DISCUSSION
We found longitudinal trends in geomorphic and ecological parameters to be very weak in these naturally disconnected study catchments, which contains disparate channel types that segment the stream. Our study also finds the geomorphic difference between the catchments to be visible in the riparian vegetation along the different catchments.
Downstream hydraulic geometry relationships of width- drainage area power functions explain at most 57% of variability in width for slow-flowing reaches in Hjuksån, and a minimum of only 4% of variability in width for slow-flowing reaches in Bjurbäcken. Combining all reach types, only 8% of variability in width is explained in Bjurbäcken and 43% of variability is explained in Hjuksån. According to Faustini et al. (2009), all of these regressions classify as ‘poor’ fits, except for slow-flowing reaches in Hjuksån; and the power relationships in Bjurbäcken have poorer fits than for any region in the conterminous United States (Faustini et al., 2009). In the Hjuksån catchment, located below the FHC, which contains more fine deltaic sediment in addition to coarse glacial sediment, there are slightly stronger downstream hydraulic geometry relationships for all reach types, indicating potential for self-adjustment in response to the current flow regime as seen in alluvial channels (Singh et al., 2003; Wohl, 2004). The fact that rapids and the catchment above the FHC show poorer DHG relationships, fits with Wohl’s (2004) observations from coarse mountain streams, where low ratios of the stream power to D84 lead to poor DHG fits. Exponent (β-) values, which describes the spatial rate at which width increases with increased drainage area, typically range 0.2-0.4 (Faustini et al., 2009). Combined reaches and slow-flowing reaches in Bjurbäcken fall within that range (0.34 and 0.26, respectively) in addition to rapids in Hjuksån (0.39); however, the remaining reach types have fairly large β-values (ranging 0.56- 0.75) (Figure 3), indicating a fairly rapid downstream increase in channel width, potentially caused by a y-value >1 (Equation (2)), due to the more dendritic drainage basin in Hjuksån compared to a more linear drainage basin in Bjurbäcken (Figure 2) and greater connections to groundwater (Burgers et al., 2014).
The presence of lakes in the Hjuksån catchment, containing finer sediment and thus having more alluvial characteristics, create wider channels directly downstream of the lake than if the reach was further downstream of a lake (Table 1). Lakes are serving to reset the longitudinal sediment conveyor belt by trapping fine sediment, increasing width: depth ratios directly downstream of lakes (Arp et al., 2007). In the Sawtooth Mountains region of Idaho, US, channel shape recovered by 50% within 1.0-1.8 km downstream of lakes and required up to 10-20 km to recover by 90%. Given the high spatial density of lakes in catchments in northern Sweden, channels may never recover and thus never reach a non-lake influenced equilibrium form. In addition to the significant effect of lakes, several surficial geology types, reflecting erosivity of streambanks affect the drainage area- channel width relationship, where the coarse till has a narrowing effect and fine deltaic, and subglacial sediment has a widening effect (Table 1). Peat has a widening effect on rapids but a narrowing effect on slow-flowing reaches, likely because the intrinsic cohesivity of clay in peat will decrease lateral erosion but compared to coarse till found in rapids the presence of peat will allow greater erosion. Likewise, bedrock has a narrowing effect in rapids and the slow-flowing reaches in Hjuksån but a widening effect in slow-flowing reaches in Bjurbäcken; bedrock outcrops found in slow-flowing reaches are smoothly rounded features composing the bed or small parts of the banks, whereas bedrock in rapids will lead to high slopes that naturally form narrower channels or make up streambanks that will hinder any lateral erosion. In general, the low R2 values and generally low β-values reflect the low ability for semi-alluvial channels to adjust their channels to the contemporary flow regime (Polvi, 2021).
The positive relationships between increased distance between reaches and reach similarity indicates low connectivity between reaches in terms of hydrochoric seed or propagule dispersal. However, models like linear regressions are sensitive to outliers and therefore these models are significant but with a very weak R2. We therefore interpret these results a lack of connectivity since reaches closer to each other are less similar than reaches that are further apart. The geomorphic trend of weak longitudinal relationships with lakes is also mirrored in the ecological data by a slight trend in decrease in longitudinal species density of riparian vegetation in Hjuksån, and no significant relationships in Bjurbäcken.
Poorly developed downstream hydraulic geometry relationships with regards to channel width, combined with low support for connectivity between riparian vegetation communities in semi-alluvial stream-lake systems, indicate that these are highly fragmented catchments where local factors steer geomorphic form and biotic communities. All rivers are part of a landscape context with varying degrees of connectivity between segments and reaches. Previous work has shown that dispersal is strong in well-connected areas compared to isolated headwaters, and patterns of connectivity among sites in a network can affect population dynamics (Swan & Brown, 2017). Hence, biodiversity has often been shown to increase towards the middle of the catchment (Nilsson et al., 1989; Kuglerová et al., 2015). However, stream networks in northern Fennoscandia are commonly naturally fragmented and consist of three types of process domains (rapids, slow-flowing reaches and lakes), which all differ in morphology and hydraulics that influence their capacity to facilitate the plant dispersal (Su et al., 2019a, b). In a connected stream network, one would expect a downstream increase in species richness density (Andersson et al., 2000; Kuglerová et al., 2015). However, in our naturally fragmented river system we found the opposite pattern, with a reduction in species density with increasing distance downstream. These differences in patterns can partly be explained by the high spatial resolution and that our study was conducted continuously along two streams, in comparison to previous studies with lower spatial resolution and discontinuity in sampling (e.g. Nilsson et al., 1994; Andersson et al., 2000; Kuglerová et al., 2015). The increased spatial resolution and sampling continuity allowed us to find an overall reduction in species density, as a function of the lack of hydrochory due to the presence of numerous lakes (Su et al., 2019a, b). Lakes trap ~80% of seeds during the spring flood and only allow downstream transport if the lake outlet is aligned with the wind direction (Sarneel et al., 2014; Su et al., 2019a). In a connected stream network, we would also predict higher similarity between reaches close to each other; however, in our disconnected systems we did not find such a relationship. Su et al. (2019b) showed that these three process domains have differing plant species communities, and thus seed banks may therefore provide a local source for metacommunity control, or perhaps diversity depends mostly on local site conditions (Green et al., n.d.), and not connectivity.
Rather than longitudinal controls of increasing discharge (as a function of drainage area) and inter-reach connectivity on channel width and riparian vegetation communities, our results indicate that our study catchments are driven by local-scale geomorphic and ecological controls. These local-scale controls, such as surficial geology (presence of coarse glacial sediment) and local seed banks and source populations, are therefore more important in recovery of stream processes and communities than connectivity of flows, sediment, and propagules between reaches in responding to disturbances, which include habitat restoration. Therefore, passive ecological recovery, with recolonization by hydrochory, is not as likely in disconnected stream networks, and thus manual planting of riparian vegetation may be necessary. Similarly, geomorphic recovery through channel adjustment based on the current flow regime is unlikely in disconnected networks with semi-alluvial process domain segments; thus, physical channel manipulation should play a larger role in stream restoration. In addition, interactions between channel width and riparian vegetation communities throughout the catchment may serve to further shape reach morphology and riparian zones (Anderson et al., 2004).
With weak downstream hydraulic geometry relationships, traditional stream ecology concepts, which assume increasing width as drainage area increases, may not apply. Geomorphic relationships with channel size have also been used to model nutrient uptake in river networks (Ensign & Doyle, 2006). Thus, given the abundance of lakes in northern latitudes (Messager et al., 2016), this study can have widespread implications on the understanding of fluvial processes and stream ecosystems in boreal, and (sub)arctic regions. For example, we may not be able to rely on past cornerstones of river science, such as downstream hydraulic geometry and the river continuum concept, in certain catchments in understanding river dynamics, designing stream restoration projects, and predicting responses of communities after restoration or other natural disturbances. Furthermore, disconnected fluvial systems with abundant lakes may also be analogous to pre-anthropogenic stream networks that contained abundant log jams and beaver dams, which create various degrees of natural disconnectivity (Wohl & Beckman, 2014; Green et al., n.d.). These naturally occurring discontinuities should be distinguished from anthropogenic dams that are nearly complete barriers to propagule, sediment and water fluxes. Natural fluvial disconnectivity features buffer fluxes and allow transport in temporal pulses and through leaky barriers; thus, lakes can provide a model for how beaver dams and log jams, which are temporally and spatially heterogeneous, affect functional connectivity of processes affecting geomorphic form and ecological communities.