Patchiness in flow refugia use by macroinvertebrates following an artificial flood pulse

Flow refugia, locations that maintain substrate stability and low hydraulic stress during periods of high flow, can ensure riverine resilience in the face of increasing hydrological unpredictability. Despite their known importance, they have been overlooked in recent years with work on drought refugia currently seeing greater attention. Moreover, research on the role of flow refugia during artificial flood pulses in regulated rivers, where flood disturbances are no longer part of the hydrograph, is essentially absent. Here, we compared flow refugia for benthic macroinvertebrates among six habitats (main channel, side channel, riffle, margin, lentic including a floodplain pond, and inundated floodplain) within four different sites in response to an artificial flood pulse. We found that the grain‐size distribution and macroinvertebrate community composition changed at each site following the flood. Macroinvertebrate assemblages became longitudinally homogeneous, but within‐site beta diversity and taxa richness remained temporally stable following the flood pulse, suggesting the presence of flow refugia. In this respect, margin, inundated floodplain and lentic (a floodplain pond) habitats provided important flow refugia locations, particularly for the mobile mayfly Rhithrogena sp. In contrast, low substrate stability in riffle and side channels resulted in limited refugia potential for most taxa. Refuge use was however patchy with high levels of intra‐habitat variability being evident for Rhithrogena sp. and the amphipod Gammarus fossarum in margin and side channel habitats. Further work is required to advance our knowledge of flow refugia in rivers with differing flow regimes to enable their integration into management and restoration schemes.

intense disturbances, macroinvertebrate communities are highly resistant, often returning to pre-flood densities in a few weeks or months (Angradi, 1997;Matthaei, Uehlinger, & Frutiger, 1997;Robinson, Uehlinger, & Monaghan, 2003). An important factor that can mitigate the impact of flood disturbance on benthic assemblages is the use of refugia which enables recovery times to be shorter than the generation times of most invertebrate species (Sedell, Reeves, Hauer, Stanford, & Hawkins, 1990;Van Looy et al., 2019).
Anthropogenic activities continue to modify riverine ecosystems globally, diminishing habitat diversity and the potential availability and quality of refugia habitats (McCluney et al., 2014;Wohl, 2019). It is therefore urgent that the role of refugia is better quantified for different river types that cover a range of flow regimes, especially as high flows are being increasingly implemented in river restoration schemes.
Further, this urgency is particularly relevant given that climatic change is leading to more unpredictable hydrological conditions with both flood and drought events predicted to increase in frequency and intensity (Asadieh & Krakauer, 2017;Yuan, Jiao, Yang, & Lei, 2018).
Identifying habitats which could act as flow refugia during flood disturbances is therefore vital to ensure the persistence of freshwater biodiversity globally. This knowledge would enable freshwater refugia habitats to be conserved and incorporated in management and restoration strategies, something that is currently absent (Hermoso, Ward, & Kennard, 2013;Keppel et al., 2015;Selwood & Zimmer, 2020). As working on the implications of flood events is inhibited by the relative unpredictability of flood events and the logistics of working during high flows (Death, 2008), artificial flow pulses can also provide valuable scientific opportunities for testing ecological theories (Konrad et al., 2011;Olden et al., 2014).
In this study, we therefore sought to assess how substrate stability affected the provision of instream habitats that provide flow refugia from an artificial flood pulse in the regulated river Spöl, Switzerland. The objectives were to: (a) assess the role of flow refugia in maintaining the core taxa present in the Spöl despite potential reductions in abundances following the artificial flood; (b) assess the implications of the artificial flood pulse on substrate stability and the provision of in channel refugia habitats; and (c) assess the effectiveness of different habitats to act as flow refugia during an artificial flood pulse.

| Study location
The river Spöl is located in the central Alps, flowing into Switzerland from Italy in the lower Engadine ( Figure 1). The river is regulated by two hydroelectric dams before entering the Inn River at Zernez, Switzerland. Flow regulation in the form of residual flow (a minimum discharge that is set) commences downstream of Livigno reservoir (Punt dal Gall dam) where the Spöl flows $5.7 km through a canyonconfined valley in the Swiss National Park and into the lower Ova Spin reservoir. From this reservoir, the Spöl flows a further 5.5 km to its confluence with the Inn. For more details regarding the hydropower setup on the river Spöl please see Scheurer and Molinari (2003).
Prior to regulation in 1970, the river Spöl exhibited a natural snowmelt / glacial meltwater flow regime, with high flows in summer and low flows in winter. Periodic floods from heavy rainfall occurred during summer / early autumn with peak discharges between 20 and 60 m 3 /s . The average annual flow of the Spöl at the Punt dal Gall fluctuated between 12.5 and 6.6 m 3 /s, but post-completion of the dam the residual flow was set to an average flow of 1 m 3 /s. In the lower flow regulated section from Ova Spin to Zernez, the flow is even lower, being permanently set to 1 m 3 /s in summer and 0.3 m 3 /s in winter (Scheurer & Molinari, 2003).
Land use within the 295 km 2 catchment (Bundesamt für Umwelt, 2020) is predominately coniferous forest (Picea excelsa and Pinus mugo) with some grassland and sedges present in the lower floodplain.
Climate in the region is continental with high seasonal variations in temperature, but low precipitation (Barry, 1992). River sediments originate primarily from dolomitic and calcareous scree from rocky, high gradient valley slopes and locally from remnant glacial moraines (Trachsel, 2001;Trümpy, Schmid, Conti, & Froitzheim, 1997). Bedrock is present in many areas of the riverbed. In many places, forest vegetation has colonised gravel bars, and roots of woody vegetation have stabilised banks decreasing riparian dynamics often present in natural free-flowing rivers (Mürle, Ortlepp, & Zahner, 2003).

| Artificial flood, study sites and habitats
The most notable feature of the flow regime of the regulated Spöl was the absence of peak flow events. As a result, the Engadine power company, Swiss National Park and state authorities began to implement artificial floods, predominately in the upper regulated part of the Spöl in 2000. Thirteen artificial floods have also been undertaken in the lower flow regulated section -2017Kevic, Ortlepp, Mürle, & Robinson, 2018). In September 2018 (the focus of this study), a controlled artificial flood was released in the lower flow regulated section, from the outlet of the Ova Spin reservoir over an 8-hr period. A peak discharge of 25 m 3 /s was achieved during the flood that lasted around 2 hrs, with rising and falling limbs being incremental. Although the event was shorter than natural flood events, previous artificial floods of similar discharge have been shown to be sufficient enough to mobilise bed sediments and reduce algal levels without causing high fish mortality Uehlinger, Kawecka, & Robinson, 2003). It should be noted that there were no implications for water temperature during the studied flood as the dam is a hypolimnetic release reservoir, and thus the thermal regime remains relatively constant (Jakob, Robinson, & Uehlinger, 2003).  Table S1 for a summary of all flow velocities for site-specific habitats before and after the flood and Table S2 for a breakdown of the sampled habitats on each occasion. As refugial habitats were sampled when present, some habitats were sampled only at one site (e.g., floodplain and riffle) and some were sampled on one or two occasions for some sites.

| Grain-size distributions
The GSD of benthic substrates changed following the flood pulse, being habitat and site dependent. At the site level, changes in the GSD were most evident at sites 1 and 2, with gravels becoming coarser immediately following the artificial flood, indicating the loss of fine sediment surficial deposits (Table 1). At sites 3 and 4, changes in GSD were not as substantial but still present, with some increases in the finemedium pebble fractions (4-16 mm) at both sites, resulting in a finer distribution (Table 1) (Table 1).
At the habitat level, there were no differences in GSD profiles over time at the main channel habitats (Figure 2a). In contrast, considerable changes in the GSD were evident at side channel and riffle habitats with coarsening evident following the flood (Figure 2b,c). Riffle habitats displayed fining 10 months later with an increase in the fraction <30 mm, whilst no further change was present in the side channel habitats.  (Table S3). Despite alterations to Spöl assemblages following the flood, beta diversity T A B L E 1 Summary of grain size metrics for the study period based on Wolman pebble count. Post-flood is immediately after the artificial flood      Table S4).

| DISCUSSION
We examined how benthic macroinvertebrates responded to an artifi- homogenisation of Spöl assemblages associated with the flood, within-site beta diversity remained temporally stable in all but one instance. The stable temporal beta diversity at the site level suggests that refugia provision was present in the Spöl with taxa being able to persist through the flood, but that the more tolerant generalists dominated. The exception to this was at site 1 that supported the highest taxa richness pre-flood, and which comprised a core set of taxa but also some rarer species. It is likely that the flood eradicated some of these rarer taxa with some remaining in a few samples, thereby causing the observed increase in community heterogeneity post-flood.
The provision of refugia also can be assumed through the lack of significant differences in overall Spöl taxa richness pre-and post-flood.
Taxa richness can provide a good proxy to assess the presence of refugia during periods of disturbance (following Sueyoshi et al., 2014;Van Looy et al., 2019). Given the stability in this metric, we can assume that refugia were present in the Spöl and therefore that refugia provision was essential in enabling the resistance of benthic taxa to the artificial flood (also see Robinson, Aebischer, et al., 2004).
Following the flood, lateral margin areas and the inundated floodplain displayed increased abundances and the lentic habitat (comprising a previously disconnected floodplain pond) displayed increased taxa richness, suggesting either active flow refuge seeking behaviour to avoid unfavourable conditions in the river channel, or passive draft/transport during the flood followed by persistence in the refugial habitats. Robinson, Aebischer et al. (2004) and  also recorded high invertebrate drift within inundated riparian areas following floods and shoreline / marginal areas have been similarly cited as refugial areas (Rempel, Richardson, & Healey, 1999). Margin habitats exhibited a shift in composition to one similar to lentic habitats in our study, reflecting the refugial role both these habitats played for taxa preferring slow-flowing waters. Sueyoshi et al. (2014) similarly found that habitats with slow-flowing water acted as refugia during a snowmelt flood. In contrast, the instream habitats of riffle, side and main channel habitats became more homogenous in their community composition post-flood.
Reductions in abundance and taxa richness were also evident in the riffle and side channel habitats, most likely reflecting a considerable loss of taxa. Inundated floodplain communities represented a 'bridge' between the two clusters of instream (riffle, side and main) and margin / lentic habitats, suggesting some degree of intermediate habitat conditions. The active use of riparian and floodplain habitats as refugia highlights the importance of river systems being able to maintain lateral connectivity during hydrological flood disturbances (Chanut, Datry, Gabbud, & Robinson, 2019;Ward, 1989). Globally, many river channels are increasingly being channelised, resulting in flood peaks being funnelled within the channel with little connection with lateral riparian / floodplain areas that naturally would occur. This lack of lateral connectivity in highly modified river channels limits habitat provision and in turn the likelihood of taxa finding suitable refugia (Negishi et al., 2002;Sueyoshi et al., 2014;Williams et al., 2020). Our results provide further evidence that lateral connectivity is essential to maintain biodiversity following hydrological disturbances.
Flow refugia use was particularly evident for the mayfly Rhithrogena sp. with occupation in the floodplain habitat, and increased abundances in margin and side channel habitats following the flood. This result was mirrored by a reduction in abundances in their dominant habitat of the main channel prior to the flood. Refugia use has been found to be highly species dependent (Lancaster & Hildrew, 1993b;Sueyoshi et al., 2014), however Rempel et al. (1999) also cited Rhithrogena sp. as showing active refugia behaviour following a large flood. This taxon is highly mobile, being observed as early recolonists following instream disturbances (Matthaei, Uehlinger, Meyer, & Frutiger, 1996) and therefore can seek refugia when hydraulic stress increases during flood events.
Mobility in accessing refugia is also crucial to ensure that taxa are not stranded in temporary habitats formed during a flood, such as inundated floodplains or marginal areas . In the case of the studied flood pulse, morphological change in the channel form dictated that marginal and inundated floodplain habitats were still present 7 days after the flood. Therefore, it can be expected that taxa would have been able to make it back to their residential habitats with low risk of stranding following the flood.
In contrast to Rhithrogena sp., the mayfly Baetis sp. and the dipteran Chironomidae displayed significant reductions in abundance following the flood. Bruno, Cashman, Maiolini, Biffi, and Zolezzi (2016) found that these two taxa were particularly sensitive to dislodgement from high flows, demonstrating the highest drift rates during hydropeaking flood pulses. Although both taxa are particularly affected during artificial flood pulses, they have been recorded to exhibit rapid recovery to pre-disturbance abundances Robinson, Uehlinger, & Monaghan, 2004).
Use of refugia is often highly patchy in space, reflecting the microdistribution of hydraulic stress and substrate stability amongst other factors (Dole-Olivier et al., 1997;Lancaster & Belyea, 1997;Palmer et al., 1996). This patchiness was evident in the Spöl with  (Death, 2008). Artificial floods in this sense represent a unique scientific opportunity to further our knowledge base of flow refugia (Konrad et al., 2011;Olden et al., 2014).
The importance of refugia habitats in the Spöl can be placed in the context of changing sedimentological conditions. High hydraulic stress during flood events can lead to significant displacement of benthic macroinvertebrates, whilst stable riverbed patches can act as flow refugia (Effenberger et al., 2006;Lancaster & Hildrew, 1993a). We observed that all sites underwent changes in their GSD, which may partially explain the altered composition of macroinvertebrate communities following the flood. In particular, riffle and side channel habitats underwent considerable coarsening with the artificial flood mobilising fine sediments. Other studies examining the role of artificial floods in the river Spöl under earlier flood settings also observed a coarsening of riverbed sediments, with active transport of bed material and the deposition of loose gravel . We observed similar in this study. The high degree of sediment transport most likely explains why these habitats displayed the greatest changes in macroinvertebrate communities following the flood in terms of composition, abundance and richness, with large-scale drift often being initiated via dislodgement due to saltating grains or enhanced shear stresses (Gibbins, Batalla, & Vericat, 2010;Gibbins, Vericat, & Batalla, 2007). In marked contrast to the side and riffle habitats, the main channel, although also being subjected to the highest hydraulic pressures of the flood pulse, displayed no changes in the GSD. This habitat was subjected to high energy velocities prior to the flood pulse and it can be expected that the artificial flood would not change the structure of the riverbed. Here, taxa richness and abundances were low prior to the flood but remained temporally stable as did community composition to some degree. Understanding substrate stability is therefore vital to be able to assess the relative importance of refugia for maintaining resilience.
Our knowledge of flow refugia remains severely limited compared to that of drought / low flow refugia. Our results provide additional knowledge to this neglected resilience concept in the context of an artificial flood pulse. We found that although macroinvertebrate assemblages became homogenised longitudinally at the four river locations monitored, the diversity of invertebrate communities at the site scale remained similar to pre-flood levels and taxa richness remained stable over time. Overall, these results indicated that although generalist taxa dominated the community following the flood, sufficient refugia must have been present to enable the persistence of more flow-sensitive taxa that contribute to diversity. Riparian margin areas, inundated floodplains and lentic habitats (a floodplain pond) acted as important refugia areas. Low substrate stability in riffle and side channels resulted in limited refugia potential. However, our results also highlight that refugia use is patchy in space, with significant intra-habitat variability being evident. We believe this study and the recognition of refugia as a means of river resilience by Van Looy et al. (2019) should stimulate further research on flow refugia functioning, which although prominent in the late 1990's has since been essentially neglected. Ensuring refugia functioning under increasingly unpredictable hydrological extremes by incorporation into restoration and management schemes is vital to ensure the persistence of freshwater biodiversity and management (Selwood & Zimmer, 2020).