Dam busy : beavers and their influence on the structure and function of river 1 corridor hydrology , geomorphology , biogeochemistry and ecosystems

1 Soil Geography and Landscape group, Wageningen University, Droevendaalsesteeg 3, 6 6708 PB Wageningen, The Netherlands, annegret.larsen@wur.nl (corresponding author) 7 2 The University of Manchester, Oxford Rd, Manchester, M13 9PL, United Kingdom 8 3 School of Geography, Earth and Environmental Sciences, University of Birmingham, 9 United Kingdom 10 4 Birmingham Institute of Forestry Research (BIFoR), University of Birmingham, United 11 Kingdom 12 5 Institute of Earth Surface Dynamics, Université de Lausanne, 1015 Lausanne, Switzerland 13

1 Abstract 28 Beavers (castor fiber, castor canadensis) are the most influential mammalian ecosystem engineer, 29 heavily modifying river corridors and influencing hydrology, geomorphology, nutrient cycling, and 30 ecosystems. As an agent of disturbance, they achieve this first and foremost through dam construction, 31 which impounds flow and increases the extent of open water, and from which all other landscape and 32 ecosystem impacts follow. After a long period of local and regional eradication, beaver populations 33 have been recovering and expanding throughout Europe and North America, as well as an introduced 34 species in South America, prompting a need to comprehensively review the current state of knowledge 35 on how beavers influence the structure and function of river corridors. Here, we synthesize the overall 36 impacts on hydrology, geomorphology, biogeochemistry, and aquatic and terrestrial ecosystems. Our 37 key findings are that a complex of beaver dams can increase surface and subsurface water storage, 38 modify the reach scale partitioning of water budgets, allow site specific flood attenuation, alter low 39 flow hydrology, increase evaporation, increase water and nutrient residence times, increase 40 geomorphic heterogeneity, delay sediment transport, increase carbon, nutrient and sediment storage, 41 expand the extent of anaerobic conditions and interfaces, increase the downstream export of 42 dissolved organic carbon and ammonium, decrease the downstream export of nitrate, increase lotic 43 to lentic habitat transitions and aquatic primary production, induce 'reverse' succession in riparian 44 vegetation assemblages, and increase habitat complexity and biodiversity on reach scales. We then 45 examine the key feedbacks and overlaps between these changes caused by beavers, where the 46 decrease in longitudinal hydrologic connectivity create ponds and wetlands, transitions between lentic 47 to lotic ecosystems, increase vertical hydraulic exchange gradients, and biogeochemical cycling per 48 unit stream length, while increased lateral connectivity will determine the extent of open water area 49 and wetland and littoral zone habitats, and induce changed in aquatic and terrestrial ecosystem 50 assemblages. However, the extent of these impacts depends firstly on the hydro-geomorphic 51 landscape context, which determines the extent of floodplain inundation, a key driver of subsequent 52 changes to hydrologic, geomorphic, biogeochemical, and ecosystem dynamics. Secondly, it depends 53 on the length of time beavers can sustain disturbance at a given site, which is constrained by top down 54 (e.g. predation) and bottom up (e.g. competition) feedbacks, and ultimately determines the pathways 55 of river corridor landscape and ecosystem succession following beaver abandonment. This outsized 56 influence of beavers on river corridor processes and feedbacks is also fundamentally distinct from what 57 occurs in their absence. Current river management and restoration practices are therefore open to re-58 examination in order to account for the impacts of beavers, both positive and negative, such that they 59 can potentially accommodate and enhance the ecosystem engineering services they provide. It is 60 hoped that our synthesis and holistic framework for evaluating beaver impacts can be used in this 61 endeavor by river scientists and managers into the future as beaver populations continue to expand in 62 both numbers and range. 63 64 Contents 2 Introduction 118 Beavers (Castor fiber, Castor canadensis) are semiaquatic mammals partial to freshwater 119 environments. They have the somewhat unique ability to create their own ecological niche at relatively 120 large scales by actively engineering their habitat through dam construction. They get busy doing this 121 most effectively in smaller channels, either of lower order streams and their associated floodplains, or 122 in floodplain and side channels of larger rivers (Butler and Malanson, 2005;Gurnell, 1998;Laland and 123 Boogert, 2010; Westbrook et al., 2013). This dam construction has the potential to alter the hydrology, 124 geomorphology, biogeochemistry, and ecosystems of river corridors and the feedbacks between them, 125 thus the beaver is also increasingly recognized as an 'ecosystem engineer' (e.g. Jones et al. (1996), 126 Wright et al. (2002)). Both species of beaver can have environmental impacts across wide swaths of 127 the Northern Hemisphere, and following a long history of eradication and now partial recovery (Halley 128 et al., 2012), their (re)-introduction is increasingly being advocated for in many cases to aid ecosystem 129 restoration in regions once part of their historical range ( Figure 1). Whilst some differences in litter size (Parker 131 et al., 2012) and dam building frequency (Whitfield et al., 2015) may exist between the two species, 132 for the purpose of this review, which focuses on landscape and ecosystem process impacts, and given 133 the highly inconclusive data on the biological and ecological differences, we make no further 134 distinctions between them. Although beavers occupy a range of habitats by burrowing (e.g. on large 135 rivers and lakes (Bashinskiy, 2020), it is their unique ability to construct dams within river corridors and 136 the consequences for landscape and ecosystem process that forms the focus of this review. Beavers 137 build dams to help engineer their food supply of riparian and wetland vegetation, to create water 138 bodies sufficiently deep that do not completely freeze during winter in higher latitudes, and as a 139 protection from potential predators. The sound of running water is also apparently sufficient 140 stimulation to trigger the busy dam repair behavior (Müller-Schwarze, 2011). The size of individual 141 beaver dams can be large, especially across floodplain and wetland habitats, however within free 142 flowing river reaches it appears beavers generally prefer to build across river widths of 4 -6 m or less 143 (Hartmann and Törnlöv, 2006) (Suzuki and McComb, 1998) and lower slope gradients (Suzuki and 144 McComb, 1998, Pollock et al. 2003), but also with relatively wide river valleys (e.g. valleys width > 4 145 stream widths) (Suzuki andMcComb, 1998, Pollock et al. 2003) where beaver meadows can also 146 develop (Figure 2 b). In addition, a single dam may not be built in isolation, with multiple dams over a 147 reach termed a beaver dam cascade, and in this case lower peak discharges and higher river valley 148 slope appear to be more important in allowing higher dam numbers to be constructed per cascade 149 (Neumayer et al., 2020) (Figure 2a). This is not to say beavers do not construct dams outside these 150 ranges (Pinto et al., 2009), or that other habitat factors such as vegetation (see section 6.5) are not 151 important, only that they appear to be the preferred conditions for dam construction within a wide 152 distribution of activity. Once constructed, dams may be actively maintained for years to decades, 153 become abandoned, breached by floods, filled with sediment, or modified by human activity (James 154 and Lanman, 2012; Johnston, 2015). Whatever their fate, both species of beaver have an amazing 155 capacity to engineer streams across a wide spectrum of environmental gradients, which also shapes a 156 range of positive and negative perceptions concerning their influence. On the one hand beavers may 157 be perceived as undermining existing river engineering schemes and current land use activities, and 158 thus creating conflict (Andersen and Shafroth, 2010). On the other hand, beavers may be seen as an 159 alternative to traditional 'hard' engineering in river restoration (Polvi and Wohl, 2013), with their 160 presence potentially improving river restoration success (Mika et al., 2010). 161 Recognizing the ever increasing interest in beavers and their works (Goldfarb, 2018), their increasing 162 population numbers and range (Halley et al., 2012), and especially their capacity to shape the river 163 corridor landscape (Naiman and Rogers, 1997), the aim of this paper is to synthesize our current 164 understanding on the process controls and impacts of beavers on river corridor hydrology, 165 geomorphology, biogeochemistry and ecosystems, as well as the feedbacks between them. This is 166 structured using seven sections: The first four deal with the primary impacts of beavers on processes 167 and dynamics: (3) hydrology; (4) geomorphology; (5) biogeochemistry; and (6) stream and riparian 168 ecosystems. In section (7) we integrate the knowledge gained from these separate areas to explore 169 the feedbacks between them, in section (8) we discuss the idea that beavers can promote alternate 170 stable states for river corridor ecosystems, and in section (9) we discuss the interpretation and 171 perception of natural landscapes and beaver impacts, as well as the role of beavers in stream 172 management and rehabilitation. A concise overview of these findings along with selected references 173 is provided in Table 1. Finally, in section (10) we use the outcomes of our synthesis to develop a holistic 174 framework in which beaver impacts can be evaluated as the hydrological and geomorphic contexts of 175 the river system change. 176 177 3 Beaver impacts on hydrology 178 Beavers first impact the overall water balance, and through this downstream flow regimes. Beavers 179 build dams, and the initial hydrological impact of beaver dam construction is a reduction in water 180 velocity and local increase of the in-channel water level, creating a beaver 'pond', with backwater 181 effects on the inflowing channel (Figure 4,). These ponds can be spatially extensive, grade into 182 wetlands and meadows, and can be relatively shallow in less confined rivers and floodplains (Chaubey 183 and Ward, 2006;Naiman et al., 1988), and vice versa in steeper and more confined river sections. 184 Through flow diversion of stream water ( Figure 4) and the accompanying rise in groundwater levels 185 (Figure 9 b, c), floodplain inundation can also be far more extensive than would otherwise occur 186 without beaver dams, especially during flood events (Westbrook et al., 2006). In a semi-or unconfined 187 valley river-floodplain system, beaver dam complexes (Figure 5 b) are likely to create more spatially 188 complex flow networks when compared to the river without beaver dams (Figure 5 a) (Green and 189 Westbrook, 2009). In areas with exceptionally low relief, beaver damming may even divert channels 190 across watershed divides (Westbrook et al., 2013). These observations suggest that the impact of 191 beaver dams on the hydrology of river systems varies widely, according to the processes that 192 determine the relative change in water level, water storage, and subsequent water redistribution 193 within the landscape that beaver dams come to occupy. These processes are discussed below. 194 195 3.1 Changes to storage and open water area 196 A change in water storage capacity is the key hydrological modification from which other impacts 197 follow. Analogous to artificial reservoirs, beaver dams create additional surface water storage whose 198 magnitude depends on whether the rise in water level behind the dam (to create a beaver pond) 199 remains confined to the channel. Examples of confined ponds include incised channels, or where the 200 channel is very large relative to dam size. If this is the case, then the surface storage impacts of beaver 201 dams are related only to the channel volumes, which can in itself be significant (Jin et al., 2009). If the 202 channel water level exceeds the local floodplain height, either permanently or on a seasonal or event 203 basis, the floodplain will be inundated to some extent and create larger areas of ponded and slowly 204 flowing water. This increases the frequency of channel-floodplain connectivity and provide access to 205 greater floodplain spaces to store and move water. Changes to energy losses and stream slope will 206 also be important as these will control the partitioning of discharge rises between increases in velocity 207 and increases in depth for in-channel flow and hence the ease of connection between river and 208 floodplain. Thus, the stream-valley morphology is also a critical determinant of the potential 209 hydrological impacts of beaver dams. Depending on these geomorphic and hydrologic conditions, the 210 increase in water storage is usually most clearly manifest as an increase in the areal extent of open 211 surface water, which have been measured to be up to 9 -12 times the pre-beaver open water extent 212 (Hood and Bayley, 2008 inundation extent can be profound over long (e.g., 50 -60 year) time periods (Figure 6a, b), with Hood 215 and Bayley (2008) finding a 9-fold increase in water surface area over this time in Alberta (Canada). 216 They can also be profound within a single reach as dam densities increase (Figure 6 c), and even 217 seasonally within a single pond and wetland complex (Figure 6 d). This increase in open water area 218 with reduced turbulence is therefore an important hydrological consequence of beaver dam 219 construction in river systems, and can have profound implications for the water balance, 220 biogeochemical processes, and ecosystems. 221 Floodplain storage capacity may be further enhanced as beavers modify their habitat, for example 222 through the excavation of small floodplain channel networks and ponds (Johnston and Naiman, 1990; 223 Stocker, 1985). Although the surface storage capacity of individual beaver dams (pond and floodplain) 224 is small relative to artificial reservoirs, the cumulative surface storages of multiple dams within a 225 beaver dam cascade may significantly increase their hydrological impact (Figure 6a and b) (Puttock et 226 al., 2017, Nyssen et al., 2011). Published dam density estimates range between less than 1 (e.g. 0.1) to 227 > 70 dams per km of river reach (Gurnell, 1998;Zavyalov, 2014, Pollock et al., 2003, although 228 considerably lower density estimates were compiled by Johnston (2017). At high densities, even small 229 individual dam storage capacities (L 3 ) relative to inflow rates (L 3 T -1 ) can in the aggregate substantially 230 modify water balances, water residence times, and flow regimes. These topics will be discussed in the 231 following sections. 232 There are at least four ways in which the comparison between beaver dams and artificial reservoirs or 233 weirs diverge, with important implications for the interpretation of storage dynamics. First, the dam 234 structure itself is permeable (Burchsted et al., 2010), and will make a largely unknown contribution to 235 outflow rates (discussed in the section below). Second, the relatively low dam height compared to 236 valley width results in very high surface area to volume ratios which can enhance losses to infiltration 237 and evaporation. Third, beaver dams are typically constructed within alluvial valleys of moderate to 238 low stream power (Pollock et al., 2003;Suzuki and McComb, 1998), conditions that are favorable to 239 higher hydraulic connectivity between the surface and shallow alluvial aquifers. This means that the 240 subsurface storage volume changes have the potential to be comparable to, if not larger than, the 241 surface storage volume changes, a point discussed in more detail in the surface -groundwater 242 connectivity section 3.5. Finally, the physical location of beaver dams can be highly dynamic in space 243 and time, adding significant complexity to how storage changes evolve within river reaches, especially 244 those with multiple dams over short distances. All these processes can change the water storage 245 dynamics in catchments and have important implications for the way the hydrological cycle is balanced 246 over a range of timescales.

249
(2017), Johnston and Naiman (1990)) 250 251 3.2 Water balance 252 The water balance from the perspective of the storage influenced by a beaver dam (e.g. a pond) can 253 be written as 254 where ⁄ is the change in total storage created by damming over the timescale of interest , is 255 the inflowing discharge, is the evaporation from the beaver modified system, and is the 256 outflowing discharge (Figure 4). The units for the terms on the right-hand side can be volumetric fluxes 257 (L 3 T -1 ), or rates normalized to the area occupied by the beaver dam system (LT -1 ). permeability; and 4) underflow, the flux seeping below the dam structure based on the nature of 273 contact between the dam base and the substrate, not including subsurface flow ( ). These 274 mechanisms of loss may also vary with dam age and level of maintenance by beaver populations 275 (Woo and Waddington, 1990). A survey of 51 beaver dams of varying age in Germany found gapflow 276 was by far the dominant mechanism of water release (Neumayer et al., 2020). Crucially, these 277 observations suggest that the quantification of the hydraulics of beaver dams is difficult when based 278 upon analogies with human-engineered instream structures (e.g. broad-crested weirs), particularly if 279 their hydraulic impacts are to be modelled, emphasizing the need for more detailed studies of beaver 280 dam hydraulics (Feng and Molz, 1997). 281 As mentioned in the previous section, it may be conceptually useful in the case of beaver dam systems 282 to separate the total storage into surface and subsurface terms, noting the likely interaction between 283 them: 284 where is the change in surface storage, and is the change in groundwater storage. 285 is also further divisible into the beaver pond (water impounded behind the dam) and water 286 diverted onto the floodplain. 287 Over shorter timescales (i.e. sub-annual), changes in the total storage term can have significant 288 hydrological impacts and are discussed in the next sections in terms of flow regimes. However, over 289 annual and longer timescales, this change in storage should be largely balanced by the outflow terms 290 (i.e. and ), assuming regional groundwater flow remains minor relative to the surface fluxes. If 291 the partitioning between and remains the same following beaver dam construction, then the 292 storage changes have had negligible impact on the overall water balance. However, if the partitioning 293 between and changes following beaver dam construction (e.g. an increase in and decrease 294 in ), then the changes in the way water is stored will also likely impact the water balance. There 295 are very few quantitative analyses of beaver dam impacts on all components of the water balance at 296 the annual scale (but see: Chaubey and Ward, 2006;Johnston, 2017;Woo and Waddington, 1990), 297 highlighting a clear and profound knowledge gap in how beavers may impact hydrology. In a beaver-298 dammed sub-arctic catchment, Woo and Waddington (1990) found total was reduced relative to a 299 paired non-beaver impacted catchment, suggesting that storage changes are capable of increasing 300 fluxes (c. 40%) at the expense of at the annual scale. In a boreal environment, (Johnston, 2017) 301 also found was diminished at the expense of increasing and groundwater recharge. Correll et 302 al. (2000) also compared annual changes in a beaver impacted and control watershed within the 303 Atlantic Coastal Plain (USA), and found a reduction in presumed to be at the expense of increasing 304 , however, the full water balance comparison was not reported in this study. In the seasonally dry 305 coastal plain of Alabama (USA), Chaubey and Ward (2006) also found a large increase in due to a 306 single beaver dam. However, because of the large increase in wetland and pond surface area at this 307 site relative to the catchment area, the increase in was largely subsidized by an increase in direct 308 rainfall on the wetland rather than as a loss to . It is also worth noting that Devito and Dillon (1993) 309 constructed full seasonal and annual water balances for a beaver pond in central Ontario, Canada, 310 however no comparison with pre-or non-beaver impacted sites were made. In any case, there is a 311 consistent message from a small number of studies (n = 8) that tends to diminish downstream of 312 beaver dam complexes (Figure 16e). 313 The mechanisms by which beaver dam systems can increase total may involve some combination  increases in floodplain open water extent downstream of dams due to substantial flow  318  diversion during flood events, inundations which can persist for weeks to months (Levine and Meyer,  319 2014; Westbrook et al., 2006). This increase in open water extent is likely to be a fairly common 320 feedback affecting the partitioning of and across all beaver impacted systems, and potentially 321 also the local climate (Hood and Bayley, 2008), yet the feedbacks remain poorly understood. Burns 322 and McDonnell (1998) also found overall streamflow was reduced in a beaver impacted catchment and 323 attributed this to increased . Although this was not quantified at annual timescales, the influence 324 of increased evaporation was evident in the clear offset of streamflow stable isotopes from the local 325 meteoric water line in water samples collected downstream of the beaver dam complex (Burns and 326 McDonnell, 1998 vegetation will also influence evaporative losses depending on the vegetation conditions they replace. 335 Although not yet examined in beaver impacted systems, evaporation from wetlands with a mix of open 336 water and wetland vegetation can be extremely complex and may be higher or lower than the open 337 water rate depending on how the local atmospheric demand influences stomatal conductance 338 (Anderson and Idso, 1987;Wetzel, 2001). It is clear though that for an equivalent surface area and 339 atmospheric conditions, the rate of losses should be higher where wetland vegetation cover is 340 greater than unobstructed open water (Wetzel, 2001), and is likely the cause of the large diurnal 341 variations observed in some beaver pond water levels (Johnston, 2017;Ward and Chaubey, 2000 (Figure 8). In principle, any increase in storage 363 capacity can allow greater buffering or hydrologic stability to be imposed on . This modification 364 may apply to all flows, but in terms of hydrological impact is especially important to determine for high 365 flow and baseflow conditions. 366 The ability of beaver dam systems to attenuate and delay peak flows depends on the available surface 367 storage capacity immediately preceding streamflow rise (i.e. freeboard), relative to the inflowing flood 368 volume. The freeboard available behind beaver dams is in general likely to be small as the water depth 369 behind dams is usually engineered by the beaver to be close to the dam crest height (Figure 7) (Devito 370 and Dillon, 1993 (2017) monitored both and in a beaver-impacted system, finding significant attenuation in 380 the flood hydrographs caused by a complex of 5 -6 beaver dams in Belgium in the case of the former 381 (Figure 8), and 4 -10 beaver dams in England in the case of the latter. Given the already mentioned 382 wide range in beaver dam densities in cascades, a major limitation to understanding flood attenuation 383 impacts is the cumulative storage and flow diversion processes that can occur both within and between 384 beaver dams. This is likely why modelling studies of beaver flood impacts that do not explicitly include 385 flow diversion find minimal impact on flood water storage, and relatively small effects on hydrograph 386 attenuation (Beedle, 1991). This is not to say that once floodplain diversion is included, all river systems 387 with beaver dams will have significant attenuation. Neumayer et al. (2020) conducted a  388  comprehensive 2D hydrodynamic model experiment by numerically inserting beaver dam cascades  389 into two sites in southern Germany for a wide range of flood event conditions. Interestingly, they found 390 flood volume attenuation and the delay in flood peak timing was only significant for smaller discharge 391 events and were much more pronounced at the site with lower slope and higher floodplain 392 connectivity. However, for flood events matching the 2-year return interval and above, in both sites 393 the impact on attenuation and delay was minimal or absent, even with large increases in floodplain 394 inundation area. These findings highlight the possibility that in many cases, once all factors are 395 considered, beavers may still have minor to negligible impacts on flooding, especially for very large 396 flood events. However, until the full flow diversion and storage changes for river corridors across a 397 wide range of topographic and geomorphic conditions is considered, the extent of beaver impacts on 398 flooding is at risk of continually being misjudged. Some parallels may be made with the work of Dixon wave propagation through a river basin network. Importantly, this work shows that the catchment 401 scale effect of debris dams in total is not the same as the sum of the impacts of each debris dam 402 individually, emphasizing the need to look in more detail at precisely how multiple beaver dams impact 403 flood attenuation. In the absence of information on both and , flood attenuation impacts from 404 beaver dams can also be assessed indirectly using the paired catchment approach (e.g. Woo and 405 Waddington (1990)), through discharge time series evaluation at a downstream point that contains 406 both pre-and post-beaver dam periods (Nyssen et al., 2011;Westbrook et al., 2006), or using 407 geochemical tracers (Burns and McDonnell, 1998). Whatever the method, there is a clear need for 408 better and more accurate assessments of the capacity for beaver damming to modify the full range of 409 catchment flood magnitudes. This urgency is enhanced by an increasing desire to re-introduce beavers 410 for the explicit purpose of flood management, despite insufficient science to understand how beaver 411 impacts might actually achieve this (BBC, 2017). 412 Floods may also cause dam breaches or failure, potentially leading to flood amplification (Butler and 413 Malanson, 2005;Hillman, 1998 At low flow, the potential impact of beaver dams is heavily dependent on the mechanisms by which 437 storage is released, which for is some combination of and , assuming is very 438 small (Woo and Waddington, 1990). Dams with high throughflow rates will more rapidly deplete 439 surface storage as the level declines (Woo and Waddington, 1990 underflow) are small (e.g. Devito and Dillon (1993)). In contrast, dams with higher underflow loss rates 443 may sustain a higher contribution to that is proportional to the rate of decline in pond 444 water level. 445 If is the primary storage regulating baseflow in beaver impacted systems, then any increases in 446 evaporative losses, especially in the summer months, will negatively impact baseflow. This appears to 447 be the case in some water balance and spot discharge measurement studies (Correll et al., 2000;448 Meentemeyer and Butler, 1999;Woo and Waddington, 1990). However, if is sufficiently large then 449 baseflow reductions may be either offset to some degree, or even increase following beaver dam 450 construction. If baseflow does increase, the overall water balance is likely to be maintained through 451 high flows that replenish (and contribute to some increase in ), but that are also able to 452 recharge to . Increased baseflow in beaver impacted systems has been hypothesized or reported 453 by a number of authors (Johnston, 2017;Macfarlane et al.;Puttock et al., 2017;Stabler, 1985, Smith 454 et al., 2020). Majerova et al. (2015) found an increase in downstream mean daily discharges following 455 beaver impact, which could be attributed directly to measured increases in surface and groundwater 456 storage, with the magnitude of this impact increasing with the number of beaver dams in the reach 457 over time. In a comparative before and after beaver impact study, Smith et al. (2020) found a large 458 increase in flow recession duration and reduced diel flow variability, suggesting beaver damming 459 increased flow buffering. Although there was no significant change in mean discharge, an increase in 460 due to beaver damming allowed a significant tempering and delay to low flow releases. Beyond 461 these studies, there is also considerable observational, anecdotal, and in some cases experimental, 462 support for a positive impact of beaver damming on low flows across a range of climatic and landscape 463 settings (Pollock et al., 2003;Rosell et al., 2005;Stabler, 1985). This underscores the strong need for 464 more quantitative studies in this area, as a sustained enhancement of baseflow would have profound 465 ecological implications, especially in otherwise ephemeral river systems and in drier climates (Gibson 466 and Olden, 2014). In addition, under conditions of hydrological and meteorological drought, as 467 streamflow declines or even ceases, beaver ponds and the wetlands they sustain may themselves 468 retain significant amounts of water (Hood and Bayley, 2008a), raising the interesting prospect that 469 they may act as critical ecosystem 'refugia' in the aquatic landscape during drought (Hood and Bayley,  470 2008a) and even as landscape buffers against fire (Fairfax and Whittle, accepted, Wheaton et al., 2019). 471 It should also be noted that the very nature of beaver dams also complicates our ability to model how 472 storage changes should impact downstream discharge. This is because the influence of beaver dams 473 on the hydrological processes described above are largely dependent on highly localized factors such 474 as substrate type, construction materials, design integrity (Muller and Watling, 2016), and age 475 (Meentemeyer and Butler, 1999), properties which may not be easy to transfer between different 476 beaver impacted systems, or even between individual dams. Additionally, the large variability in dam 477 locations and densities means their influence on the total storage capacity can be highly dynamic in 478 space and time. This makes it very difficult to undertake meaningful hydrological model calibration 479 without explicit knowledge and tracking of all the changes in the storage and flows occurring in the 480 river corridor. 481 482 3.5 Ground and surface water interactions 483 The extent to which increased groundwater storage ( ) may supply river baseflow is itself dependent 484 on the hydraulic characteristics of both the river and the aquifer. The total volume of available aquifer 485 storage is driven by the aquifer geometry (bounded by the valley) river channel, and how stratigraphy 486 controls the hydraulic properties. Provided high open water levels in beaver ponds and backwater 487 areas can be maintained, they may serve as an effective recharge pathway, either via the channel 488 boundary or as floodplain infiltration, causing a rise in local groundwater levels (Figure 9 a, b) (Karran 489 et al., 2018, Zahner, 1997. The effectiveness of this pathway will be heavily determined by hydraulic 490 conductivity, which may vary by many orders of magnitude in alluvial settings. In the context of beaver 491 impacted systems, the deposition of fine sediment in the ponds and around dam structures, and 492 potentially upon floodplain wetlands, can lower the hydraulic conductivity at these interfaces 493 (Johnston, 2017), similar to what has already been found in other river channels (e.g. Stewardson et 494 al. (2016)) and floodplain (e.g. Nowinski et al. (2011)) settings. Nonetheless, even though rates of  495  exchange at a point may be reduced, this impact may also be counteracted to some degree by the  496  expanded area over which ground and surface water interactions will occur. This potential tradeoff  497 between the areal extent vs rates of river aquifer exchange is also an important knowledge gap in 498 beaver impacted systems. 499 Beaver impacts may therefore introduce an interesting set of changed hydraulic gradient boundary 500 conditions that in an idealized case can be divided into being either upstream or downstream of an 501 individual beaver dam. In this case, we would generally consider beaver impacted systems as generally 502 'losing' (i.e. net water exchange from the surface to the aquifer) upstream of beaver dams, and 503 'gaining' (i.e. net water exchange from the aquifer to the surface) downstream, analogous to the 504 dynamics that occur across many man-made instream structures (Hester and Doyle, 2008). If high 505 beaver dam densities exist within a reach, such an idealized case will be too simplistic as many nested 506 flow paths may develop between the dams but may be valid over the whole reach scale. Despite the 507 clear potential for significant changes to the longitudinal hydraulic gradient, the variation in magnitude 508 of upstream losing and downstream gaining conditions within beaver dam impacted systems is not 509 well constrained. This is critical to understand, as it is likely to be a key control on the magnitude of , 510 and whether baseflow is likely to increase or decrease as a result of beaver impacts. This sequence of 511 interactions is broadly consistent with the findings of Lowry (1993) in an alluvial river of north-central 512 Oregon (USA), where a groundwater 'wedge' developed upstream and adjacent to a beaver dam 513 (Figure 9 b). This increase in groundwater storage (an additional ~89m 3 of drainable storage) driven by 514 the losing hydraulic gradients upstream of the dam, in turn sustained groundwater flow back to the 515 river downstream of the dam (i.e.: switch to gaining conditions) (Lowry, 1993 2020; Zahner, 1997) have also found significant changes in upstream and downstream groundwater 521 dynamics in close proximity to beaver ponds. In all cases there was a rise in groundwater levels (as a 522 result of increased ) following dam construction, and in the case of (Zahner, 1997) showed 523 relatively rapid declines in level once the beaver dam was removed (Figure 9 a,b). In addition, 524 depending on local topography and aquifer properties, recharge during flood events may be sufficient 525 to cause local groundwater flooding, and thus contribute to the overall surface inundation (Westbrook 526 et al., 2020). Interestingly, groundwater models have been under-utilized in examining potential 527 impacts from beaver structures. Whilst this would be an imperfect representation of the beaver 528 impacts on groundwater, such an approach has the potential to be a useful tool in evaluating the 529 storage and water balance impacts of beaver dams from the perspective of the aquifer. This in turn 530 will be critical to better understand potential baseflow impacts, especially where is expected to 531 play an important role. Over the longer term, as beaver dams are breached or fill with sediment and 532 beavers abandon or decrease activity, wetland and meadow development may decrease in , 533 however they may still retain significant , especially relative to the pre-impact landscape (Grygoruk 534 and Nowak, 2014). If this finding from Grygoruk and Nowak (2014) in Poland is generalizable, it has 535 significant implications for the long-term water storage and flow dynamics of beaver impacted river 536 systems, where unique wetland and meadow successional landscapes with increased water storage 537 may persist even in the absence of actively maintained beaver dams and ponds. 538 A related hydrologic process impacted by beaver dams is hyporheic exchange, distinguished from 546 broader ground and surface water interaction as water that enters and returns from the subsurface, 547 with a flow field typically induced by variations in channel topography (e.g. wood, bedforms, weirs, 548 etc) (Figure 4c). Although the total flux of water within this flow path is small relative to that in the 549 channel, it is important to consider given the role of the hyporheic zone in the biogeochemical cycling 550 of river systems. Vaux (1968) developed an analytical description that is useful to illustrate the 551 potential effects of beaver dams on the vertical component of hyporheic exchange at the interface 552 between the streambed and surface water ( ) 553 where is the streambed interface pressure (FL -2 ), is viscosity, ℎ is water depth (L), is the stream 554 length (L), is mean permeability (L 2 ), is the depth of the streambed containing the hyporheic flow 555 field, and ⁄ is the downstream variation in the streambed surface elevation. Positive values of 556 at a point indicate vertical hyporheic flow from the streambed to the river (i.e. upwelling) and negative 557 values indicate flow from the river into the streambed (i.e. downwelling). A key dynamic is introduced 558 by ⁄ , i.e. whether travelling in the downstream direction the streambed is broadly concave and 559 promoting upwelling, or convex and promoting downwelling. For the case of a single beaver dam, the 560 change in ⁄ is not gradual, but abrupt. Nonetheless, the shape can be approximated as a strongly 561 concave element and therefore conducive to upwelling. The effect of an abrupt change rather than a 562 gradual concave profile is to 'tighten' the flow net (or velocity flow field) beneath the dam, and thus 563 increase the magnitude of upwelling downstream (Figure 4c). In very flat terrain or a channel 564 without pronounced bedforms, beaver dams may provide the only significant hyporheic exchange 565 element in the system, and therefore introduce a large local change in subsurface flow dynamics. In 566 steeper environments, or where channels have considerable variation in the channel bed elevation 567 (e.g. large pool and riffle sequences), beaver dams will represent one component of the overall 568 hyporheic exchange (though still likely distinct given the abruptness of changes in ⁄ across a 569 beaver dam). In addition to the influence of ⁄ , and ℎ will likely decrease downstream of beaver 570 dams due to the abrupt decrease in water level, which also serve to increase downstream. The data 571 collected by Hartmann and Törnlöv (2006) nicely demonstrates that the capacity for beaver dams to 572 generate increased vertical hydraulic gradients is much greater where the downstream water depth is 573 lower (Figure 10), imposing an additional constraint on beaver dam influences on hyporheic processes. conditions upstream of beaver dams, albeit with considerable variability tied to the river morphology 582 and streamflow conditions. These results are also consistent with the hyporheic response expected 583 across man-made channel structures (Hester and Doyle, 2008), especially ones that span the full 584 channel width (as is typical for beaver dams). 585 There are some important caveats that will moderate the potential influence of beaver dams on 586 hyporheic exchange. As in any river system, the degree of exchange will also depend on the overall 587 regional ground and surface water gradients, which are not explicitly included in Equation (4). Thus, 588 strongly losing or strongly gaining conditions will also influence the relative impact of beaver dams on 589 . In an extremes case, an isolated beaver dam within strongly losing or strongly gaining systems 590 would be unlikely to have a significant impact on hyporheic exchange at the reach scale. The 591 considerable heterogeneity in riverbed will also exert a strong influence on . As already discussed, 592 there is a higher likelihood of encountering lower permeability flow paths upstream of beaver dams 593 due to deposition of finer sediments which will reduce local downwelling rates, and thus also reduce 594 any downstream upwelling, even if again increases downstream. It is also important to emphasize 595 that any impacts of beaver dam induced hyporheic exchange will be highly localized, and that the 596 impact will therefore be enhanced when many dams are present within a reach, but less impactful 597 when a reach has fewer dams. Nonetheless, equation 4 illustrates the potential for considerable 598 enhancement of hyporheic exchange driven by beaver dams, especially compared to most other 599 channel roughness features typically encountered in river corridors. This influence on hyporheic flow 600 has important implications for overall water residence times (section 3.7), and influence the extent to 601 which biogeochemical reactions can occur there (see section 5). 602 603 Figure  Any enhanced hyporheic flow as described above will be but one mechanism by which water residence 608 times are increased in beaver impacted river reaches. Overall, any system in which the storage capacity 609 increases to capture a greater proportion of inflowing water necessitates that the residence time of 610 the water leaving the system also increases. In the case of beaver impacts, even though the increase 611 in hyporheic and subsurface flow and storage will be large, it is the nature of the surface water storage 612 changes to which residence times will be most sensitive, since this is the storage with which the vast 613 majority of the flow will be interacting. The simplest characterization of the water residence time ( ) 614 for a beaver impacted system is the nominal residence time ( ) 615 Where is the total (nominal) volume of surface water storage (L 3 ) in the beaver system, and is the 616 volumetric flow rate (L 3 T -1 ). There is a longstanding ambiguity as to which flow rate should be used, 617 , , or an average of the two. Ideally, it would be preferable to use the latter if sufficient 618 monitoring information is available, and in this case is often referred to as the through-flow rate. 619 However, as in all natural systems, flow mixing leads to zones of faster and slower flowing water in the 620 ponds and wetlands. This means that over seasonal or annual timescales not all the water will 621 participate in active flow through the system and that is almost always an overestimate of actual 622 residence times. Therefore, it is important to understand the volume of storage engaged in active flow 623 ( ) 624 Where represents the volumetric efficiency of the beaver impacted system, which lumps together 625 several factors that may generate stagnant pockets of water (such as vegetation, large woody debris, 626 and irregular hypsometry) as well as any uncertainties in the estimates. Thus, a better 627 representation of in beaver impacted systems is 628 Unfortunately, there is no a priori theory to predict and thus from information on and 629 alone. Therefore, snapshot measurements of tracers that 'track' the flow of water are essential since 630 this will capture the full mixing process of the system and allow the key moments of the residence time 631 distribution (e.g. mean and variance) to be extracted. Majerova et al. (2015) conducted tracer 632 experiments over a relatively short (~750m) river reach before and after the construction of ~10 beaver 633 dams in a first order perennial mountain stream in Utah, and found residence times had increased 634 from 27 to 89 minutes (a 230% increase). Devito and Dillon (1993) also reported residence time 635 estimates, however the exact method was not specified, but they are likely to be based estimates 636 and thus overestimate actual τ to some extent. Nonetheless, assuming the pre-beaver residence time 637 over the reach would have had the same structure as the outflow, they report average annual 638 residence times have increased from 6 hours to 47 days. However, it is also worth noting this is an 639 average of two distinct flow regimes operating in this system, namely high snowmelt dominated water 640 fluxes in spring with very short residence times, and of very low water fluxes over the summer and 641 autumn periods with very long residence times. Given the paucity of results on the impact of beaver 642 damming on water residence times, it is useful to also note some similarities with debris dams, which 643 although are far more porous structures, have nonetheless consistently been found to also increase 644 reach scale water residence times across a variety of flow conditions (e.g. Ehrman and Lamberti (1992)) 645 For future research it is important to note that will also be dynamic over time in two important ways 646 in beaver impacted systems: 1) through the impact of changing on depending on the pond 647 hypsometry of the system, and 2) through the seasonal growth and decay of vegetation and its impact 648 on . Therefore, we should expect large variation in as both flow and vegetation vary seasonally in 649 beaver impacted systems. For larger values of , there is also an increasing likelihood that does not 650 remain constant for the duration of the tracer measurement, and that and infiltration can also 651 impact , all of which can confound the interpretation of tracer based estimates. A thought 652 experiment comparing residence times between water and sediment as the number of beaver dams 653 in a system increases is provided in the geomorphology section (section 4). 654 3.8 Water temperature 655 The changes in hydrology due to beaver impacts described above also have potential implications for 656 water temperatures within a beaver impacted reach, as well as downstream of beaver dams. Any 657 regulation of Q will have some impact on the advective component of the river reach energy budget, 658 but it may not necessarily be a large impact. An increase in surface water storage area can increase 659 the influence of the radiative component of the river energy budget, especially if this is accompanied 660 by a decline in riparian vegetation cover. This means the ponds behind beaver dams are likely to be 661 the main water body influencing any changes to the temperature regime downstream. This is 662 supported by Harthun (1998) and Harthun (2000) who found beaver ponds were on average 2.3 °C 663 warmer than adjacent stream sections in central Germany. It is also likely that beaver ponds are usually 664 too shallow to develop significant temperature stratification (Naiman and Melillo, 1984), except in 665 ponds that experience lengthy ice formation (Devito and Dillon, 1993), or in littoral zones with 666 abundant macrophytes (Majerova et al., 2020). An increase in groundwater storage can increase the 667 supply of water at the local groundwater average temperature, provided this is also contributing to 668 downstream . Groundwater temperatures are typically slightly above the local mean annual air 669 temperature (Benz et al., 2017), and considerably less variable in time than surface water 670 temperatures. However, if the groundwater recharge rate has increased as a result of beaver ponding, 671 the temperature of recharging stream water can also have a substantial legacy effect on the shallow 672 groundwater temperatures (Lowry, 1993 Edwards (1994) found no relationship between the size or number of beaver ponds and the extent of 677 warming, however Majerova et al. (2015) did find that temperature increased cumulatively with the 678 number of dams. Moreover, within a single beaver pond and wetland system, there is considerable 679 spatial heterogeneity in the thermal regimes that itself mirrors the increased habitat variability, with 680 the more marginal and shallower wetland and pond regions exhibiting the most warming and variation 681 (Majerova et al., 2020). The increased surface water storage following beaver damming has also been 682 found to act as a buffer of summertime low flow temperatures, increasing minimum and decreasing 683 maximum diel ranges without a change in the mean temperature (Weber et al., 2017). This study also 684 found an increase in localized groundwater upwelling which provided isolated zones of colder water 685 refugia (Weber et al., 2017). In terms of overall downstream impact,  found water 686 temperatures were higher downstream of a beaver dam complex in spring, summer, and autumn, and 687 potentially colder during winter. Interestingly, Avery (2002) found that beaver dam removal in some 688 Wisconsin (USA) streams led to an overall decrease in average stream temperatures, and in the 689 western Great Lakes region (USA) there are numerous catchment studies where beaver dams have 690 been found to elevate stream temperatures, except in streams with higher groundwater inputs 691 (Johnson-Bice et al., 2018). There is therefore sufficient evidence to suggest beaver dam building and 692 pond creation has the potential to increase the average downstream water temperature, however this 693 is by no means universal and the overall energy budget dynamics that determines the magnitude of 694 this increase remains poorly understood. This is especially the case at shorter time scales where the 695 relative importance of site specific conditions on water temperature increases. The magnitude of these 696 potential water temperature changes is particularly important to understand given their local influence 697 on aquatic ecosystems, and fish in particular (section 6.3), through both metabolic and dissolved 698 oxygen controls. likely to be constructed than dams. It is important to note dams are also constructed at larger channel 710 widths, just with far lower frequency. Beaver dams also rarely appear in very steep headwater streams, 711 indicating that stream power might be a factor controlling dam constructing activity. Taken at face 712 value, these results suggest the scale of hydro-geomorphic impacts from beavers is likely to decrease 713 with river size, and therefore with increasing stream order, meaning only minor construction activity 714 should be expected in larger river systems (Levine and Meyer, 2014;Naiman et al., 1988). However, 715 many larger river systems also have increasing levels of anthropogenic modifications to floodplain and 716 channel environments and flow regulation, meaning the reduction in dam construction frequency on 717 larger river systems may be difficult to disentangle from the increase in human influence. This section 718 explores the geomorphic impact of beavers on 1) sediment transport and deposition, 2) erosion 719 (including beaver dam breaches) and channel stability, and 3) long-term river valley formation. 720 721 4.1 Sediment transport and deposition in beaver systems 722 An important geomorphic impact of beaver dams is to reduce the longitudinal (downstream) 723 hydrological and sediment transport connectivity in rivers ( Figure 4). The reduced velocity upstream 724 of dams (backwater effect) causes a decline in sediment transport capacity, with bedload initially 725 deposited as sediment wedges against the dams (Figures 11, 12 a), and over time some suspended 726 load will settle out as the still-water area of the beaver ponds expand to cover the bedload deposits. 727 These dam-wedge and pond deposits are also rich in particulate organic carbon (POC), which is partly 728 produced by the decomposition of in-situ aquatic vegetation, but also transported from upstream. 729 Additionally, beavers add organic matter to the stream by felling trees, encouraging habitat for 730 macrophyte and biofilm growth, and intentionally submerging vegetation for winter food storage (see 731 sections 5, 6.2, 6.5) . Sediment wedges have their highest thickness at the dam and decrease in 732 thickness with distance from the dam in the upstream direction (Figures 11, 12a) and are also 733 influenced by active construction and modification by beavers themselves. However, dam-wedge 734 sedimentation dynamics and geometry can be difficult to quantify and is therefore rarely taken into 735 account in assessments of overall beaver pond sediment deposition and storage. 736 Whilst the sediment wedge against the dam is often the thickest area of deposition within a beaver 737 pond, the progressive development of backwater environments can also result in the upstream 738 deposition of bedload as delta-like deposits (Harthun, 1998) (Figure 12 b), although this has not been 739 reported in all studies (de Visscher et al., 2014). Delta-like deposition can often be generated due to 740 the supply of a sediment pulse from the breach of an upstream beaver dam (see below), and might 741 therefore be more common in systems that have had the opportunity to develop multiple dams. These 742 sedimentation patterns may also reflect the influence of distinct flow stages, e.g. wedge deposition 743 during high flows, and delta-like deposition during low and medium flows. However, further research 744 is needed to better understand depositional patterns in beaver impacted reaches. 745 Across these range of sedimentation mechanisms, it is clear that beaver dams and ponds trap 746 sediments to a much greater extent than would otherwise occur in their absence (Table 2). However, 747 these sedimentation rates also vary widely, with estimates ranging between 0.2 up to 45 cm yr -1 (Table  748 2). These comparatively large rates demonstrate that sediment trapping efficiency of beaver ponds 749 can be very high (Giriat et al., 2016). However, the large variability also attests to the importance of 750 local conditions in controlling the overall trapping efficiency and sediment supply, which can also be 751 seen in the comparatively high sedimentation rates in beaver systems from more mountainous 752 regions, and generally reduced sedimentation rates in lowland regions (Table 2). It is important to note 753 however, that this is a 'between catchment' spatial trend and does not track downstream changes in 754 sedimentation rates in a single system, or at a single site over time. Most research has focused on 755 'snapshots' of sedimentation within beaver pond cascades, but this storage capacity is also transient 756 over longer timescales because beaver dams either eventually breach or the associated ponds fill with 757 sediment, and hence the capacity of dams to store additional sediment will diminish to become 758 negligible over time (Demmer and Beschta, 2008;Levine and Meyer, 2014;Persico and Meyer, 2009). 759 This is also supported by the observation that deposition rates in ponds can be very high just after dam 760 construction, but reduce with age (Meentemeyer and Butler, 1999). Even if the variation in sediment 761 rates over time is not well known, there is in principle an upper limit to the sediment storage capacity 762 of beaver dams. The simplest expression of this maximum sediment storage ( ) for a single beaver 763 dam, represented as a triangular prism, can be formulated following Pollock et al. (2003) as: 764 where is the beaver dam height, is the pond or valley, and is the valley or river slope. This is a 765 highly idealized estimator, and therefore may not be applicable over shorter term timescales (e.g. < 766 10 1 -10 2 years) where irregular storage geometries across multiple beaver dams will be highly 767 influential. This also makes Equation (8)  Within beaver dam cascades (Figure 2a, 12d) the relationship between age and deposition rate breaks 775 down when sediment released by dam breaching is simply re-captured by other beaver ponds 776 downstream ( Figure 12d), a process which significantly delays the overall timescales of sediment 777 transport downstream. It also implies that sediment storage in space and time within beaver ponds is 778 not a linear function that can be extrapolated from shorter-term deposition rate estimates. In addition, 779 the resuspension and downstream transport of pond sediments is possible without dam breaching 780 (e.g. de Visscher et al. (2014)) ( Figure 12c), which may also account for some of the variability in 781 sedimentation rates that can found within a cascade of beaver dams. In systems with valley bottom 782 spanning beaver ponds and beaver meadows, the longer-term mid-late Holocene sediment deposition 783 rates on the floodplain have been found to be much lower (0.05 cm yr -1 ) than shorter-term pond 784 deposition rates . These floodplain sediments are however usually distributed 785 over a much larger area, and given they are much less influenced by shorter-term dam breaches, the 786 volume of sediment stored on floodplains due to beaver activity is likely to be far more significant over 787 the longer term ( Figure 12c). This is supported by the finding that steeper headwater catchments seem 788 to not preserve longer-term records of beaver pond deposits despite their higher aggradation rates, 789 compared to lower gradient streams which can preserve a wealth of alluvial activity (Persico and 790 Meyer, 2009). 791 It is therefore clear that some sediment will be trapped and sequestered over longer timescales, and 792 some fraction of sediment will continue to be transported through a beaver dam cascade system albeit 793 with some delay. Although we are not aware of previous attempts to do so, it is possible, in principle, 794 to combine these elements into a complete sediment mass balance of this system, from the 795 perspective of beaver dam 796 Where is the storage volume available behind beaver dam , is the concentration of sediment 797 in suspension or available to be transported on the bed behind dam , is the volumetric water flux 798 (inflow or outflow), −1 is the concentration of sediment flowing into dam (potentially from the 799 dam immediately upstream), and is the long-term sediment deposition rate that sequesters 800 sediment away from the active transport pathways. Where many beaver dams occur in a cascade, 801 Equation (9) would be integrated across all dams in the system. We propose Equation (9) because it is 802 conceptually useful, although we also note there are considerable limitations to its use in practice 803 given the paucity of reliable data. However, it is also interesting to use Equation (9) to ask to what 804 extent a system of beaver dams may delay the downstream transport of sediment that is not being 805 sequestered over the longer-term. Analogous to water residence times (section 3.7), we can define 806 = ⁄ as the residence time (or transport delay) of sediment from a single beaver dam. If we 807 then assume all beaver dams have equally sized storages and equal values for (i.e. the delay in 808 sediment transfer is the same between all dams), it is possible to consider how a pulse of sediment (or 809 water) acting as a tracer would pass through this system. Although it is beyond the scope of this paper 810 to provide the full working, substituting into Equation (9) and then performing a Laplace 811 transform, it is possible to evaluate the sediment tracer outflow from the th downstream beaver dam 812 as 813 Equation (10) is a result well known across different fields by different names, for example as the tanks 814 in series residence time distribution used in chemical engineering (Fogler, 2006), and also as the very 815 popular Nash storage cascade rainfall-runoff model in hydrology (Nash, 1957), though takes on a 816 differnt meaning in these separate applications (and is implicitly 0 for the Nash cascade in hydrology). 817 This approach can also be used for tracers of water, however there is often a very large difference 818 between values for (water), which may be on the order of 0.2 -2 days and , which may be closer 819 to the order of 100 -1000 days. Given this important difference, we can apply Equation (10) in a useful 820 thought experiment to consider the implications for tracer outflow as the number of dams increases. 821 If we consider a system where the number of beaver dams ( ) is increasing from 2 to 5, and then to 10 822 beaver dams, = 0 and the time taken for 50% of the water or sediment tracer outflow to be released 823 from the system ( 50 ), then 50 for water will increase from 2.2 days (2 dams) to 9.2 days (10 dams), 824 while 50 for sediment outflow increases from 0.46 years (2 dams) to 2.6 years (10 dams) (Table 3). 825 The assumption of and being equal between all dam structures in a cascade is of course 826 unrealistic. Nonetheless, the thought experiment does show the potential for creating very long delays 827 in sediment transport through beaver dam systems compared to water, especially as the number of 828 dams ( ) becomes large. 829 830 831

832
Established beaver dam cascades reduces the potential for streams to incise, mimicking to some extent 833 artificial grade control structures. However, if and when beaver dams breach, outburst flows can be 834 large and have been reported as damaging roads, rail tracks and pipelines, and also causing mortalities 835 (Butler and Malanson, 2005). The stability of beaver dams depends on many factors, which are largely 836 unexplored, and have been discussed in more detail in the hydrology section. Beaver dams mostly 837 breach during high discharge events when sediment transport capacities and load are at their peak. A 838 breach not only releases water that was previously retained in the beaver pond, but also sediment 839 eroded from the bed directly upstream of the dam. Beaver dams can breach centrally or laterally, and 840 if the latter can also trigger further bank and floodplain erosion as well as channel widening (Demmer 841 and Beschta, 2008). The water and sediment released during dam breaching adds to the already high 842 event discharge and sediment load, however the overall contribution to the event may be small. 843 However, little is known about the longer-term fate of sediments released from breached beaver 844 dams, due to the difficulty of monitoring rare flood events (Jakob et al., 2016). In North America, dam 845 breaches have been documented to easily erode previously deposited pond sediments, re-incising the 846 streams to their previous base level but with minimal lateral bank erosion (Butler and Malanson, 2005). 847 In central Europe, local fisherman observed no noticeable change in channel shape or sediment 848 transport after a managed breach of a beaver dam, until a larger natural flood event initiated a sandy 849 sediment slug which then moved progressively through the downstream river reaches (personal 850 communication, local fishery department Karlstadt, Germany). Hillman (1998) also reports channel 851 incision occurring upstream of a beaver dam breach in the beaver pond deposits, with some evidence 852 for boulder transport, testifying to high sediment transport capacities over short distances following a 853 breach (Butler and Malanson, 2005). One explanation for high transport capacities over short distances 854 might be the local initiation and rapid migration of an alluvial knickpoint at the step in the long-profile 855 created by the sediment wedge on the lee side of beaver dams ( Figure 11 wedge deposited against the dam, which is commonly reported to be between 1 -2 m in thickness 858 (example in Figure 11, section 4.1). Once initiated, the knickpoint then migrates upstream until the 859 slope equilibrates with the upstream and downstream reaches. Knickpoint migration would explain 860 the high but localized increase in sediment transport, and the creation of downstream sediment slugs. 861 Knickpoints can also develop where water has been diverted on the floodplain because of beaver 862 activity and re-enters the channel as return flow via a channel bank (John and Klein, 2004). In this case, 863 knickpoint migration beginning where the return flow breaches the channel bank can also initiate 864 floodplain channel erosion. As already described above, sediment eroded during and following beaver 865 dam breaches may largely be trapped by subsequent beaver dams if a cascade system exists (Burchsted 866 et al., 2010) (Figure 12 c).
Although not yet investigated, it is interesting to speculate that the sediment-867 laden flows generated by beaver dam breaches may also counteract any bed incision that would 868 otherwise occur directly downstream of the breach (Butler and Malanson, 2005; Meentemeyer and 869 Butler, 1999). Beavers dig small channels within floodplains to extend their habitat mobility (Harthun, 1998;Hinze, 874 1950; Hood and Larson, 2015). Beavers also dig channels on the pond floor, which may create sufficient 875 water depths such that the ponds do not completely freeze during winter (Hood and Larson, 2015). 876 These channels have average widths of 60 -90 cm, a depth of 35 -70 cm, relatively steep slopes and 877 can extend more than 100 m in length (Gurnell, 1998;Hinze, 1950), in some instances even up to 300 878 m (Hood and Larson, 2015). They are often interspersed by deeper sections, which are probably used 879 as a refuge. Sediment removed during the digging process is not typically observed adjacent to be the 880 beaver channels on floodplains, so it is likely pushed into the main river channel where it is available 881 for transport further downstream. One study has estimated the magnitude of sediment removed from 882 these smaller channels to be 22,300 m 3 over a 13 km 2 area populated by beavers in Alberta, Canada 883 (Hood and Larson, 2015), thus dpending on the size and transport capacity of the main channels, this 884 may be a significant source of sediment. The development of beaver floodplain channels are also likely 885 to play an important role in increasing the hydrological and ecological connectivity between rivers and 886 floodplains (Hood and Larson, 2015), and in the transport and retention of surface water on floodplains 887 (Westbrook et al., 2013) ( Figure 6, section 3.1). Importantly, these channels greatly improve the areal 888 extent of floodplain wetland development. In Alberta (Canada), the construction of floodplain channels 889 by beavers lead to a 575 % increase in wetland area in one study (Hood and Larson, 2015). If reasonable 890 hydraulic conductivity values can be maintained, they may also facilitate the rise in shallow ground 891 water levels typically found adjacent to beaver dams (section 3.5). However, the creation of channels 892 may already depend on relatively high floodplain ground water levels in the first place, as beavers may 893 preferentially construct channels when the height difference between in-channel water level and 894 floodplain is relatively small (Stocker, 1985). This may be because in more incised river systems beaver 895 channels could be very effective in draining the floodplain surface, and thus be counterproductive in 896 terms of wetland habitat creation. 897 In addition to building dams, beavers also burrow into channel banks and floodplains, and can steepen 898 river banks and lead to destabilization and collapse (Figure 13 c, d). The length of these burrows is 899 usually less than 10 m, but they may extend up to several 100 m, and are around 15 -30 cm in diameter 900 with occasional widened sequences (Djoshkin and Safanow, 1972). Studies have found a complicated 901 network of burrows in the subsurface of older beaver colonies (Djoshkin and Safanow, 1972), meaning 902 that their influence on bank stability can potentially be significant. When beaver burrows collapse, 903 they can create preferential flow paths for infiltration, which can further enhance bank erosion, and 904 finally promote channel widening. This mechanism has been suggested to enhance lateral migration 905 of streams (Giriat et al., 2016), but quantitative studies examining the extent to which this may occur 906 are still needed. Collapsed beaver burrows have also been observed to create spillways and the 907 diversion of stream water around the main dam, which over time are likely to incise and create side 908 channels (Giriat et al., 2016). Within beaver ponds, underwater digging activities by beavers (e.g. 909 removal of sediments from the base of banks after failure) in combination with sediment instability 910 due to pore water pressure changes and fluvial erosion and deposition processes lead to a general 911 widening of the beaver pond, which then contributes to a widening of river sections in the case of dam 912 breaching (Figure 13 b,e) (Giriat et al., 2016). In contrast, Polvi and Wohl (2013) argued that beavers 913 increase bank stability because they promote the deposition of finer sediment on floodplains, which 914 provides more cohesive and higher river banks. Abandoned dams incorporated into the stream banks 915 may also reinforce bank stability, thus helping to limit channel migration and promote a combination 916 of bed incision and high-angle channel bends ( Figure 14). Also important for bank stability is the 917 possible rise in shallow groundwater levels near beaver dams (see section 3), and any change in 918 riparian vegetation root mass, which can shift if there is dieback of existing tree species and a 919 promotion of pioneer species vegetation assemblages (see section 6.5). There is also the importance 920 of changes in pore pressure as surface water recedes following dam breaching and pond drainage in 921 promoting bank instability. In summary, whether or not beaver activity enhances or reduces bank 922 stability will depend on the extent of burrowing activity, the frequency of dam disruption and pond 923 drainage, fine sediment deposition, and groundwater-vegetation feedbacks over the longer term. 924 Further long-term research is clearly needed to better understand the relative importance of these 925 different drivers. It has been long suggested that beavers have had an important influence on long-term valley 932 formation. Beaver damming activity was descried by Rudemann and Schoonmaker (1938) as 933 generating "gently graded, even valley plain, horizontal from bank to bank" river corridors, as the agent 934 of valley floor aggradation that is enhanced over time by their valley-wide beaver dam construction 935 (Ives, 1942). Their medieval eradication in western Europe has also been put foreward as one 936 explanation for the expansion of braided river planforms, at the expense of more channelised patterns 937 with wetlands, across post-glacial river valleys draining from the European Alps (Rutten, 1967). These 938 earlier studies argued that although beaver dams disappear over time, their accumulated floodplain 939 and meadow deposits remain, forming fertile river valleys. Buried beaver dams found in the Colorado 940 headwaters also lend some weight to this hypothesis (Ives, 1942 In any case, the long-term aggradation rates on floodplains and meadows influenced by beaver 952 damming is low compared to ponds (table 2), and also heterogeneous in time and space due to the 953 highly variable beaver occupation and landscape constraints (Persico and Meyer, 2009;Polvi and Wohl, 954 2012). Most beaver-induced changes to long-term valley floor evolution are attributed to the creation 955 of wet beaver meadow complexes (Ives, 1942;Polvi and Wohl, 2012), which are considered to develop 956 due to a combination of: (1) damming and flow diversion onto floodplains, facilitating sedimentation, 957 (2) the silting-up of shallow ponds on floodplains, (3) the introduction of wood into channels, further 958 facilitating flow diversion and a decrease in stream power, (4) beaver floodplain digging activity 959 channelizing flow diversion, and (5) rising shallow ground water levels and associated vegetation 960 feedbacks, promoting grasses and sedges which can also effectively trap sediments, and the reduction 961 of tree species (see section 6.5, Figure 13). Following the introduction of beaver dams, some of the 962 largest terrestrial ecosystem impacts are within beaver meadows and wetlands (see section 6).The 963 persistence of beaver meadows and implications for vegetation, nutrient cycling, and carbon storage 964 is covered in section 7.2. 965 One of the most profound long-term geomorphic influences of beavers is their suspected capacity to 966 change postglacial fluvial channel patterns, with implications for the aquatic and terrestrial ecosystems 967 within these river corridors Wohl, 2013, Rutten, 1967). Examining gravel-bed river corridors 968 with a snow-melt hydrological regime and set in semi-confined mountain valleys partially dammed by 969 glacial moraines, Polvi and Wohl (2013) hypothesize that beavers came to occupy postglacial 970 environments after they had transitioned from braided to single thread, meandering channel 971 planforms, since this would have provided the riparian vegetation necessary for beaver populations to 972 thrive. This may not be an exclusive transition, and changes to anabranching systems with vegetated 973 islands may have also be sufficient. Beavers may also promote anabranching channel planforms due 974 to (1) the water diversion processes as a result of damming, (2) fine sediment accumulation on valley 975 floors, and (3) increased wood in streams, forming, for example, log jams and promoting partial flow 976 diversion (Polvi and Wohl, 2013). More specifically, Polvi and Wohl (2013) hypothesize that beaver 977 occupation and meadow development follows a long-term sequence from the post-glacial recovery of 978 vegetation leading to the creation of log-jams within early post-glacial braided rivers, which in turn 979 promotes fine sediments deposition, and the initial creation of floodplains. Beaver meadow vegetation 980 is well adapted to inundation, which then sufficiently stabilizes banks, islands and floodplain patches 981 to create avulsion and promote stable anabranching channel patterns. In contrast, the removal of 982 beaver dams and log-jams would promote incision and contraction to a single, mostly meandering 983 channel system. It has also been suggested that the widespread and rapid removal of beavers from 984 dryland, discontinuous streams in the US ('arroyos') is one reason for post-European settlement 985 channel incision response, and to the evolution of the modern continuous stream networks (Cooke 986 and Reeves, 1976, Fouty, 2018). A key feature of discontinuous streams is a relatively stable 987 aggregational surface within a section of the channel and floodplain, a feature that is often associated 988 with local wetlands. The historical accounts of these wetlands in US drylands have all the 989 characteristics of beaver meadows and their wetland complexes, though this is not definitive evidence 990 of causation since beaver wetlands can appear very similar to non-beaver wetlands (Fouty, 2018). It 991 has therefore been suggested that once beavers were removed from these streams, the wetlands 992 dried up, the vegetation cover disappeared, and the channels incised and became continuous (Cooke 993 and Reeves, 1976; Fouty, 2018). In the gravel-bed rivers of non-glaciated regions in the north-east USA, 994 the pre-European Holocene deposits dominated by fine-grained organic-rich sediments have been 995 interpreted as the product of small anabranching channels within extensive vegetated wetlands 996 (Walter and Merritts, 2008), an interpretation that is also consistent with beaver meadow 997 characteristics. In Europe, the long-term influence of beavers on river valleys are difficult to determine, 998 because of the widespread eradication of beavers between ~ 1000 -150 years ago (Zahner et al., 2005). 999 However, John and Klein (2004) have also observed an anabranching planform emerge in southern 1000 Germany a decade after beaver re-introduction. Nonetheless, the suggested geomorphic feedbacks 1001 between beaver engineering and long-term river corridor vegetation dynamics may re-inform 1002 traditional models of biogeomorphic succession (e.g. Corenblit  Changes to the biogeochemical functioning of beaver impacted systems, and therefore their potential 1011 impact on riverine water quality and ecosystem processes, can be divided into their influence on (i) 1012 pathways, i.e. modification of existing pathways or introduction of pathways not previously present, 1013 (ii) the spatial extent of these pathways and their rates, and (iii) the degree to which water flowing 1014 through the system can interact with these pathways (i.e. residence time and hydraulic efficiency). 1015 Impacts on these processes have important consequences for aquatic and terrestrial ecosystem 1016 processes and productivity, which in turn will also produce positive or negative feedbacks on the 1017 biogeochemical cycling. Thus, from a mass balance perspective the development of beaver ponds, 1018 wetlands and meadows may create both sources and sinks of e.g. carbon, nitrogen, and phosphorus 1019 in the riverine nutrient cycles ( Figure 15). However, it remains unclear when and how these process 1020 modifications should interact over different spatial (e.g. one vs many beaver dams) and temporal (e.g. 1021 event, seasonal, annual) scales. 1022

1023
In terms of potential changes to biogeochemical pathways, the combination of increased surface water 1024 inundation extent, turbulence reduction, higher temperatures, and higher floodplain water tables can 1025 combine to diminish dissolved oxygen concentrations and enhance the extent of anaerobic conditions 1026 present in beaver impacted systems (Dahm et al., 1987;Naiman et al., 1994). This spatial enhancement 1027 of anaerobic conditions is typically focused along saturated boundaries with limited turbulent 1028 exchange, for example within benthic ponds and wetland areas where biofilm communities are 1029 abundant, which typically contain a variety of aerobic and anaerobic metabolic pathway communities 1030 (Battin et al., 2016) or within permanently or seasonally saturated floodplain or meadow soils. The 1031 enhancement of anaerobic conditions is important since a shift from aerobic to anaerobic metabolism 1032 will tend to slow the overall rate of organic matter cycling, and utilize electron acceptors beyond 1033 dissolved oxygen, such as nitrate (NO3 -), iron (Fe) and manganese (Mn) oxides, sulfate (SO4 2-), and 1034 eventually CO2. This in turn creates new loss pathways for the nitrogen, carbon and sulfur cycles via 1035 reduction to atmospheric nitrogen (N2) (or nitrous oxide -N2O), methane (CH4), and hydrogen sulfide 1036 (H2S) respectively, as well as concentration enrichment pathways for Fe, Mn, and aluminum (Al) via 1037 the dissolution of their respective oxides. The breakdown of organic matter containing appreciable 1038 nitrogen under anaerobic conditions will also yield ammonium (NH4 + ), which can be subsequently 1039 oxidized to NO3 -(via nitrite -NO2 2-, i.e. nitrification) if transported back into aerobic conditions or 1040 internally cycled within biofilm communities. This potential re-oxidation pathway has the capacity to 1041 counteract or diminish any reduction in NO3 -(due to denitrification) downstream of beaver dam 1042 complexes, depending on the rates and extent of mineralization (NH4 + production) and subsequent 1043 nitrification (to NO3 -). NH4 + can also be taken up directly by many plant communities, which may be an 1044 important pathway in beaver meadow or wetland development (Naiman et al., 1994). Enhanced 1045 anaerobic conditions also have implications for the phosphorus cycle, as organic matter breakdown 1046 may release orthophosphate, in addition to the phosphorus absorbed onto mineral surfaces (e.g. Fe 1047 oxides) that is released as these minerals dissolve following the transition from oxic to anoxic 1048 conditions. With the enhancement of anaerobic conditions and associated biogeochemical pathways 1049 in beaver impacted systems, a key question is therefore how these biogeochemical pathways and rates 1050 will act in combination with changes to the overall storage of nutrients to influence any net changes in 1051 water quality and ecosystem dynamics. These feedbacks, over a range of timescales, are critical to 1052 understand since they will determine the implications of beaver modification for the riverine carbon, 1053 nitrogen, and phosphorus cycles and the ecosystems which depend on them ( Figure 15). 1054

1055
In terms of the carbon cycle, a key consideration in determining the relative impact of beavers is the 1056 carbon storage existing within the landscape prior to beaver modification. If floodplain forests are 1057 present, then the standing carbon stored in woody biomass will be greatly reduced as a result of 1058 floodplain inundation and rising water tables (Naiman et al., 1994), in addition to species specific tree 1059 felling and consumption by the beaver populations (see section 6.5) (Martell et al., 2006;Mitchell and 1060 Niering, 1993). The death and felling of these forests following inundation may in some cases create 1061 substantial storages of submerged woody biomass; (Johnston, 2017; Thompson et al., 2016). If 1062 widespread floodplain forest is not initially present, at the very least, reductions in riparian zone woody 1063 biomass is likely (Martell et al., 2006;Stabler, 1985). Thus, as beaver modifications promote the 1064 expansion of lentic open water area and anaerobic conditions, there is the potential for significant net 1065 transfers of carbon stored as woody biomass carbon to herbaceous and grass biomass, as well as 1066 increased sediment carbon storage (Johnston, 2014;Naiman and Melillo, 1984; (Wohl, 2013). 1067 Furthermore, much of the woody biomass that enters the beaver system, either from landscape 1068 conversion, or via the fluvial network, may not be very labile relative to other carbon inputs 1069 (Hodkinson, 1975). In general, woody biomass can provide some soluble sugars and cellulose during 1070 the initial stages of decomposition, however the large fraction of remaining lignin in woody biomass is 1071 notoriously slow to decompose (Reddy and DeLaune, 2008). Adding to this context, a very important 1072 experimental finding from Naiman et al. (1986) was that the expansion of anerobic conditions due to 1073 beaver daming considerably reduced the decomposition rates (by 81% and 61%) of both labile and 1074 non-labile woody biomass inputs respectively, compared to downstream aerobic riffle environments. 1075 This promotion of anerobic environments, slower decomposition rates, and abundance of refractory 1076 woody carbon is therefore condusive toincreased long-term carbon storage. Beavers can themselves 1077 also directly import large masses of plant detritus and woody material into the river corridor that 1078 contributes to carbon storage. Additional mechanisms by which beavers can increase carbon storage in river corridors include 1) 1087 trapping of allochthonous particulate organic carbon (POC) inputs, and 2) through greater 1088 autochthonous inputs derived by increasing net aquatic ecosystem productivity (NEPaq, or gross 1089 primary production minus respiration). In terms of 1), POC inputs can include: leaf litter and small twigs 1090 and branches (macro-organics), as well as coarse and fine POC fractions which come in various stages 1091 of decomposition and from a variety of sources. These sources of POC may have some overlap with 2), 1092 increased NEPaq, especially for the fine POC fractions. These overlaps arise depending on the scope of 1093 NEPaq feedbacks considered within beaver systems. If NEPaq from only the lentic (pond) zone is 1094 considered, benthic biomass increases but is generally a small percentage (e.g.: 4 -12%) of the carbon 1095 budget for beaver impacted systems (Hodkinson, 1975;Stanley et al., 2003). In contrast, if the 1096 promotion of new littoral zone and wetland habitat vegetation is also considered, the increase in 1097 NEPaq, and therefore autochthonous inputs to C storage, may be far more substantial (Hodkinson, 1098 1975; Stanley et al., 2003). This increase in NEPaq is also discussed in section 6.1, suffice to say it is 1099 critical to recognize as it builds a foundation for changes to carbon cycling and storage in river corridors 1100 impacted by beavers (Mann and Wetzel, 1995 , and even within a single pond 1127 (Weyhenmeyer, 1999;Yavitt et al., 1992). These increased CH4 fluxes, and to some extent CO2 fluxes, 1128 along with their high spatial and temporal variability, are a result of the expanded benthic anaerobic 1129 conditions following beaver impacts promoting metabolic pathways that include methanogenesis. 1130 However, CH4 fluxes are also higher in beaver ponds per unit area compared to similar water bodies, 1131 which as Weyhenmeyer (1999) notes, raises the question as to whether this is due to higher methane 1132 production rates, differences in methane oxidation rates in the sediments and water column, or some 1133 combination of both. In terms of CH4 production rates, this could be due to higher organic carbon 1134 quality (Weyhenmeyer, 1999), perhaps as a result of inputs from the the relatively high ecosystem 1135 productivity noted earlier, though this remains speculative and needs further research. In terms of 1136 differences in oxidation rates, this question may come down to the relative importance of ebullition, 1137 which Weyhenmeyer (1999) found to dominate (65%) over diffusive fluxes in a beaver pond in Ontario, 1138 Canada. Though only a single study, this is important as it would shift the dominant controls on CH4 1139 flux sensitivity being mainly due to water depth in the case of diffusive fluxes, which have been shown 1140 to be susceptible to significant oxidization in the water column, even in relatively shallow beaver ponds 1141 (Yavitt and Fahey, 1994;Yavitt et al., 1990), and more towards atmospheric pressure and sediment 1142 temperatures (Weyhenmeyer, 1999). Nonetheless, even if the diffusive fluxes are a smaller 1143 component, they are still likely to be significant enough to permit water depth, and thus also beaver 1144 pond hydrology and wetland hypsometry, to play an important role. Indeed, Fahey (1994) 1145 found the CH4 tended to be higher, though not always, in beaver ponds with shallower water depths. 1146 An interesting result was also found by Yavitt et al. (1990) where the flowing water river sections 1147 between beaver dams tended to have higher CH4 fluxes than the ponds themselves. This makes sense 1148 from the perspective of the streams having higher turbulent fluxes, but only if a high CH4 supply can 1149 be maintained, suggesting hyporheic and groundwater flow from the upstream ponds and wetlands 1150 are in this case able to subsidize the downstream CH4 fluxes from the stream. In terms of CO2, it is 1151 important to note that some anaerobic pathways produce, and others consume, CO2. Thus, it is difficult 1152 to make general speculations on the extent to which CO2 fluxes should increase. Nonetheless, small 1153 water bodies are known to disproportionately contribute to natural CO2 and especially CH4 evasion 1154 (Holgerson and Raymond, 2016), and the areal extent of small water bodies generated by beavers is 1155 increasing (Hood and Bayley, 2008;Nisbet, 1989;Whitfield et al., 2015), especially in boreal zones 1156 (Nisbet, 1989). For this reason, it is important to consider the role of beavers on regional and global 1157 CH4 emissions, and Whitfield et al. (2015) have estimated a ~20x increases in CH4 emissions from 1158 expanding beaver ponds and wetlands over the last century across Europe and North America. This 1159 outsized influence on CH4 emissions per unit water area led Moore (1988) to wonder "whether the 1160 beaver is aware the greenhouse effect will reduce the demand for fur coats". Nonetheless, it is critical 1161 to emphasize that speculation regarding beaver impacts on CO2 and CH4 emissions should be placed in 1162 the context of both the total greenhouse gas emission flux (~0.001% of total CH4 emissions) as well as 1163 the full carbon mass balance of the aquatic system being studied, especially the increase in carbon 1164 storage, which is discussed in greater detail later in this section. 1165 An additional mechanism of carbon export from beaver systems is downstream fluvial transport, which 1166 comprises three main components: dissolved inorganic (DIC), dissolved organic (DOC), and particulate 1167 organic (POC) carbon. Within fluvial systems, DOC is typically the dominant export mechanism 1168 interacting with the organic carbon storages (Regnier et al., 2013). However, with the expansion of 1169 anaerobic conditions following beaver modifications, HCO3is also produced via multiple pathways 1170 (e.g. NH4 + production, Mn 2+ , Fe 3+ , and SO4 2reduction) which typically dominates total DIC under the 1171 pH range of natural surface waters (Reddy and DeLaune, 2008). Given sufficient concentrations, HCO3 -1172 will also contribute to additional CO2 outgassing and even to stream biofilm precipitates. Cirmo and 1173 Driscoll ( immediately downstream of beaver dams, which then tended to decrease with distance downstream. 1175 This suggests the production of higher concentrations of HCO3in beaver systems were being 1176 subsequently diminished by conversion in the carbonate system to CO2 (Cirmo and Driscoll, 1993;1177 Margolis et al., 2001), which is another potentially important source of CO2 evasion related to beaver 1178 impacts, but one that is not captured by the focus on pond water quality measurements behind the 1179 dams. 1180 In terms of DOC export fluxes, a largely consistent finding is an overall increase in DOC concentrations 1181 downstream of beaver systems (Figure 16). Although this result only considers the direction of change 1182 in DOC and not the magnitude, it nonetheless suggests sufficient reactive transport interaction 1183 between the increased organic carbon production, storage and residence times of flowing water within 1184 beaver systems to drive net increases in DOC concentrations. This represents a profound change in 1185 riverine DOC behavior relative to what would occur in these same river reaches in the absence of 1186 beaver impacts, with important implications for carbon export dynamics and ecosystem processes. It 1187 is also largely consistent with the impact of similar within stream network lakes and wetlands that 1188 buffer river flow and enhance DOC concentrations (e.g. Kalinin et al. (2016); Kling et al. (2000). This is 1189 because a comparatively low NEPaq environment (e.g. the forested stream) flows into a higher NEPaq 1190 lentic environment (e.g.: lake, wetland, beaver pond) which as a result has to establish enhanced 1191 carbon storage and cycling feedbacks (Kalinin et al., 2016;Kling et al., 2000;Wetzel, 2001). This is also 1192 supported by the few studies that have examined sub-annual dynamics (e.g. seasonal, monthly, event) 1193 in beaver impacted systems, where the majority have found outgoing DOC fluxes, and to some extent 1194 DIC fluxes, to be strongly seasonal, likely reflecting the importance of wetland vegetation and algal 1195 biomass production and breakdown as well as hydrological feedbacks (Mann and Wetzel, 1995). The 1196 hydrological feedbacks include enhanced riparian soil carbon interaction as beaver dams cause water 1197 levels to rise (on average, as well as seasonally), which has been found to increase pond DOC 1198 concentrations (Hill and Duval, 2009;Wang et al., 2018). This is also a potential mechanism that can 1199 explain the increase in DOC concentrations following beaver related water level increases in Finnish 1200 lakes (Vehkaoja et al., 2015). However, Nummi et al. (2018) suggest the initial DOC sources following 1201 damming are from the decay of existing organic matter stocks rather than new interactions with 1202 riparian and littoral zone organic matter. This mechanism is in contrast to most other studies examining 1203 DOC source and export dynamics that emphasize the importance of hydrological feedbacks with the 1204 riparian zone, however it does highlight the need to better understand the unique DOC source-sink 1205 dynamics that may occur in beaver systems. 1206 Changes in the quality of DOC could also provide insights into the availability of these different carbon 1207 sources as well as the implications for downstream ecosystem carbon cycling. However, there is 1208 relatively little information available on DOC quality from beaver impacted systems. Two studies that 1209 have examined DOC quality changes, found either no change in total DOC (Koschorreck et al., 2016) or 1210 a decrease (Kothawala et al., 2006) in total DOC due to beaver impact, results which are unusual 1211 compared to the majority of findings (Figure 16). The decrease in DOC found by Kothawala et al. (2006Kothawala et al. ( ) 1212 was accompanied by a corresponding decline in the molecular weight of DOC, with both these factors 1213 potentially dependent on the unusually high DOC inputs from the headwater swamp upstream. 1214 Koschorreck et al. (2016) found no significant difference in either DOC or quality (as measured by UV 1215 indices) from sites draining beaver dams, though by study design (paired catchment, rather than 1216 upstream -downstream comparison) these results are somewhat inconclusive. The quality of DOC and 1217 its concentration within beaver ponds is also likely to be dependent on the age of the system given the 1218 observed evolution in biogeochemical cycling from initial damming to pond systems that have been 1219 functioning for >10 years (Catalán et al., 2016). In this case, there is a hypothesized increase in labile 1220 carbon during the early stages of beaver impact which then diminishes with age (Ecke et al., 2017). 1221 However, the extent and timescales over which this should occur remain speculative. In an already 1222 well-established beaver dam system, Mann and Wetzel (1995) found the increase in DOC due to beaver 1223 impacts is not necessarily accompanied by a change in bioavailability, however the limited sample 1224 comparisons emphasize the clear need for further work in this area. is clear from spatial snapshots beaver systems can act as significant sinks for coarse and fine POC, 1241 further research is clearly needed to examine the significance of POC within the overall carbon budget, 1242 especially given the near ubiquitous increase in woody debris introduced by beavers to river corridors 1243 (Anderson et al., 2014;Thompson et al., 2016). This is also important because the POC filtering vs 1244 production effectiveness of beaver systems will regulate the downstream delivery of this important 1245 component of the aquatic carbon cycle. 1246 The full mass balance of changes to the storage and fluxes of carbon that can occur as result of beaver 1247 modifications, especially across the spectrum of terrestrial and aquatic carbon sources and sinks, 1248 remains poorly understood (Nummi et al., 2018;Wohl, 2013). This is partly because the mass balance 1249 strongly depends on the spatial and temporal frames of reference considered, and the availability of 1250 suitable controls for context. For example, some studies consider the change in storage and fluxes with 1251 respect to the beaver pond , and others the change in carbon storage within the 1252 beaver modified wetlands and floodplains (Wohl, 2013). Such frameworks are potentially confusing, 1253 since beaver modifications can both create conditions for enhanced storage as well as aquatic and 1254 terrestrial primary production (e.g. wetland vegetation and biofilms). Thus, the increase in exported 1255 fluxes (POC, DOC, CO2, CH4) is likely to be due to some combination of increased allochthonous carbon 1256 storage, as well as enhanced in situ carbon production (NEPaq) and decay, both of which can be highly 1257 interactive with water flow paths through the system. As already mentioned, the large expansion of 1258 anaerobic conditions is likely to be a key driver of these increases in both aquatic Driscoll, 1259 1993;Naiman et al., 1986) and terrestrial (Johnston, 2014;Wohl, 2013) carbon storages in beaver 1260 modified systems. These changes to carbon storage and fluxes also have implications for the residence 1261 time of carbon in river channel and floodplain systems, which will increase as storage increases in order 1262 to maintain continuity in the carbon mass balance, although this is unlikely to ever reach steady state 1263 given the large variation in timescales over which the different storages and fluxes operate (see also 1264 section 7.2). 1265 1266 1267 5.3 Beaver impacts on the Nitrogen cycle 1268 In terms of changes to the nitrogen cycle, the documented increase in organic carbon storage within 1269 beaver impacted systems is likely to also be accompanied by some increase in total organic nitrogen 1270 storage (Naiman and Melillo, 1984). Francis et al. (1985) estimate large increases in organic nitrogen 1271 accumulation once beaver ponds are established, relative to what would accumulate in their absence 1272 (e.g. within riffle sequences). This is not necessarily because nitrogen uptake rates are enhanced, but 1273 rather due to the large spatial increase in biofilm extent across beaver pond sediments (Francis et al., 1274(Francis et al., 1985, as well as the expanded sequestration of initial and new organic matter inputs (Devito and 1275 Dillon, 1993). Naiman and Melillo (1984) also found beaver impacted systems greatly enhanced 1276 nitrogen storage (per unit length or area) within beaver pond sediments, and similarly found this was 1277 likely to be due to the increased biofilm uptake of nitrogen. However, it remains unclear as to whether 1278 such large increases in nitrogen storage are restricted to more nitrogen-limited systems (Naiman and1279 Melillo, 1984), and whether this should change as nitrogen availability also changes. Beaver vegetation 1280 consumption and waste can itself also be a considerable input of nitrogen and phosphorus to the 1281 system (Naiman and Melillo, 1984). Uptake of inflowing nitrogen (primarily NO3and NH4 + ) by wetland 1282 vegetation has been found to be a key seasonal storage component (Devito and Dillon, 1993;Naiman 1283 and Melillo, 1984). However, the degree of long-term sequestration is unclear since this biomass also 1284 undergoes seasonal decay. Within sediment and soil pore waters, NH4 + diffusively released during the 1285 biomass decay process (mineralization) will also increase the total nitrogen storage provided anaerobic 1286 conditions are maintained and the advective transport is slow. This is supported by evidence from 1287 Dahm reported an order of magnitude increase in NH4 + concentrations (as well as very low NO3 -1289 concentrations) due to organic matter breakdown within beaver impacted sediment pore waters 1290 relative to sites without beaver impacts. In colder climates, the capacity for beaver ponds to develop 1291 ice cover also been found to promote both increased anaerobic conditions and NH4 + production 1292 (Devito and Dillon, 1993). In terms of export, downstream increases in NH4 + due to beaver damming 1293 have been found within the majority of studies in which NH4 + concentrations have been reported 1294 (Figure 16 c). However, NH4 + export or retention may have a large seasonal bias (Devito andDillon, 1295 1993;McHale et al., 2004), and the production of higher NH4 + concentrations will not necessarily be 1296 sustained for significant distances downstream given the likelihood of nitrification to NO3 -. 1297 In addition to these potential storage changes for nitrogen, the increase in anaerobic conditions 1298 provides an important avenue for denitrification, primarily within benthic biofilms and subsurface 1299 microbial communities (Lazar et al., 2015). This increase in denitrification capacity, in some 1300 combination with biomass uptake, likely explains the general decrease in NO3concentrations 1301 downstream of beaver impacted systems identified in the majority of published studies (Figure 12b). 1302 However, it should be noted that the magnitude of this reduction varies markedly between studies. As 1303 already noted NH4 + can also be converted to NO3 -, meaning the overall impact of beaver modifications 1304 on downstream nitrogen fluxes is not clear. Studies that have tracked both NH4 + and NO3with 1305 increasing distance downstream of beaver systems have found the initial increases in NH4 + are 1306 subsequently diminished while NO3increases (Błȩdzki et al., 2011, Harthun, 2000, strongly suggesting 1307 nitrification may be an important pathway to consider downstream of beaver systems where aerobic 1308 conditions again dominate. All these uncertainties in combination highlight the need for a more 1309 comprehensive mass balances of nitrogen dynamics within beaver impacted systems. 1310 Despite these knowledge gaps, the literature seems clear on the increased likelihood of net retention 1311 of NO3 - (Figure 12b) and net export of NH4 + (Figure 12c), within the caveats already mentioned above, 1312 and a less clear likelihood of increased organic nitrogen retention (Devito and Dillon, 1993; McHale et  1313 al., 2004) within beaver impacted systems (also see section 5.6 for further discussion on source vs sink 1314 behaviour). Increasing atmospheric fluxes as from beaver ponds as N2 have also been found (Lazar et 1315(Lazar et al., 2015. Interestingly, this study also found that pond conditions were sufficiently anaerobic to allow 1316 complete denitrification, thus limiting the fluxes of N2O and allowing most atmospheric losses to occur 1317 as N2 (Lazar et al., 2015). Taken together, these findings are largely consistent with syntheses of 1318 nitrogen dynamics in river systems interacting with wetlands and lakes without beaver impacts, 1319 whereby the mechanisms of nitrogen retention in order of decreasing importance have been found to 1320 follow: denitrification > sedimentation > biomass uptake (Saunders and Kalff, 2001). If this sequence 1321 also holds in beaver impacted systems, this suggests the reduction in downstream NO3is being driven 1322 primarily through an increase in the atmospheric losses, and secondarily as increasing within-system 1323 storage, however the limited evidence thus far on full nitrogen cycling in beaver systems highlights 1324 much more work remains to be done in this area. 1325

1326
The development of beaver ponds and wetlands is likely to lead to a large increase in the storage of 1327 total sorbed and particulate phosphorus (Devito and Dillon, 1993;Maret et al., 1987), given it also 1328 creates a large storage capacity for suspended sediment and organic matter, to which a large fraction 1329 of available phosphorus is sorbed (e.g.: Fe oxides) or complexed within. Although the total storage of 1330 phosphorus may increase, so too will the likelihood of sediment exposure to anaerobic conditions in 1331 beaver modified systems. Thus, phosphorus sorbed to redox-sensitive mineral phases such as Fe or 1332 Mn oxides may be readily released as dissolved orthophosphate (PO4 3-) as these phases dissolve under 1333 anoxic conditions (Klotz, 1998). Separately, PO4 3concentrations may also increase under anaerobic 1334 conditions due to the mineralization of organic phosphate (Roden and Edmonds, 1997). However, the 1335 extent to which these mechanisms separately contribute to phosphorus dynamics in beaver impacted 1336 systems is not understood. This contrast between increased storage potential and the ability to release 1337 phosphorus under anaerobic conditions may explain the lack of consistency in the downstream 1338 behavior of PO4 3concentrations in beaver impacted systems across all published studies (Figure 16d). 1339 Seasonal biomass uptake of phosphorus and release during decay may also contribute to this lack of 1340 trend, although this effect is likely to be smaller in magnitude than the influence of storage changes 1341 and the availability of anaerobic flow paths (Reddy and DeLaune, 2008). Fuller and Peckarsky (2011) 1342 found beaver systems were more likely to retain or release phosphorous depending on whether the 1343 vertical hydraulic gradient over the dam(s) was low or high respectively. This interesting result doesn't 1344 reveal a clear mechanistic explanation but highlights the need to better understand how the extent of 1345 anaerobic conditions, transport and residence times, and increases in phosphorous storage conspire 1346 to determine the magnitude of phosphorous retention or export downstream of beaver systems. 1347 Moreover, the export or retention of phosphorous may depend on the form measured, Devito and 1348 Dillon (1993) monitored the outflow of a beaver pond in Canada and found that PO4 3was more likely 1349 to be retained, and organic phosphorous was more likely to be released. This may also explain the 1350 results found by Smith et al. (2020), in which PO4 3concentrations diminished downstream of a beaver 1351 pond in Germany, but total phosphorous concentrations remained the same. The variability in PO4 3-1352 responses downstream of beaver systems (Figure 16d) therefore presents some difficulty in terms of 1353 broader mechanistic interpretations, however some constraints are possible to outline. If PO4 3-1354 decreases downstream, then it is likely that any increase in phosphorus storage occurred without 1355 sufficient exposure to anaerobic flow paths. Conversely, if PO4 3increases downstream, then it is likely 1356 that increases in phosphorus storage were exposed to sufficient anaerobic flow paths, and that the 1357 conditions at the point of sampling did not yet diminish these increased concentrations via re-sorption 1358 or biomass uptake as aerobic conditions returned. There may also be a beaver dam age effect; in large 1359 review, Ecke et al. (2017) found on average beaver dams released phosphorus (albeit with considerable 1360 variation), but that this was mostly in younger beaver dams, with older dams more likely to retain 1361 phosphorus. In any case, the clear lack of dominance in either response, as well as the large frequency 1362 of 'no change' in downstream PO4 3concentrations (Figure 16d) also suggests these competing 1363 mechanisms are likely to be of similar magnitudes in beaver impacted systems. 1364 These mechanisms are important to consider because phosphorus is often considered to be the key 1365 limiting nutrient for primary production in freshwater ecosystems. However, under natural conditions 1366 (i.e. limited human impact), and depending on the stoichiometry of primary producers, nitrogen can 1367 sometimes be equally limiting. Thus, the degree of phosphorus or nitrogen limitation within beaver 1368 impacted systems, and therefore the overall impact on downstream water quality, will depend to some 1369 extent on the supply from upstream land use, as well as atmospheric deposition in the case of nitrogen. 1370 Given the high seasonal loadings of nitrogen in many areas of Europe and North America, it is 1371 reasonable to expect phosphorous also to be the limiting nutrient and thus its downstream availability 1372 may be determined to a large extent by beaver dam construction and whether these new conditions 1373 promote phosphorus retention or release. 1374 5.5 Impacts on iron cycling, mercury, and additional contaminants 1375 Aside from the cycling of the major nutrients, beaver impacts also have potential implications for other 1376 nutrients and contaminants, especially those that are redox sensitive given the expansion of anaerobic 1377 conditions that can occur. As already mentioned in the phosphorus cycle (section 5.4), Fe-oxides are 1378 particularly sensitive to changing redox conditions, and high concentrations of Fe 3+ , due to the 1379 reduction of Fe 2+ , have been found in the pore water of beaver impacted systems (Donahue and Liu 1380 1997). This is a pathway for the liberation of sorbed phosphorus, and also for some metal contaminants 1381 such as arsenic. The cumulative effects of these expanded pathways are not well known in beaver 1382 systems, but it is nonetheless a mechanism to increase the concentration of Fe 3+ and associated metals 1383 and nutrients in solution, which may then in turn be re-oxidised by a variety of abiotic and biological 1384 mechanisms if these pathways re-enter downstream anaerobic surface waters ( Figure 15). although it is important to emphasize the data on this potential impact remains quite limited. 1403 Given the array of hydrological and biogeochemical changes that beaver impacts may introduce to 1404 river systems, it is likely they will have a role to play in the cycling of additionally important and 1405 emerging contaminants, such as pesticides, pharmaceuticals, and microplastics, all of which remain to 1406 be examined. This is especially the case in river systems under the burden of industrial or urban 1407 pollution, and that also may have re-emergent beaver activity. The demonstrated capacity of beaver 1408 impacts to increase water, sediment, and nutrient storage within expanded anaerobic conditions is 1409 likely to influence the storage, residence time, and cycling of pesticides and pharmaceuticals with a 1410 wide variety of breakdown pathways (e.g. redox or photo oxidation sensitivity). Microplastics and 1411 other particulate urban or industrial pollution may also find a high storage and retention capacity 1412 within beaver dam complexes, and one that has the potential to be far more efficient than river 1413 reaches without beaver impacts. 1414 5.6 Impacts on source vs sink behavior, and the evolution of overall water quality and 1415 its variability 1416 Understanding the diversity of water quality impacts from beaver modifications requires some insights 1417 from the coupling between water transport and biogeochemical reactions, and how these are likely to 1418 change. However, a formal quantitative analysis is difficult given the need to derive full mass balances 1419 of both nutrients and water within beaver modified systems, which are unlikely to be in steady state 1420 at sub-annual scales (e.g.: water) or even at annual (e.g.: nitrogen) or decadal (e.g.: carbon and 1421 phosphorus) time scales. Nonetheless, it is an important issue to address since it can help explain the 1422 extent to which a river corridor will act as a source or sink, which can be far more dynamic following 1423 beaver impacts (Wegener et al. 2017), as well as how efficiently each source or sink may be operating. 1424 An insightful analysis in this regard was provided by Stanley and Ward (1997), who compared the net 1425 retention of different nitrogen components (total nitrogen, NO3 -, NH4 + ) and water (discharge), as: 1426 (Fluxin -Fluxout)/ Fluxin, where the nitrogen fluxes have the units MT -1 and water L 3 T -1 (Figure 17). 1427 Consistent with the discussion in the preceding hydrology (section 3) and biogeochemistry (section 5) 1428 sections, there was net retention of water, NO3and NH4 + (i.e.: Fluxin > Fluxout) for the majority of 1429 monthly sampling intervals, with only 2 winter months displaying net release (i.e.: Fluxout > Fluxin). 1430 However, it is important to note that the correlation between net water and nutrient fluxes is partly 1431 spurious, since the same discharge values contribute to both axes, and is a common issue in water 1432 quality analysis. Nonetheless, variation about the 1:1 balance can be informative, since Fluxin -Fluxout 1433 is representative of the total change in storage of water or nutrients (named here ΔSQ or ΔSN 1434 respectively) at the time of sampling. Within this beaver modified system on the coastal plain of 1435 Alabama (USA), NO3fluxes were almost always retained to a greater extent than water, while water 1436 fluxes were generally retained to a greater extent than NH4 + fluxes, which had a much higher frequency 1437 of net release (Figure 17). This result is important because it emphasizes the first order control of water 1438 storage changes on the downstream water quality dynamics, which are likely critical to many other 1439 beaver impacted systems. In addition, it also demonstrates important second order effects, such as 1440 the far more efficient retention of NO3fluxes compared to NH4 + , even when both are operating overall 1441 as net sinks, due to their different reaction and production mechanisms (discussed in the nitrogen 1442 impacts section 5.3). These results are also similar to DeVito and Dillon (1993), who demonstrated the 1443 capacity of a beaver dam to retain nitrogen and phosphorus was controlled to the first order by the 1444 extent of water retention and runoff, with the added complexity of seasonal ice cover enhancing 1445 reducing conditions and therefore also the seasonal release of some fraction of NH4 + and PO4 3-. Higher 1446 frequency monitoring of discharge, carbon and nutrient fluxes is also important, and a recent study by 1447 Wegener et al. (2017) found net release of all these fluxes during high flows, and net retention during 1448 low flows in a beaver impacted river reach. In combination, these studies highlight the need for more 1449 studies accounting for the full mass balance of both water and nutrients, which involves higher 1450 frequency monitoring of changes in water and nutrients over a fixed reach or volume, and over 1451 identified flow paths, which can reveal far greater insights into the overall water quality dynamics 1452 beyond only characterizing system behavior as being either a net source or sink. 1453 In terms of the temporal variability in biogeochemical dynamics, only c. 40% of studies examined in 1454 Figure 16 reported 'sub annual' dynamics (e.g. variation at seasonal, monthly, or event timescales). 1455 From these studies that do examine sub-annual dynamics, it is clear that many of the export fluxes 1456 display considerable seasonal variation ( Over the longer term (i.e.: > 1 yr), it is clear that increased storage of water and nutrients (per unit 1471 length) should also increase their residence times. However, this increase in residence time must be 1472 mediated to some extent by the observed increases in outflowing fluxes such as DOC, N2, CO2, CH4, 1473 NH4, and in some cases PO4 3- (Figure 15). There is also likely to be large variability in the relative 1474 magnitude of residence times between these components, e.g.: carbon > phosphorus > nitrogen > 1475 water. Indeed, Naiman et al. (1988) estimated an order of magnitude increase in pond sediment 1476 carbon residence times as the storage increased. This may be especially important when considering 1477 the long-term resilience of beaver modified systems to climate and anthropogenic change, as well as 1478 how beavers can be used in river management, since water and nitrogen fluxes will likely be more 1479 sensitive to short term fluctuations than phosphorus and carbon, however these suggestions remain 1480 purely speculative. The long-term carbon feedbacks are discussed further in section 7.2. In natural 1481 wetland and lake systems, residence times, and therefore biogeochemical functioning, is linked to the 1482 degree of hydraulic connectivity between inflowing and outflowing water fluxes (Cohen et al., 2016). 1483 Although longitudinal (downstream) hydrological and biogeochemical connectivity is reduced in the 1484 short term by beaver dams (and thus increasing residence times), over seasonal and annual time scales 1485 the vast majority of water flow must still pass through and interact with the beaver impacted river 1486 reach. In contrast, many other wetland and lake systems in river networks usually interact with a much 1487 smaller fraction of total flows (Cohen et al., 2016). This is important when considering the potential for wetland, lake, or beaver modified systems to influence the evolution of downstream water quality 1489 and attenuate water quality problems such as high nitrate concentrations, since the overall 1490 effectiveness may be higher within beaver modified systems as they can provide increased water 1491 residence times whilst still interacting with the majority of water flow in the system. 1492 1980; Ward and Stanford, 1995). Broadly, the RCC states that lower order streams are dominantly 1506 heterotrophic, receive most of their organic matter as inputs from the terrestrial ecosystem, and have 1507 macroinvertebrate community compositions adapted to break down and filter these inputs. As stream 1508 order and size increases downstream, light availability increases which means more organic matter 1509 can be provided through aquatic primary production, and macroinvertebrate communities diversify to 1510 filter material from both benthic and water column environments. The RCC also places an emphasis 1511 on nutrient cycling and ecosystem stability, with the extent of biological activity and disturbance in low 1512 order streams having an influence on the net retention or export of nutrients to downstream and 1513 higher stream order ecosystems. 1514 Reach-scale beaver modifications to the physical process templates upon which ecosystems adapt and 1515 function therefore disrupt this traditional RCC framework, especially in low order stream habitats, with 1516 important consequences for our conceptualization of river ecosystem processes. The primary reason 1517 beaver modifications pose such a disruption to the RCC is because of the increasing extent of ponded 1518 surface water behind individual dams, and collectively within beaver dam complexes, which constitute 1519 an abrupt reach-scale shift from almost exclusively lotic (flowing water) to a complex mix of lentic (still 1520 water) and lotic conditions and transitions between them (Naiman et al. 1998). This variation between 1521 lotic and lentic ecosystems has been covered in conceptual models that include anthropogenic dams 1522 in regulated river systems (e.g.: the serial discontinuity concept of Ward and Stanford (1995) Specifically, beavers facilitate a mix of finer sediment and particulate organic matter benthic habitat in 1535 deeper water lentic environments (e.g. beaver pond and backwater channels), a replacement of lotic 1536 'riparian' zones with lentic 'littoral' zones, which are shallow water vegetated environments (e.g. 1537 beaver meadow and wetlands), and coarser sediment and particulate in shallow water lotic 1538 environments (e.g. immediately downstream of beaver dams) ( Figure 13). In addition, a rather unique 1539 feature of beaver impacts is the very large increase in large woody debris within aquatic habitats, 1540 especially within dams themselves but also elsewhere in the channel productivity including the beaver pond, littoral zone and wetland habitats, then there is likely to be a 1560 mix of autotrophic and heterotrophic ecosystem components, with increased productivity from beaver 1561 created wetlands and littoral zones contributing substantial new biomass, and through its breakdown 1562 an increased supply of coarse and fine particulate organic matter to the heterotrophic ponds and 1563 ecosystems downstream (Hodkinson, 1975;Naiman et al., 1986). It is this integrated mix of 1564 heterotrophic and autotrophic components in addition to the lentic and lotic transitions that makes 1565 beaver influenced ecosystems such a departure from the traditional RCC concept. This highlights the 1566 profound role of wetland vegetation and the littoral zone biomass production can have on NEPaq once 1567 lentic conditions are introduced, and by extension probably helps explain the widespread increase in 1568 net DOC export from beaver impacted systems (Figures 15, 16). This is also consistent with findings 1569 from other wetland and small lake ecosystems where productive littoral zones can be maintained 1570 (Wetzel 2001). The creation of lentic habitats can generate a larger abundance of particulate organic matter, plant 1587 tissue and nutrients within the ponded section, which increases the numbers of shredders and 1588 gatherer/collectors, which can otherwise usually only be found in low percentages within lotic reaches 1589 . Although the new lentic habitats created by beavers may have more restricted 1590 assemblages compared to the lotic habitats, it is the capacity of beavers to facilitate and maintain a 1591 mosaic of both habitats and the transitions between them that allows reach scale assemblage diversity 1592 to increase (Robinson et al., 2020). However, the influence of beaver ponds on benthic 1593 macroinvertebrates can be highly seasonal, which needs to be considered in studies targeting these 1594 differences . The larger diversity found in beaver influenced reaches may also be 1595 influenced by the increase in woody debris, with submerged wood adding considerable habitat 1596 diversity for macro-invertebrates in streams, which is known to increase macroinvertebrate numbers 1597 and species diversity (Benke and Wallace, 2003). Submerged large woody debris also creates pools on 1598 the channel bed, providing additional habitat for many invertebrate species (Benke andWallace, 2003) 1599 as well as the wood dam structures themselves becoming a potential hotspot for macroinvertebtrate 1600 habitat (Rolauffs et al., 2001). Hence, it is likely that beavers can increase not only the diversity of 1601 invertebrate species in the habituated stream section, but also potentially throughout entire stream 1602 reaches through the pervasive increase in large woody debris increasing the abundance of macro-1603 invertebrate taxa specialised in wood herbivory. However, these larger spatial scale effects of 1604 increased large woody debris on macro-invertebrate assemblages depend strongly on the local hydro-1605 geomorphologic conditions and requires further study in order better understand the influence of 1606 beaver impacts on macro-invertebrates in the aquatic food chain across a gradient of stream order 1607 sizes. Drift dispersal is also a critical component of many macro-invertebrate life cycles, and it can be 1608 expected that beaver dam construction might delay or filter this dispersal to some extent. However, 1609 in a comparative study Redin and Sjöberg (2013) surprisingly found no impact on drift density 1610 downstream of beaver dams. This may suggest beaver dam filtering of drift dispersal is not likely to be 1611 significant, although lags may still exist. Given this is a single study, further work is clearly also needed 1612 to understand drift dispersal responses across beaver impacted reaches in a wider variety of landscape 1613 contexts. 1614

1615
The potential impacts (positive or negative) of beaver dams on fish populations can be separated into 1616 migration, habitat, growth, population dynamics and diversity, and thermal regulation. It should not 1617 be controversial to state the following based on the process feedbacks already discussed in this review: 1618 1) constructing a beaver dam will restrict (but not necessarily stop) fish mobility, just as it does the 1619 transport of water and sediment, relative to the same river with no dam, 2) habitat diversity will 1620 increase, especially lentic habitat but also potentially in lotic zones through the general increase in 1621 large woody debris availability, and 3) river shading has the potential to decrease, and therefore locally 1622 increase water temperatures (see section 3.8), with flow regulation from dams potentially also 1623 stabilizing downstream temperatures. If these statements are largely without controversy, the fishy 1624 question therefore becomes, are these changes likely to have noticeable positive or negative impacts 1625 on fish populations? 1626 In terms of mobility impacts, there is an important dependence on the migratory needs of the species 1627 being considered, and thus whether the species is potamodromous (i.e. freshwater only), e.g. pike, or 1628 diadromous (i.e. migrating between salt and freshwater), e.g. salmonids. In addition, the timing and 1629 developmental stage during migration is critical, and especially whether higher mobility periods tend 1630 to occur during high or low flow regimes and whether they embark as juveniles or adults. As a result 1631 of these caveats, there is enormous variance in the research findings concerning fish mobility impacts. 1632 The cases with the largest negative impact on mobility have been found for juveniles migrating 1633 downstream (Mitchell and Cunjak, 2007;Schlosser, 1995;Virbickas et al., 2015), or on adult mobility 1634 during low flow periods (Bylak et al., 2014;Collen and Gibson, 2000;Cunjak and Therrien, 1998;1635 Mitchell andCunjak, 2007;Schlosser, 1995;Taylor et al., 2010). In one study over 4 summers, large 1636 fractions of total upstream and downstream fish movement over dams occurred over only a 1 -2 day 1637 period that had slightly elevated streamflow, though not all days with elevated streamflow had 1638 increased mobility (Schlosser, 1995). In some cases, the restricted mobility may even be seen as an 1639 ecological benefit, for example (Mitchell and Cunjak, 2007) found that beaver dams on coastal rivers 1640 prevented upstream migration of salmon, which through competitive exclusion increased fish species 1641 diversity upstream. These are however, far from ubiquitous results for all fish, with considerable 1642 variation between taxa (Schlosser, 1995), and many studies finding limited or negligible mobility pathways around dam structures may be important in mitigating dam impacts in some of these cases 1646 (Cutting et al., 2018). However, it is important to note that relatively few beaver impact studies have 1647 used fish tracking or tagging, and many instead rely on downstream vs upstream, or beaver site vs 1648 control site abundance, which is a far less reliable measure of actual mobility, and may in fact over-1649 estimate the mobility impacts of dams (Johnson-Bice et al., 2018). Thus, given this wide range of 1650 uncertainty, it is probably most apt to consider beaver dams as 'semi-permeable' barriers to fish 1651 movement (Schlosser, 1995 accounted for only ~2.5% of the river area but produced ~50% of the juvenile salmon in the river 1664 (Müller-Schwarze, 2011). The importance of succession in beaver dam habitat was also emphasized by 1665 Snodgrass and Meffe (1998), who also found species richness was highest in 'middle age' (9-17 yrs) 1666 abandoned dams and ponds, with species richness lower in both younger active dams, and older (>17 1667 yrs old) abandoned dams. Moreover, this result was only for headwater streams, with lowland sites 1668 exhibiting little difference in species richness with pond age. At more local scales, there is some 1669 concern that the coarse bed sediment habitat required for salmonids may be reduced by finer 1670 sediment deposition induced by beaver damming (see section 4), since if this is too extensive, it can 1671 result in some salmonid species being outcompeted by others (Müller-Schwarze, 2011). However, the 1672 finer sediment ponds may be advantageous for other fish species, for example in Sweden these finer 1673 beaver pond sediments have been found to be preferred habitat for minnow spawning (Hägglund and1674 Sjöberg, 1999). Over time, beaver ponds may also select for species more tolerant of oxygen stress 1675  given the tendency of ponds to have diminished dissolved oxygen, 1676 especially at depth (see section 5). Finally, beaver dam impacted rivers can also provide critical habitat 1677 refugia for fish during drought and summer low flow periods (Hägglund and Sjöberg, 1999;Hanson and 1678 Campbell Bylak, 2010). Beaver ponds also seem to be a net positive in terms of growth rates, particularly for 1683 salmonid juveniles (Sigourney et al., 2006). These increased sizes and growth rates are likely possible 1684 through a combination of reduced energy expenditure by the fish and greater food availability (e.g. 1685 macroinvertebrates) due to the higher overall ecosystem productivity (Pollock et al., 2003), and also 1686 perhaps due to the reduced mobility imposed by dams. However, some surveys also report no impact 1687 on growth rates (Malison and Halley, 2020). 1688 It is evident that water temperatures can rise both in beaver ponds and downstream, but this is far 1689 from ubiquitous and contains many nuanced dynamics (see section 3.8). The questions regarding water temperature and fish impacts are therefore 1) whether any temperature increase reaches the 1691 thermal tolerance thresholds for the species of interest, and 2) whether sufficient thermal refugia exist 1692 or are created through habitat modification that can mitigate against any stream sections that may 1693 now reach these thermal thresholds. Of particular concern here are cold water fish species, especially 1694 salmonids, which is particularly sensitive given their economic importance in many regions to fisheries 1695 and recreation. It is also likely that many cold-water species may already have a spatial range reflective 1696 of their thermal stress limits, and thus any temperature increase due to beaver impacts may at the 1697 very least lead to a constriction in the spatial distribution of these species. It is therefore not surprising 1698 that many studies do find a negative link between beaver impacts on increased water temperatures, 1699 and It is important to note that beavers and fish were presumably able to co-exist across a wide range of 1704 conditions prior to the large-scale declines in beaver populations across Europe and North America. 1705 However, modern river corridors cannot easily return to these conditions, with considerable human 1706 regulation of the landscape, and population dynamics of both beavers and fish that may be interacting 1707 outside their previous ranges, together means that the past may not be a terribly good guide to 1708 evaluating current impacts and potential management strategies. Modern stream habitats and their 1709 management ideals are also in many cases likely quite different from those during the beaver -fish 1710 co-existence of the distant past, meaning their re-unification may not easily revert to the desired 1711 harmonious balance of old. Many fish species of concern may also not be native, further complicating 1712 this dynamic. On the other hand, it may be the case that many of the documented impacts (positive 1713 or negative) on fish are too short term in focus. Provided sufficient time and space is available, as a 1714 river corridor begins to experience beaver dam and habitat succession, intact individual dams may 1715 collapse or promote channel avulsion, and the relatively closed habitat of intact single dams can 1716 become a mosaic of lentic and lotic habitats with sufficient migratory passages and thermal refugia. 1717 However, in many current river corridors, the luxury of the necessary time and space to achieve this 1718 successional mosaic may not be available. 1719 In practice, effective management of beaver impacts for the potential benefits for fish such as 1720 increased growth rates, and assemblage and habitat diversity, against the potential negatives such as 1721 temperature and mobility, may be difficult, especially as the balance between overall net positive or 1722 negative can shift over time (  (Figure 18). Dragonfly species 1738 have been shown to be 89% higher when compared to reaches not dammed by beavers (Schloemer, 1739 2014). In central Europe, amphibian species were observed to increase by 85 to 100% in beaver ponds 1740 compared to lotic reaches Dalbeck et al., 2007). In North America, beaver pond 1741 construction attracted much higher colonization rates of some, but not all, endangered amphibians 1742 (Hossack et al., 2015). The common frog (Rana temporaria) is known to benefit from the development 1743 of shallow beaver ponds, which creates large breeding areas (shallow ponds) during times of re-1744 production . Waterbird diversity and density is also much higher in beaver created 1745 wetlands (Grover and Baldassarre, 1995). These results indicate a close association between beaver 1746 impacts and many wetland-dependent species and hence their potential to facilitate the recovery of 1747 many of these fauna and flora, of which many of these species are critically endangered (Hossack et  1748 al Sweden (Hartman, 1996). In terms of initial impacts, when permanently inundated, most deciduous 1760 canopy trees will die within a year, and smaller sub-canopy species even earlier (Härkönen, 1999; Müller-Schwarze, 2011), but given more variable surface inundation or a slowly rising groundwater 1762 table from below, trees at the margins or at slightly higher elevations may die a slower death or even 1763 survive, albeit potentially under sub-optimal growing conditions and thus with stunted growth 1764 (Härkönen, 1999;Reddoch and Reddoch, 2005 (Figure 19), but with 1770 some capacity for both deciduous and evergreen tree survival at the margins. 1771 Trees within river corridors that survive or surround inundated areas are not breathing a sigh of relief, 1772 as they are also subject to browsing, girdling and felling by beavers. There are a large number of studies 1773 documenting tree preference on the basis of species, size, and foraging distance (Haarberg and Rosell, 2006;Jenkins, 1980;Martell et al., 2006). However, there is no clear definitive list of these preferences, 1775 given that studies vary considerably in species and size availability, as well as in the timescale of beaver 1776 impact on the riparian vegetation being studied. It is generally accepted however, that all these 1777 preferences are constrained by 1) optimal foraging theory, in which the beaver seeks to maximize net 1778 energy intake during foraging from a central location per unit time (Belovsky, 1984;Fryxell andDoucet, 1779 1993;Jenkins, 1980;McGinley and Whitham, 1985), and 2) by the need to overcome plant chemical 1780 defenses (secondary metabolites) through generalist herbivore foraging strategies 1781Basey et al., 1990. The impact of these constraints can be seen across many studies that find e.g. 1782 browsing intensity (Haarberg and Rosell, 2006;Jenkins, 1980;Martell et al., 2006;McGinley and1783 Whitham, 1985), as well as tree size and species preferences (Basey and Jenkins, 1995;Fryxell and1784 Doucet, 1993;Haarberg and Rosell, 2006;Jenkins, 1980;Raffel et al., 2009) of beavers clearly shifting 1785 with increasing distance from water. Consistent with optimal foraging theory, this is likely because the 1786 foraging time costs increase with distance from a central water location compared to the energy 1787 gained (Belovsky, 1984), and also because tree species and their size vary considerably in terms of 1788 energy availability and secondary metabolites . However, the choices available to 1789 beavers are not everywhere the same, thus beavers cannot always be religious in tree selection and 1790 local species availability will be a strong constraint on preference. Nonetheless, it is possible to infer 1791 the broad upper and lower bounds of woody species preferences, with willow (genus Salix), aspen (or 1792 poplar, or cottonwood -genus Populus) and birch (genus Betula) species clearly preferred when 1793 available, mixed results for alder (genus Alnus), oak (genus Quercus) is less preferred, and there is a 1794 clear avoidance of conifer species, though even these will be consumed under duress to varying extents within these preference ranges as part of the generalist herbivore strategy, subject 1797 to the caveats already mentioned above. There is also a considerable seasonal cycle to woody 1798 vegetation consumption, which dominates beaver diets over winter (Svendsen, 1980) and especially 1799 in ice covered regions within submerged food cache's that are progressively compiled underwater in 1800 ponds for overwintering (Hartman and Axelsson, 2004). Apart from dietary intake, it has been noted 1801 that less palatable species will often be felled for use in dam construction (Pinkowski, 1983). However, 1802 this is not likely to be a consistent result, since beavers are only targeting the inner bark, leaves, and 1803 twigs of woody plants for consumption, thus depending on the tree sizes available there can be a 1804 considerable volume of wood left over from many species across the palatability spectrum for use in 1805 dam construction. 1806 The combined impact on riparian trees is therefore likely a local decrease in diversity (Nolet et al., 1807(Nolet et al., 1994, that may also come to be dominated by quickly regenerating tree species able to grow as shrubs, 1808 as well as those that are less palatable to beavers (Barnes and Mallik, 2001;Naiman et al., 1988;Pastor 1809Pastor et al., 1988. Importantly, this also results in a distinct shift in both the age and size demographics of 1810 the riparian forest towards younger and smaller trees, albeit with a strong dependence on distance 1811 from water. This substantial impact on riparian forest cover is in flagrant disregard of many current 1812 forestry and conservation management practices (Martell et al., 2006), though it is unclear whether 1813 any fines or other penalties have been issued. Thus, if retaining forested riparian areas in combination 1814 with beaver occupation is a desired management outcome, as it may be in many areas of the world, 1815 managers would be wise to consider a composition dominated by species less palatable to the beaver, 1816 or even potentially using the leaves of less palatable species as protection (Basey, 1999). 1817 Although tree species diversity may decrease locally, this is usually not the case at the landscape scale 1818 if forested areas away from the riparian and inundation zones remain. Indeed, beaver impacts are 1819 generally considered to increase overall vegetation species richness at the landscape scale by creating 1820 a new mosaic of terrestrial and aquatic vegetation habitats (Wright et  pioneer species in beaver meadows tend to invest more in biomass production than chemical defenses 1856 during regrowth (Veraart et al., 2006), but they may also be flexible in their chemical defense 1857 investments in juvenile sprouts in response to beaver cutting ). This likely create a 1858 complicated mix of poorly understood negative and positive feedbacks that may allow some 1859 vegetation species to maintain a dynamic equilibrium with beavers (Pollock et al. 1995), and others to 1860 decline, all of which remains poorly understood. However, it is important to note there is a strong bias 1861 towards higher latitudes in terms of our understanding of herbivory restriction and resource depletion, 1862 and many more studies from lower latitudes as beaver ranges expand are needed. 1863 The net result of reduced herbivory is to force beaver migration or population decline, which in 1864 principle allows later successional species to return to the meadow, with the nature of this succession 1865 depending primarily on the ongoing flooding frequency and water retention capacity of the site 1866 ( This is the first of three sections that discuss the emergent issues synthesized from the findings of this 1901 review. Thus far, this review has summarized the key changes and processes dynamics stemming from 1902 the impact of beaver damming of river corridors on hydrology, geomorphology, biogeochemistry, and 1903 ecosystems (table 1). Whilst many important connections between these fields have already been 1904 described, it is useful to examine how all these impacts are connected in a more comprehensive way. 1905 7.1 Initial and shorter-term impacts: the importance of floodplain inundation and Disturbance by beaver activity has a cascading series of consequences for river corridors that begins 1908 with their primary impacts, namely the damming of river channels, digging riverbank and floodplain 1909 burrows and channels, and actively gnawing woody vegetation on riparian and floodplain areas (yellow 1910 circles Figure 20). Tree felling provides material for dam construction, and dam construction can result 1911 in profound increases to water storage and hydrology (blue circles), sediment storage and river 1912 corridor geomorphology (brown circles), nutrient cycling and storage (red circles), and terrestrial (light 1913 green circles) and aquatic ecosystems (aqua circles). Our perceptual model of the links between all 1914 these feedbacks is not intended to be definitive, but it does highlight that floodplain inundation 1915 emerges as a central initial driver of many subsequent feedback connections (Figure 20). 1916 Floodplain inundation is a hydrological feedback caused by backwater ponding behind dams that 1917 reaches above the level of the adjacent floodplain, which can also extend downstream of the dam as 1918 shallow overland flow or as new wetlands. Thus, in terms of hydrology, beaver damming decreases 1919 longitudinal hydrological connectivity, but can increase lateral and vertical (e.g. hyporheic) 1920 connectivity. The scale of these feedbacks depends on the capacity of river systems to convert the rise 1921 in surface water behind dams to an increase in the areal extent of water. This geomorphic context 1922 dependency is discussed in greater detail in sections 4 and 10. The extent of floodplain inundation is 1923 important because it can: (1) increase aquatic habitat area and diversity, which in turn expands the 1924 interface between terrestrial and aquatic trophic chains and increases net aquatic ecosystem 1925 productivity (section 6, Figure 18), (2) increase surface and groundwater water storages, and may in 1926 some cases be linked to increased flood retention capacity and to locally enhanced baseflow (see 1927 section 3, Figures 4, 6, 8). In terms of biogeochemical processes, floodplain inundation allows (3) an 1928 expansion of anaerobic conditions, via diminished oxygen transport and increased organic matter 1929 storage and production. This allows a larger diversity of biogeochemical pathways and fluxes to 1930 emerge, which in combination with enhanced vertical (hyporheic) exchange can diminish NO3export 1931 (via increased denitrification and biomass uptake) and enhance DOC export (see section 5, Figure 15 (section 4, Figure 12, 13). This change in depositional environment, in combination with tree loss and 1937 vegetation shifts due to (5) higher soil water content, increased flood disturbance, and herbivory 1938 (Figure 19), as well as beavers digging new floodplain channels, and the substantial increase in large 1939 woody debris within the river, may in turn encourage (6) river corridor planform shifts to anabranching, 1940 multi-thread flow patterns, and an increase in floodplain carbon storage (Sutfin et al., 2016;Wohl, 1941 2013). In summary, the cascading impacts stemming from beaver damming, in which hydrological 1942 feedbacks through the extent of floodplain inundation can be a key moderating factor, has the 1943 potential to create a distinct environmental functioning of the entire river corridor in which the 1944 hydrology, geomorphology, biogeochemistry, terrestrial and aquatic ecosystems, and the multiple 1945 feedbacks between them have to adjust to new steady-state conditions ( Figure 21). 1946 7.2 Longer-term impacts: Perpetual succession of landscapes and ecosystems, and 1947 feedbacks driving carbon sequestration potential 1948 As beaver occupation of a river corridor extends in timescale, especially > 10 1 years, the initial 1949 landscape impacts that follow on from the hydrological changes described above will remain 1950 important, but will also be modified as the river corridor adjusts towards a state of 'perpetual 1951 succession'. In this context, 'succession' is meant in a holistic sense and refers to landscape and 1952 ecosystem processes changes that take longer timescales to manifest ( Figure 21). Thus, we suggest the 1953 critical impact of beavers on river landscapes is to amplify the natural mechanisms of adjustment that 1954 operate over these longer timescales, which they do by (1)  assemblages driven by water availability and herbivory (section 6.5). These impacts are 'perpetual' 1965 only so long as the disturbance from beaver activity can be maintained, which may include cycles of 1966 abandonment and re-occupation. Therefore, following abandonment the state of perpetual succession 1967 may be largely reversible ), or they may trend towards alternate states, discussed 1968 in detail in section 8. The net effect of perpetual succession through beaver impacts is to create, as 1969 described by Naiman et al. (1988), a 'spatial and temporal mosaic' of environmental conditions and 1970 habitat complexity along the river corridor, that cannot develop without prolonged beaver activity. 1971 The fate of the increased carbon storage facilitated by beaver impacted river corridors (see section 5), 1972 and alluded to in point (4) above, is the subject of considerable interest and speculation. In particular, 1973 the question is how much, carbon will remain in storage over longer timescales (e.g. > 10 2 -10 3 yrs), 1974 and how much of the shorter-term carbon storage is likely to be exported downstream. In terms of the 1975 aquatic component of this system, Naiman et al. (1988) reported order of magnitude increases in 1976 organic matter residence (or turnover) times in beaver ponds up to ~161 years. Such a large increase 1977 in residence times are to be expected in beaver ponds where the relative increase in carbon storage is 1978 very large, however it is of course unlikely that individual beaver ponds and the carbon stored within 1979 them will remain intact for this length of time, given many dams can be abandoned or breached over 1980 the 1 -10 1 yr timescale. Thus, the actual long-term fate of the aquatic carbon storage in beaver systems 1981 is likely to be set by the frequency of dam disruption on the one hand, and the geomorphic capacity of 1982 the river system to sequester any remaining pond deposits within a water saturated alluvial 1983 stratigraphy on the other (e.g. via overbank deposition whilst keeping water tables relatively high). As 1984 a result of these constraints, it is likely that only a small fraction of the available aquatic carbon storage 1985 will be sequestered over the long-term. In terms of riparian zone soil carbon, the 'reverse succession' 1986 process promoting pioneer vegetation on beaver meadows enables higher biomass input rates to the 1987 soil (Rosell et al., 2005), resulting in higher soil carbon accumulation in beaver meadows (Westbrook 1988(Westbrook et al., 2011Wohl, 2013). However, similar to the challenges in preserving aquatic carbon over the 1989 long-term, this increase in soil carbon may difficult to retain unless the high biomass inputs from the 1990 meadow and higher water tables can be also maintained by continuous beaver occupation, or 1991 alternatively sequestered within water saturated alluvial deposits. Given beavers do not occupy sites 1992 indefinitely, beaver meadow soil carbon stocks can diminish over time once abandoned (DeAnna and  1993 Wohl, 2019), likely though a combination of reduced biomass inputs and declining water tables. The 1994 overall long-term carbon storage potential in beaver impacted river corridors therefore seems to be 1995 most sensitive to 1) whether or not continuous beaver activity (or at least cycles of re-occupation) can 1996 be maintained, and 2) the geomorphic and hydrologic capacity of the corridor to stratigraphically 1997 sequester the carbon deposits. These constraints offer some explanation as to why the long-term 1998 storage rates of carbon in beaver systems are far lower that the shorter-term rates (Wohl et al., ). 1999 It is also clear that in the case of site abandonment, the pathways of subsequent landscape and 2000 ecosystem transitions will determine the fate of the beaver assisted carbon storage. These potential 2001 pathways are covered in the following section (section 8). 2002 interpretation. For example, as an agent of disturbance, beavers must maintain this disturbance in 2064 order for beaver meadows to develop and remain. Does the meadow therefore constitute a stable 2065 state? As documented in Figure 22, and in the vegetation section (section 6.5), even following beaver 2066 abandonment, meadows may persist for considerable periods of time, but this depends on a range of 2067 initial conditions and it is clear they will inevitably undergo some landscape and ecosystem transitions. 2068 Therefore, without continued beaver activity, meadows are clearly not themselves stable systems if 2069 sufficiently long time periods are considered. However, the alternate stable state framework is very 2070 useful in highlighting the necessary role of beavers as an ecosystem engineer in enabling these 2071 landscape and ecosystem transitions that would likely not occur in their absence. For example, the 2072 trajectory of channel incision and floodplain drying following beaver abandonment in Figure 22 would 2073 be difficult to reverse without beaver re-introduction facilitating the recovery of incised channels, as 2074 was the case at Yellowstone once elk browsing pressures were reduced (Wolf et al., 2007). However, 2075 we note that the attribution of river incision solely to beaver abandonment at this site is problematic, 2076 and that a more complex interplay with climatic (Persico and Meyer, 2013) and fire (Meyer et al., 1992(Meyer et al., ) 2077 is likely involved and is also important context to consider for all beaver assisted river recovery efforts. This review has synthesized the profound impacts that beavers can have on river corridor hydrology, 2086 geomorphology, biogeochemistry and ecosystems, and the myriad of feedbacks between them. Yet, 2087 the interpretation of these impacts in terms of what is 'natural', in terms of the future role of beavers 2088 in river management and rehabilitation, and in terms of public perception and government policy are 2089 fraught with uncertainty and a large potential for misunderstanding. Are beavers an invasive pest to 2090 be removed, a natural part of landscape functioning whose impacts should be embraced, or 2091 somewhere in between as an ecosystem engineer that itself requires some level of management? 2092 Here, we briefly review the challenge of defining 'natural' landscapes, and spectrum of positions and 2093 contexts in which beaver impacts and their implications have been considered. 2094 There is comprehensive evidence for the widespread historic reduction in both the geographic range 2095 and population densities of both North American and European beavers, although the timing of this 2096 impact is much earlier in Europe than in North America (Morgan, 1868;Müller-Schwarze, 2011;Zahner 2097Zahner et al., 2005. However, estimates of these historic population densities and ranges throughout the river 2098 networks of both continents prior to human impact remains uncertain, with relatively unbounded 2099 speculations in North America ranging from 60 -400 million . This limits the 2100 context in which the current recovery in beaver populations in both North America and Europe can be 2101 placed, and will always render interpretations of 'natural' population densities and ranges, or the 2102 carrying capacity of the landscape, with some level of uncertainty. Hence, the full range of habitats 2103 that beavers can occupy remains unclear, particularly in marginal environments such as ephemeral 2104 streams with little riparian vegetation, low order streams at increasing elevation, Eurasian steppe 2105 landscapes, and streams heavily modified by humans (Bailey et al., 2019). This knowledge gap has led 2106 in some cases to the re-introduction of beavers into unsuitable habitats, and therefore delays in re-2107 introduction success (Stocker, 1985). Despite these overall limitations, it is useful to try and constrain 2108 the potential range of beaver habitat at more regional and local scales. Recent work on streams of the 2109 south-west USA used information on the permanence of water sources, available riparian vegetation, 2110 channel width, magnitude and frequency of typical floods, and channel gradient and mean discharge 2111 as predictors for the potential beaver habitat within these hydrological sensitive river networks 2112 (Macfarlane et al., 2017). More research is clearly needed to constrain potential and preferred beaver 2113 habitat ranges. 2114 However, the overall landscape carrying capacity of beavers is more complex than potential habitat, 2115 and considered from a population point of view, there are two broad constraints on beaver 2116 populations: 1) predators (e.g. wolves, where present) as a top down control , and 2117 2) food supply as a bottom up control, which includes interaction with other herbivores (see section 2118 8). However, it is not intuitive how these constraints should operate in the very common case of beaver 2119 populations that are either re-introduced or recovering. Interesting data in this case comes from 2120 beaver populations re-introduced to Sweden between 1922 and 1939, which long term monitoring 2121 reveals has followed the Riney-Caughley 'irruptive' population model for introduced ungulates, 2122 whereby they experienced a growth phase for 24 -35 years, followed by a steady population decline 2123 to a more stable (though still dynamic) level (Hartman, 1994;Hartman and Axelsson, 2004). Such a 2124 population dynamic suggests 1) that there is a general lack of top down predator control, and 2) that 2125 beavers as an expanding population may exploit food supply beyond the landscape carrying capacity 2126 and therefore decline in numbers. However, it is also important to note that this population trend is 2127 from the boreal zone and may not be as predictive of expected population expansions throughout 2128 more temperate regions. In addition, except for some regions of the USA, Canada, Poland, Latvia and 2129 Russia, beavers across many regions of the Northern Hemisphere are not expected to encounter 2130 significant top-down predation pressures (e.g. from Wolves) in the regions in which they are recovering 2131 or being reintroduced . In a separate line of evidence, river geomorphic conditions 2132 have been found to be more influential than forest type in habitat selection as beavers colonize new 2133 areas (Hartman 1996), and a general finding across Europe has emerged in which beavers first increase 2134 in habitat range before increasing in population (Halley and Rosell, 2002). This suggests the growth 2135 phase is a case of being spoilt for choice (not that vegetation availability is unimportant), with habitat 2136 selection becoming more marginal as the landscape approaches carrying capacity (Pinto et al. 2009), 2137 suggesting the eventual population decline may be due to a delayed feedback regarding food supply 2138 and the ecosystem engineering impacts of beavers discussed in detail in this review, as well as the 2139 need to eventually move into increasingly marginal habitats. Where competition with other herbivores 2140 such as elk are present, the population outcome may be much more dynamic and beaver populations 2141 may instead suffer heavy declines as the food resources are even more quickly depleted, and with 2142 fewer chances for recovery (Wohl, 2019, also see sections 6.5, 8). This longer-term relation between 2143 ecosystem engineering, food stocks, and landscape carrying capacity remains very poorly understood, 2144 and urgently needs further research. However, it is important to note than an irruptive population 2145 dynamic may not always occur, outside countries with large forested areas such as Sweden, beaver 2146 population expansion may have far greater habitat competition and conflict with human land use 2147 (Halley and Rosell, 2002). Nonetheless, as warned by Hartman (1994), it would be prudent for 2148 managers and policy makers to be cognisant of the potential beaver population consequences of 2149 having no natural predators or habitat competition given the risk of over-exploitation of food resources 2150 during population recovery and reintroduction efforts. Regardless of the uncertainty surrounding the 2151 'natural' landscape beaver carrying capacity and projected population dynamics across European and 2152 North American landscapes, any future capacity is still likely to be higher than the present population 2153 numbers in many regions. If we consider the trajectory from current population numbers to the 2154 theoretical landscape carrying capacity as a legitimate future scenario, then, as documented 2155 throughout this review, this will set in motion a large suite of landscape and ecosystem feedbacks and 2156 changes to the river corridor that will require thoughtful and potentially vexing management and 2157 policy decisions into the foreseeable future. In some cases, an expansion of beaver populations to the 2158 landscape carrying capacity may be welcome, and beavers could potentially re-establish river 2159 conditions to those present prior to European impact (Polvi and Wohl, 2013). However, in many 2160 regions it is unlikely that beaver populations reaching the theoretical landscape carrying capacity is a 2161 desired outcome as envisaged under a majority of river and landscape management scenarios, which 2162 by design must balance the needs of multiple stakeholders. Thus, the active human management of 2163 beaver population numbers and their impacts is all but certain to increase into the future as their 2164 populations expand, and this management is already well underway in some regions (BAFU, 2016; 2165 Halley and Rosell, 2002; Wróbel and Krysztofiak-Kaniewska, 2020). 2166 2167 9.2 Insufficient context can skew the interpretation of beaver impacts 2168 As this review has attempted to reveal, beaver modifications to river corridors set in motion a wide 2169 range of feedbacks between hydrology, geomorphology, biogeochemistry, and ecosystems. In 2170 addition, as beaver populations expand, the extent to which their impacts are considered positive or 2171 negative by various stakeholders also depends on management priorities, which themselves will be 2172 heavily dependent on the magnitude of change that beavers are expected to deliver within human 2173 modified or natural landscapes. In terms of placing the magnitude of beaver impacts in an 2174 experimental context (e.g. before-after-control-impact, BACI), the practice is relatively rare, but more 2175 beaver impact studies are embracing this kind of approach Conner et al., 2016;2176 Weber et al., 2017), which will be increasingly important for engaging with stakeholders on outcomes. 2177 In any case, given the wide range of feedbacks that can occur, it can be difficult to interpret these 2178 impacts if insufficient information or understanding of the underlying feedbacks are available. 2179 Therefore, a narrow process understanding of these impacts risks interpretations that can be skewed 2180 as either net positive or negative from a management or policy point of view. This means care is 2181 needed when isolating individual impacts, lest they be used to strengthen the perception of beaver 2182 impacts being either net positive or negative for the landscape in question. This lack of context is 2183 further amplified by the relative paucity of process studies that provide actual data on these feedbacks. 2184 Based on our review of the underlying processes (hydrology, geomorphology, biogeochemistry, and 2185 ecosystems) (Table 1), a set of illustrative, but not exhaustive, examples in which impacts considered 2186 in isolation could be construed net positive or net negative is provided in Table 4. Whist it is certainly 2187 interesting from a management or policy perspective to highlight positive impacts, which are often 2188 considered 'ecosystem services', it would be remiss to exclude the potential negative impacts linked 2189 to the same process or feedback. Likewise, only pointing to net negative impacts can ignore the many 2190 potential benefits that beaver impacts may provide. This highlights Floodplain forests in particular have proven to be highly favored habitats, especially since they include 2213 abundant Nothofagus pumilio and Nothofagus betuloides which have become the preferred woody 2214 species browsed by beavers in the region (Anderson et al., 2006b). However, beavers have also been 2215 able to spread into the steppe vegetation landscapes which implies the importance of woody 2216 vegetation in habitat selection is lower than generally expected (Pietrek and González-Roglich, 2015). 2217 The net result is population numbers in Patagonia have grown to an estimated ~100,000 individuals 2218 (Choi, 2008). 2219 In terms of impacts, beaver damming is flooding sub-Antarctic riparian forests and reducing canopy 2220 extent (Choi, 2008a). Vegetation succession in beaver ponds also follows a different trajectory 2221 compared to other disturbances common to the region such as forest clearings or wind-throw, and 2222 facilitate succession dominated by Nothofagus antarctica, which is the local pioneer species most 2223 adapted to high water content conditions (Martínez Pastur et al., 2006). The creation of beaver ponds 2224 and meadows has also been shown to advantage invasive bush and grass species (Anderson et al., 2225 2009), and invasive mammals such as muskrats and minks which hunt native fauna (Crego et al., 2016). 2226 Interestingly, thus far there does not appear to be a significant difference between macro-invertebrate 2227 assemblages in the natural lentic habitats and those created by beavers in Patagonia (Anderson et al., 2228 2014), suggesting the native lentic aquatic fauna have been able to expand their range. In any case, 2229 these findings are consistent with the broader ecological argument that introduced species can 2230 facilitate the expansion of additional introduced species (Anderson et al., 2009), and provides an 2231 important example of where it is possible to conclude that there are net negative ecological feedbacks 2232 associated with beaver impacts. 2233 It is also worth noting that in Finland and areas of northwestern Russia, the beaver is also an introduced 2234 species to itself. Seven North American beavers (C. canadensis) were introduced in 1937 as part of 2235 ongoing efforts to re-introduce the nearly extinct Eurasian beaver (C. fiber), which at the time were 2236 thought to be identical species (Parker et al., 2012). This is of considerable concern, since as noted by 2237 Parker There is therefore a very real chance that the invasive C. canadensis is able to displace C. fiber over the 2242 longer term and further expand into mainland Europe, thus strident eradication measures have been 2243 recommended (Parker et al., 2012), however it is unclear if any have yet been adopted. 2244

Beavers as ecosystem engineers and their role in river restoration and
The global river restoration effort is a sizeable collective business, and in many cases is does not 2247 consider whether a site is within the historical range of beavers, or the implications for restoration 2248 strategy if they returned (Burchsted et al., 2010). There has been an interest in re-introducing beavers 2249 into formerly native habitats in Europe and North America since at least the 1950s, mainly for the 2250 biodiversity benefits (see section 6) (Stocker, 1985;Zahner et al., 2005). Since the 1990s beavers have 2251 also been increasingly recognized and described favourably as ecosystem engineers (Gurnell, 1998;2252Jones et al., 1996Wright et al., 2002). In addition, the fact that beavers benefit from the ecosystem 2253 changes that they trigger (e.g. the pond as protection from predators, enhanced foraging habitat), and 2254 the large positive feedbacks they generate with the rest of the aquatic and terrestrial ecosystem, 2255 means they are now often labelled as a 'keystone species' (Mills et al., 1993). This designation as both 2256 a keystone species and ecosystem engineer mean beavers have become highly rated as a tool for river 2257 rehabilitation improved ecosystem biodiversity (Pollock et al., 2017), which is supported by the wide 2258 range of net positive impacts effect beavers can have (tables 1, 4) impacts (see section 7), and acknowledges the considerable legacy of beaver ecosystem engineering 2275 on river corridors prior to their widespread eradication. Combining beavers and the geomorphic basis 2276 of stage 0 restoration efforts is particularly well suited to address the broader problem of historical 2277 channel incision, as the multithread channel system can reduce reach scale stream power and promote 2278 deposition (Pollock et al., 2014). In combination, these processes can lead to the lateral hydrological 2279 re-connection of the floodplain-channel system (Polvi and Wohl, 2013)  to understand success in attracting beaver populations to take over as the 'stage 0' engineer, otherwise 2290 the continued maintenance of BDA efforts, and the broader feedbacks deriving from the 'perpetual 2291 succession' induced by beaver disturbance (see section 7.2), could be difficult to reach. The core goal 2292 behind the rewilding framework is the re-establishment of trophic ecosystem complexity (Bakker and  2293 Svenning, 2018), particularly top-down interactions promoted by larger wildlife species or their proxies 2294 (Svenning et al., 2016). Thus, beaver re-introduction is essentially a form of rewilding, and parts of this 2295 review have documented the trophic complexity they facilitate, particularly in aquatic and wetland 2296 meadow ecosystems (see sections 6, 7). In addition, as an ecosystem engineer beavers may 2297 substantially improve the biodiversity restoration success many rewilding projects seek to achieve and 2298 reduce the need for management interventions ( financial values for these services may be premature for widespread management and policy use. 2312 Nonetheless, as the knowledge and evidence base increases, the utility of this approach is certain to 2313 increase. In terms of distilling the place of beavers across all these restoration frameworks, it is clear 2314 from the knowledge collected in this review that there is a need to consider the profound spatial and 2315 temporal variation in the feedbacks created by beaver impacts both between and within river 2316 corridors, in all aspects of project planning and implementation. This variation is driven in large part, 2317 but not exclusively, by the context dependency of the site being considered, which is synthesized in 2318 more detail below (section 10). 2319 10 Putting beaver impacts in a holistic context 2320 Here we develop a holistic context for evaluating beaver impacts based on an inter-disciplinary 2321 synthesis stemming from the main findings of this review. This is centered on a conceptual model 2322 ( Figure 24) that emphasizes these impacts cannot be divorced from the wider landscape context in 2323 which they occur. We first consider the spatial components of connectivity (lateral vs longitudinal 2324 connectivity), and then show how in combination with climate, these gradients can impact important 2325 process timescales (e.g. water and nutrient transport). Broadly, we consider valley slope and width as 2326 placing an important first order constraint on where and how beaver damming will influence a river 2327 corridor, which is demonstrated using four river valley scenarios ( Figure 24). 2328 The extent of beaver impacts on lateral connectivity will control, amongst other things, open water 2329 extents, flood attenuation capacity, sediment, carbon and nutrient storage, extent of anaerobic 2330 metabolism and biogeochemical interfaces, water residence times and nutrient fluxes, aquatic 2331 ecosystem productivity and biodiversity, riparian vegetation mosaics, and river channel pattern. Thus, 2332 the ability of beaver dams to influence the lateral hydrological connectivity between the channel and 2333 floodplain is a key impact from which many other hydrological, geomorphic, biogeochemical, and 2334 ecosystem impacts follow. 2335 Valley slope and width will moderate the number of dams that can be built in a given reach, and thus 2336 determine the overall capacity for beavers to decrease longitudinal connectivity, but increase vertical 2337 exchanges, over a stretch of river corridor. This is because increasing the slope allows a higher density 2338 of dams per unit stream length, or a beaver dam cascade, and at lower slopes wider multi-channel 2339 systems also potentially allow a high density of dams to develop laterally across its network. Dam 2340 density defines the extent of disruption to longitudinal connectivity, as well as influencing water, 2341 sediment, carbon and nutrient storages, vertical hydraulic gradients controlling ground and surface 2342 water interaction and hyporheic exchange, hydraulic roughness, the size and number of lentic to lotic 2343 aquatic ecosystem transitions, fish migration, the extent of wood introduction to the river corridor, 2344 and the spatial constraints on meadow development. 2345 In our framework, river corridors that are highly incised or contain negligible floodplain area represent 2346 systems in which there is little capacity for increases in the width of open water area, meaning beaver 2347 impacts on lateral connectivity will be comparatively low (Figure 24 A1 -A2). However, these typically 2348 low-order and higher slope river systems represent cases where although changes to lateral 2349 connectivity may be low, the changes to longitudinal connectivity and vertical exchanges may be very 2350 high, especially relative to the conditions prior to beaver impact. The damming of low order river 2351 systems by beavers can create significant jumps in longitudinal hydraulic gradients, with sections of 2352 flatter water surfaces, ponds and wetlands, connected by short but abrupt increases in the hydraulic 2353 gradient (i.e. the dams themselves). This may greatly enhance longitudinal processes such as hyporheic 2354 exchange, and also create a mosaic of lentic ecosystem conditions and transitions within river corridors 2355 that would be highly unlikely to support them in the absence of beavers. 2356 As greater floodplain and channel space becomes available with increasing stream order and 2357 decreasing slope, the lateral connectivity associated with individual dams has the potential to increase 2358 (Figure 24 B -C). In many river corridors of the world, river-floodplain connectivity has been heavily 2359 reduced or lost due to incision and engineering modifications, leading to large losses in aquatic and 2360 terrestrial habitat and biodiversity (Schumm, 2005;Wohl, 2004;Wohl, 2005; Wohl and Beckman, 2361 2014). These streams are likely to experience the greatest increases in lateral connectivity, open water 2362 extent, and habitat complexity through beaver damming activity, often resulting in distinctive beaver 2363 meadow development through the 'reverse' succession of vegetation assemblages. 2364 The relative impact of beavers on river-floodplain connectivity will be lower when this lateral 2365 connectivity is already naturally high, such as in near-natural river systems in Patagonia with a high 2366 abundance of lakes and wetlands (Anderson et al., 2006a), in natural fen and peat ecosystems (Naiman 2367 or in larger braided or anabranching rivers , where beavers mostly 2368 dam smaller tributaries or secondary channels and therefore a much small proportion of the overall 2369 flow is impacted by beaver damming (Figure 24 D). However, even in these cases, at a local scale the 2370 influence of beaver dams on the riparian processes and ecosystems can still be significant. 2371 The climatic context will also exert considerable influence on the spatial and temporal scale of beaver 2372 impacts through its control on the supply of, and atmospheric demand for, water. If we hold the 2373 general valley geometry to be constant, then varying the climate context within each scenario in Figure  2374 24 (A -D) will lead to differential beaver impacts on the river corridor. For example, being able to 2375 increase the extent of open surface water and higher soil moisture through the construction of beaver 2376 dams will have increasingly large hydrological and ecosystem consequences as the surrounding 2377 climatic context moves to drier scenarios. This is because in very dry climates the proportion of water 2378 lost to evaporation from open water may increase, but concurrent water storage increases may allow 2379 increases to streamflow persistence downstream, and the creation of new lentic habitat and 2380 ecosystem refugia that would not otherwise exist. Thus, river corridors with temporary flow dynamics, 2381 either because they are low order systems (e.g.: steeper headwater channels), or because they are 2382 very dry, should experience very large relative changes to connectivity and residence times 2383 (hydrological and biogeochemical). In very cold climates, deeper beaver ponds with surficial ice cover 2384 may also provide new and important aquatic habitat refugia. 2385 The final context to consider is temporal. As agents of shifting connectivity, ecosystem disturbance and 2386 succession, and increased gradients, process feedbacks associated with beaver damming will evolve 2387 over time within each of the spatial contexts described above. How long beavers can maintain their 2388 activity at a site depends on both top down (e.g. humans, predators, competitors) and bottom up (e.g. 2389 food resource) constraints, and will determine the persistence of water, carbon, nutrient, and 2390 ecosystem changes they have induced. Importantly, the population constraints, length of beaver 2391 occupation, and whether cycles of abandonment and re-occupation can be established, will all help 2392 determine how river corridor landscapes and ecosystems develop once beaver occupation ceases. 2393 The legacy of beaver damming impacts for river corridor processes and ecosystems further 2394 downstream remains poorly understood and is critical to improve given the importance of river 2395 networks in the global water, carbon, and nutrient cycles. The ubiquitous increase in wood and 2396 particulate organic carbon to rivers following beaver damming (Anderson et al., 2009;Thompson et al. 2397 2016) is an example in which beaver impacts can generate a significant downstream legacy for 2398 ecosystems, carbon cycling, sediment transport, and channel evolution (Levine and Meyer, 2019). 2399 Changes to water storage also have the potential to leave a downstream legacy on streamflow regimes 2400 and water resources. In addition, changes to riparian ecosystem structures and trophic complexity 2401 through the introduction of new lentic-lotic transitions and 'reverse' succession meadows will 2402 challenge traditional concepts of how these ecosystems should vary downstream along rivers. can increase lateral and vertical, and decrease longitudinal hydrologic connectivity. This change in 2411 hydrological connectivity is the basis for all subsequent impacts, with the key process impacts 2412 summarized in Table 1. Longitudinal decreases in connectivity create ponds and wetlands, transitions 2413 between lentic to lotic ecosystems, increase vertical hydraulic exchange gradients, and biogeochemical 2414 cycling per unit stream length. Increased lateral connectivity will determine the extent of open water 2415 area and wetland and littoral zone habitats and induce 'reverse' succession in riparian vegetation 2416 assemblages. In combination, these changes in connectivity also promote increased storages of surface 2417 and subsurface water, carbon, nutrients, and sediment, and increase habitat complexity and 2418 biodiversity at the reach scale. The extent of these impacts depends on 1) the hydro-geomorphic 2419 landscape context, with the extent of floodplain inundation being a key driver of changes to hydrologic, 2420 geomorphic, biogeochemical, and ecosystem dynamics, and 2) the length of time beavers can sustain 2421 this disturbance at a given site. This large influence of beavers on river corridor processes and 2422 feedbacks is also fundamentally distinct from what would occur in their absence, and thus has 2423 profound implications for the future function and management of river systems as beaver populations 2424 continue to recover and expand. Nonetheless, considerable knowledge gaps and outstanding 2425 questions remain, which provides a rich and interdisciplinary future research agenda. 2426 Rise in river water levels due to beaver dam construction in a low-order stream in Germany (A), 2505 resulting in a rise in the shallow groundwater level in two distal piezometers (B) (modified from Zahner, 2506Zahner, 1997. Rise in water levels are apparent after the dashed vertical lines, which represents the timing of 2507 beaver dam construction. (C) Measured geometry of an idealized groundwater 'wedge' developed due 2508 to a rise in the groundwater table upstream and adjacent to a beaver dam in the Bridge River, Oregon 2509 (USA). Note the spatial dimensions in this figure are not drawn to scale. Modified from Lowry (1993). 2510 end of the beaver pond; c) deposition and erosion in beaver ponds upstream of beaver dams during a 2524 variety of flow types: during normal flow (i); re-mobilisation of beaver pond sediments during high-2525 flow events and sediment deposition on floodplains respectively beaver meadows (ii); inset floodplain 2526 of former beaver pond deposits remain after drainage (iii); and d) variability of spatio-temporal 2527 pattern of in-channel beaver ponds (i -iii) results in a delay in overall sediment transport downstream. 2528 Flow direction is indicated by thick black arrows. 2529 Backwater ponds introduce lentic, littoral and wetland (characterized by unconfined surface flow, 2573 beaver meadows) habitat for invertebrates, amphibians, and fish in otherwise faster flowing rivers and 2574 dry floodplains. By permanently flooding some of the floodplain, beavers connect aquatic and 2575 terrestrial ecotones, and create breeding and feeding ground for many animals. 2576 Cause and effect feedback loops that can be generated following beaver dam construction, digging, 2584 and gnawing (large yellow circles) in a connected river-floodplain system (Hydrology (blue), 2585 Geomorphology (brown), freshwater ecosystems (turquoise), and Biogeochemistry (red)). A link to 2586 Animal Ecology (purple) is also provided as an example case, but is not meant to be definitive. The 2587 figure indicates that conceptually, the cause of most beaver induced environmental changes in the 2588 aquatic and riparian ecosystem is caused by beaver dams being able to inundate the floodplain and 2589 pond the main channel. 2590

Figure 21 2591
Summary of shorter-term and longer-term processes and feedbacks in beaver meadows, with a visual 2592 example from the Jossa River in Germany. Within ~3 months of damming, a large shallow wetland 2593 covered a large portion of the formerly agricultural floodplain (left aerial photo). After ~20 years, the 2594 floodplain has developed into a mix of ponds, wetlands, channels, and a mosaic of organic matter rich 2595 fen, sedge, reed, and juvenile willow vegetation patches (right photo, a drone-derived orthophoto and 2596 digital elevation model, giving a spatial impression). The arrow points towards the confluence between 2597 the two Jossa channels. 2598 2599 Alternate stable states 2600

Figure 22 2601
Potential alternate riparian trajectories of river corridors depending on weather beaver occupation 2602 can be sustained. If the site is abandoned, e.g. due to resource depletion or competitive exclusion (a), 2603 the subsequent trajectory depends on the valley hydro-geomorphic, and specifically whether channel 2604 stability and high water contents can be maintained, or whether incision and drying ensues (b). 2605 Numbers refer to example references for alternate stable states: 1  This connectivity is initially hydrological, which then in turn influences geomorphic, biogeochemical 2612 and ecosystem connectivity. The horizontal transitions (A -D) represent shifts in river valley (and to 2613 some extent climatic) contexts. These represent a transition in overall valley slope, along with an 2614 increase in the size of the main channel and extent of the valley and floodplain area. The transition 2615 from landscape context B to C represents an increase in the size of the main channel such that beavers 2616 are likely to be able to dam the main channel (A -B) below this size, and unlikely to be able to dam the 2617 main channel (C -D) above this size. An important feature of the landscape (and climatic) transitions 2618 is the increase in lateral connectivity from A -D, with the relative extent of this lateral connectivity 2619 enhanced by beaver damming (1 -2), especially as valley slope decreases. 2620 The vertical transitions (1 -2) represent the change in each landscape context from pre-(1) to post-2621 (2) beaver damming. An important consequence of the pre-to post-beaver damming transition across 2622 all landscape contexts is the decrease in longitudinal connectivity. Some key consequences of this are 2623 an increase in vertical hydrological exchange gradients, increases in the storage and residence times 2624 of water (H2O) carbon, nutrients (N and P) and sediment, and an increase in the biogeochemical cycling 2625 within the river reach (per unit length). In addition, each dam introduces new ponded water, and as 2626 the number of dams increases, so too does the number of transitions between lentic and lotic 2627 freshwater ecosystem habitats. With increasing river size and natural lateral connectivity (A-B), the 2628 potential influence of beaver dams on the lateral connectivity become smaller (1 -2   See the relevant sections for more detailed discussions on these feedbacks 2 In many regions of Western Europe river valleys have been actively managed as agricultural landscapes, in some cases since the Neolithic period, and in most regions since the medieval period. The policies to maintain and protect these cultivated river valleys often describes them as cultural landscapes