. http://dx.doi.org/10.1130/0016-7606(1994) 1890/1051-0761(2002)012[0107:RVRTAD]2.0.CO;2. 106<1494:CATCDP>2.3.CO;2. Sigafoos, R.S., 1964. Botanical evidence of floods and flood-plain deposition. In: Pizzuto, J., Keeler, J., Skalak, K., Karwan, D., 2017. Storage filters upland suspended Geological Survey Professional Paper No. 485–A. USGS, Washington, D.C.. sediment signals delivered from watersheds. Geology 45, 151–154. http://dx.doi. Stella, J.C., Hayden, M.K., Battles, J.J., Piégay, H., Dufour, S., Fremier, A.K., 2011. The org/10.1130/G38170.1. role of abandoned channels as refugia for sustaining pioneer riparian forest ecosys- Powder River County Examiner, 1923. Wide Open Range on Powder River. tems. Ecosystems 14, 776–790. http://dx.doi.org/10.1007/s10021-011-9446-6. Rood, S.B., Goater, L.A., Mahoney, J.M., Pearce, C.M., Smith, D.G., 2007. Floods, fire, and Stewart, I., Cayan, D., Dettinger, M., 2005. Changes toward earlier streamflow timing ice: disturbance ecology of riparian cottonwoods. Can. J. Bot. 85, 1019–1032. http:// across western North America. J. Clim. 18, 1136–1155. http://dx.doi.org/10.1175/ dx.doi.org/10.1139/B07-073. JCLI3321.1. Rose, J.R., Cooper, D.J., 2016. The influence of floods and herbivory on cottonwood Stokes, M.A., Smiley, T.L., 1968. An Introduction to Tree-ring Dating. University of establishment and growth in Yellowstone National Park: Yellowstone cottonwood Chicago Press. establishment. Ecohydrology. http://dx.doi.org/10.1002/eco.1768. Wilcox, A.C., Shafroth, P.B., 2013. Coupled hydrogeomorphic and woody-seedling re- Sadler, Peter M., 1981. Sediment accumulation rates and completeness of stratigraphic sponses to controlled flood releases in a dryland river: coupled hydrogeomorphic and sections. J. Geol. 89 (5), 569–584. vegetation responses to floods. Water Resour. Res. 49, 2843–2860. http://dx.doi.org/ Schook, D.M., Friedman, J.M., Rathburn, S.L., 2016. Flow reconstructions in the Upper 10.1002/wrcr.20256. Missouri River Basin using riparian tree rings. Water Resour. Res. 52, 8159–8173. Williams, G.P., Wolman, M.G., 1984. Downstream Effects of Dams on Alluvial Rivers. http://dx.doi.org/10.1002/2016WR018845. Winters, D.S., et al., 2004. Aquatic, riparian and wetland ecosystem assessment for the Schumm, S.A., 1969. River metamorphosis. J. Hydraul. Div. 95, 255–274. Bighorn National Forest. In: No. Report 2 of 3: Anthropogenic Influences Report. Schumm, S.A., 1973. Geomorphic thresholds and complex response of drainage systems. USDA Forest Service, Rocky Mountain Region, Denver, CO. Fluv. Geomorphol. 6, 69–85. Wohl, E., 2013. Floodplains and wood. Earth-Sci. Rev. 123, 194–212. http://dx.doi.org/ Schumm, S.A., Lichty, R.W., 1963. Channel widening and flood-plain construction along 10.1016/j.earscirev.2013.04.009. Cimarron River in southwestern Kansas. In: USGS Numbered Series No. 352–D. Wohl, E., Bledsoe, B.P., Jacobson, R.B., Poff, N.L., Rathburn, S.L., Walters, D.M., Wilcox, Professional Paper. A.C., 2015. The natural sediment regime in rivers: broadening the foundation for Scott, M.L., Friedman, J.M., Auble, G.T., 1996. Fluvial process and the establishment of ecosystem management. Bioscience. http://dx.doi.org/10.1093/biosci/biv002. bottomland trees. Geomorphology 14, 327–339. http://dx.doi.org/10.1016/0169- (biv002). 555X(95)00046-8. (Fluvial Geomorphology and Vegetation). Xu, D., Bai, Y., Ma, J., Tan, Y., 2011. Numerical investigation of long-term planform Scott, M.L., Auble, G.T., Friedman, J.M., 1997. Flood dependency of cottonwood estab- dynamics and stability of river meandering on fluvial floodplains. Geomorphology lishment along the Missouri River, Montana, USA. Ecol. Appl. 7, 677–690. http://dx. 132, 195–207. http://dx.doi.org/10.1016/j.geomorph.2011.05.009.

239