Riparian Vegetation Distribution along the Ume River: Predicted responses of riparian plants to environmental flow modifications in run-of-river impoundments

Riparian Vegetation Distribution

along the Ume River

Predicted responses of riparian plants to environmental

flow modifications in run-of-river impoundments

Louise Berglund

Degree Thesis in Ecology 30 ECTS Master’s Level

Report passed: 04 June 2014 Supervisor: Roland Jansson

Riparian Vegetation Distribution along the Ume River

Predicted responses of riparian plants to environmental flow

modifications in run-of-river impoundments.

Louise Berglund

Abstract

River environments are complex and dynamic ecosystems, and provide valuable ecosystem services such as clean water. The species rich riparian vegetation performs many important ecosystem functions such as reducing erosion and filter inputs from upland areas. Regulated flow regimes have decreased riparian plant species richness, cover and plant performance. To restore the integrity of riparian ecosystems, mitigation measures such as re-regulation of water-level regimes toward more natural seasonal fluctuations may be needed. The aim of this study was to assess potential responses of riparian plants to changes in water-level regulation in run-of-river impoundments to better match natural flow regimes. The elevational extent of plant species on riverbanks of two run-of-river impoundments in the Ume River were surveyed and their probability of occurrence along the gradient of inundation duration was modelled and compared to their distribution in the free-flowing Vindel River. Most species showed similar tolerance to flooding in the Ume and Vindel Rivers. Changes in elevational extent in response to three simulated environmental flow regimes were predicted by using the relationship between plant occurrence and inundation duration. A simulated spring flood and low water levels during the latter part of the growing season is predicted to result in the largest increase in elevational extent, with increases of 70-80% for several riparian species. However, only 47% of the riverbanks along run-of-river impoundments in the Ume River is deemed to be suitable for plant establishment, since many riverbanks are steep and devoid of fine-grained substrate as a result of erosion.

Keywords: riparian plant species, environmental flows, water level regulation, flood duration, run-of-river impoundments

Sammanfattning

Älvmiljöer utgör komplexa och dynamiska ekosystem som tillhandahåller värdefulla ekosystemtjänster så som rent vatten. Den artrika strandvegetation bidrar till många viktiga ekosystemsfunktioner som närings- och giftupptag och till minskad erosion. Vattenregleringen med förändrade flödesregimer har minskat artrikedom, täckningsgrad och tillväxt av strandväxter. För restaurering av strandekosystemen kan omreglering till mer naturliga säsongsvariationer i vattenståndet vara nödvändigt. Den här studien syftade till att förutsäga hur utbredningen av strandväxter längs stränder i vattenkraftsmagasin potentiellt skulle förändras vid användande av miljöanpassade flöden för att mer likna naturliga flödesregimer i outbyggda älvar. Jag undersökte växternas utbredning i höjdled på stranden längs två magasin i Umeälven och beräknade sannolikheten för varje arts förekomst längs strandens översvämningsgradient. Av de arter som förekom i både Umeälven och den närliggande, outbyggda Vindelälven jämfördes växternas utbredningsgränser i respektive älv. De flesta arterna uppvisade liknande översvämningstolerans i Umeälven och Vindelälven. För att förutsäga förändringar i utbredning som respons på tre olika simulerade miljöanpassade vattenståndsregimer, jämfördes arternas översvämningstolerans vid nuvarade vattenstånd med simulerade vattenståndsregimer. En simulerad vårflod och lågt vattenstånd under sensommaren förväntas ge de största responserna i artutbredning med ökningar på 70-80% för ett flertal strandväxter. Endast 47% av älvstränderna i magasinen i Umeälven bedöms vara lämpliga för växtetablering eftersom stora delar av strandsträckorna är branta och saknar finkornigt substrat till följd av erosion.

Table of Contents

1 Introduction ... 1

1.1 Riparian ecosystems ... 1

1.2 Hydropower regulation and conservation management goals... 3

1.3 Environmental flows ... 3

1.4 Aim ... 4

2 Materials and methods... 5

2.1 Study areas ... 5

2.2 Survey ... 7

2.3 Water-level data ... 7

2.4 Environmental flow modifications ... 9

2.5 Data analysis ... 10

3 Results ... 12

3.1 Vegetation cover and species richness ... 12

3.2 Species distribution along the elevational gradient ... 14

3.3 Comparison of inundation tolerance between the Ume and the Vindel Rivers . 14

3.4 Predicted changes in extent of riparian plant species as a result of changes in

water-level regime ...16

4 Discussion... 20

4.1 Comparison of species distribution in riparian zones between run-of-river

impoundments and a free-flowing river. ... 20

4.2 Predicted plant responses to simulated water-level regimes ... 21

4.3 Conclusions ... 24

Acknowledgements... 25

References ... 26

Appendix

I Species list for Bjurfors nedre and Harrsele

II Number of days inundated per growing season and elevational limits

1

1 Introduction

1.1 Riparian ecosystems

Riparian ecosystems constitute the transition zone between terrestrial and aquatic systems and are among the most species rich and dynamic environments existing (Naiman et al. 1993; Naiman and Décamps 1997; Nilsson and Svedmark 2002; Ward et al. 2002). According to Naiman and Décamps (1997), the riparian zone consists of the area “between the low and high water marks and that portion of the terrestrial landscape from the high water mark toward the uplands where vegetation may be influenced by elevated water tables or flooding and by the ability of the soils to hold water”. Riparian ecosystems sustain organisms with several life-history strategies and disturbance adaptations, and are sites of biogeochemical processes important at the landscape level (Naiman and Décamps 1997). They often host a diverse shrub and tree vegetation, which serves as habitat for mammals and birds, and woody debris that provides suitable conditions for both terrestrial and aquatic invertebrates (Naiman and Décamps 1997). Rivers and streams are linear corridors of similar habitat across landscapes, transporting organic and inorganic material and enabling plant dispersal among riparian vegetation communities along watercourses (Nilsson and Svedmark 2002). Vegetation in riparian zones performs many ecosystem functions. Riparian vegetation stabilizes soil and reduces erosion. It is a source of organic matter and wood that serves as food and structure for aquatic organisms. The riparian vegetation is also a filter against input from upland areas, and can take up excessive nutrients and toxic compounds from surface and ground water before entering streams (Naiman et al. 1995; Naiman and Décamps 1997; Postel and Richter 2003). River ecosystems provide ecosystem services such as clean water and constitute important environments for e.g. recreation and tourism (Naiman et al. 1995; Postel and Carpenter 1997; Naiman et al. 2002; Postel and Richter 2003).

The species diversity in riparian ecosystems is influenced by flow regimes, the geographical morphology of the channel, climate and impacts from surrounding areas (Naiman and Décamps 1997). Riparian vegetation along watercourses is to a great extent controlled by the hydrology of the adjacent stream or river (Naiman and Décamps 1997; Nilsson and Svedmark 2002; Ström et al. 2011 A). Flood duration and frequency of regulated flow regimes often cause a reduced plant survival and growth (Johansson and Nilsson 2002). Species richness and the abundance of vascular plants have been found to decrease in rivers regulated for hydropower production, where water level variation differs from natural watercourses in having higher frequency and longer duration of flooding (Nilsson et al. 1997; Jansson et al. 2000). Dynesius et al. (2004) showed that this pattern is consistent even between continents (North America and Europe). To sustain the integrity of ecosystems along river courses there is a need for more natural seasonal variation in flow over the year (Poff et al. 1997). The largest ecological consequences caused by hydropower production come from fragmentation of the watercourse and the changes of the flow regime (Nilsson et al 1997; Malmqvist and Rundle 2002; Nilsson et al. 2005). An important structuring factor for riparian vegetation along rivers is the species dispersal along watercourses (Nilsson et al. 1991), which is disrupted in regulated rivers (Andersson et al. 2000). Dams prevent dispersal of propagules, sediment and organic material along the rivers, and changes in the timing and magnitude of water-level variation have altered river ecosystems (Poff et al. 1997; Andersson et al. 2000; Jansson et al. 2000).

Globally, the majority of the total discharge in the largest river systems is affected by flow-related fragmentation (Nilsson et al. 2005). Many types of river ecosystems are now lost and populations of species connected to these environments are therefore highly fragmented (Dynesius and Nilsson 1994). According to Sala et al. (2000), the biodiversity of the global freshwater ecosystems are decreasing at an alarming rate, much faster than even the most affected terrestrial ecosystems. Dudgeon et al. (2006) grouped the threats to global freshwater diversity into five categories: overexploitation; water pollution; flow modification;

2

habitat destruction or degradation; and invasion by exotic species. In addition to this, there are environmental changes, such as atmospheric nitrogen deposition, warming and shifts in precipitation and runoff patterns, which act over all the five categories (Dudgeon et al. 2006). Global climate change will likely increase the frequency of floods and droughts, thereby aggravating flow modification (Vörösmarty et al. 2000).

Day et al. (1988) described three main factors structuring riparian vegetation communities. The first was water depth, which is correlated with spring flood duration, magnitude of wave and ice erosion, fluctuations in water level and litter deposition. Riparian vegetation is distinctly zoned according to water depth (Day et al. 1988; Ström et al. 2011 A). The second gradient included primary productivity, litter deposition and litter removal during floods. The third was a fertility gradient caused by the interaction between flooding and disturbance from waves (Day et al. 1988). Kuglerová et al. (2014) argued that both river hydrology and upland inputs are main factors controlling riparian vegetation dynamics. Groundwater discharge increases the riparian species richness and is correlated to higher soil pH and higher nitrogen availability (Kuglerová et al. 2014).

River storage reservoirs and impoundments are used to store and release water when the energy is needed. Long-term regulation in storage reservoirs store water over seasons and years while run-of-river impoundments with short-term regulation meet the differences in daily needs of electricity (Renöfält et al. 2010). Water levels in run-of-river impoundments in Sweden vary only within a short interval (about 1 m) with daily and weekly shifts but lacking seasonal variation (Jansson et al. 2000). The high frequency changes in water level during winter results in unstable ice cover, with frequent ice-erosion events as push and breaks throughout the winter, and increases the total annual flood duration in impoundments. Winter flooding has been found to constitute a major limiting factor for vegetation in run-of-river impoundments (Johansson and Nilsson 2002). However, Johansson and Nilsson (2002) found that the largest impact was from the changes in duration and frequency of flooding during summer, likely because of the low inundation tolerance of the vegetation during periods of plant growth. Short-term water-level regulation results in stress and disturbance for riparian vegetation and erosion of fine grained substrate from the riverbanks, which makes the vegetation sparse except along the high-water level (Jansson et al. 2000). The riparian zones in regulated rivers are narrow and lack the zonation of the vegetation that characterizes free-flowing rivers (Jansson et al. 2000). Flows able to mobilize sediment and deposit material on the riversides are mostly lacking in regulated rivers (Jansson 2008). Jansson et al. (2000) also found that riverbanks along regulated rivers have more coarse-grained soils than free-flowing rivers. Due to the low rate of deposition of fine-coarse-grained substrate on the riverbanks and the ongoing erosion, vegetation establishment is hampered (Nilsson and Berggren 2000).

Riparian vegetation can be sorted into vegetation belts or zones depending on the ability of species to tolerate flooding and droughts (Auble et al. 1994). Auble et al. (1994) predicted that changes in maximum and minimum flows would cause significant changes in riparian vegetation. A study by Ström et al. (2011 B) showed that all vegetation belts in the riparian zone except for the amphibious vegetation are expected to decrease in species richness with anticipated climate change. They found the highest species richness in the riparian forest belt. The riparian forest belt is predicted to lose most species (3 – 7 species) (Ström et al. 2011 B). Free-flowing rivers are more capable to buffer against land-use changes and climate change through dynamic interactions within the river ecosystem and flow adjustments, while rivers with flow regulations and dams have lost much of their ability to adjust to new conditions and therefore will need more management actions to mitigate the impacts of climate change (Palmer et al. 2008).

3

1.2 Hydropower regulation and conservation management goals

Interest in implementing management actions to reduce the ecological impact of hydroelectric production has increased in recent years, for example as a result of the requirements in the EU Water Framework Directive (2000/60/EG) and the Swedish Environmental Quality Objectives. The Water Framework Directive requests that management actions toward more environmentally adapted flow regimes are implemented in regulated rivers (EU 2000). Good Ecological Potential should be reached for all waters classified as Heavily Modified Water bodies, according to the Water Framework Directive. The Swedish Environmental Objectives constitute of 16 different objectives, with one of them specifically considering the health of running water, “Flourishing Lakes and Streams”. This objective should be reached within one generation and is specified as: “Lakes and watercourses must be ecologically sustainable and their variety of habitats must be preserved. Natural productive capacity, biological diversity, cultural heritage assets and the ecological and water-conserving function of the landscape must be preserved, at the same time as recreational assets are safeguarded” (Environmental Objectives 2014). Hydropower production is the most profound impact on rivers in northern Sweden and management actions in regulated rivers are therefore prioritized to mitigate negative effects of hydropower production, according to the Water Framework Directive (Jansson 2008). There are several recommendations for management actions toward more natural flow regimes but this often induce a conflict with the hydropower production. This results in a need for finding management solutions with significant positive ecological effects and with relatively small negative impact on power generation (Renöfält et al. 2010).

1.3 Environmental flows

There has been a lack of consideration of the ecological effects of hydropower in impoundments and reservoirs from the time of development (Renöfält et al. 2010). Jansson et al. (2007) emphasised the importance of restoring ecological processes, which often means that the regulated flow regime needs to be re-naturalized to increase the connectivity of the river. The most efficient strategy to improve longitudinal and lateral connectivity is to reintroduce seasonal floods (Stromberg et al. 2007; Jansson 2008). To sustain freshwater biodiversity, conservation management actions toward more natural flow regimes are needed (Poff et al. 1997; Richter et al. 1997). In the Brisbane declaration (2007), environmental flows are described as “the quantity, timing, and quality of water flows required to sustain freshwater and estuarine ecosystems and the human livelihoods and well-being that depend on these ecosystems”. More natural variation in flow has large potential to improve the conditions for riparian vegetation (Stromberg et al. 2007; Jansson 2008). Such modifications in water-level regime can in some cases be done with relatively small losses in power production, provided that production in adjacent impoundments can be synchronized so the relative drop height remains the same. In other cases the situation might be the opposite, that the loss in power production is greater than the ecological gain (Jansson 2008).

The aim for using environmental flows can differ from improvements of a specific species to sustaining of entire ecosystems, and therefore they also differ in magnitude of mitigation measure (Tharme 2003). Tharme (2003) summarized 207 different methods of environmental flows and grouped them in four models: hydrological, hydraulic rating, habitat simulation and holistic methodologies. The first three focus most on one or a few specific species, while the holistic methodologies are built on the concept that both flow regime and ecosystem should function in a sustainable way and are probably the models that will be used most in the future. A river ecosystem needs a minimum amount of water and seasonal variations in the flow regime to be maintained, but according to the concept of environmental flows, the water in addition to these basic needs can be used for hydropower production (Tharme 2003).

Environmental Water Allocations (EWA) are examples of management actions that seek to create a water regime that follows natural variability, which includes a range of flows and not

4

just a minimum level (Dudgeon et al. 2006). The Range of Variability Approach (RVA) includes ecological relevant components of a flow regime to sustain river ecosystems: variability, magnitude, frequency, duration, timing and rate of change (Richter et al. 1997). Another framework, called “the ecological limits of hydrologic alteration” (ELOHA), have been proposed to set socially acceptable and ecological favourable goals for regional flow management, based on existing techniques and environmental flow methods (Poff et al. 2010).

There is a demand for studies concerning which aspects of natural flow dynamics that would favour different species (Jansson et al. 2007). There are only a few quantitative predictions of how the riparian vegetation will be affected by changes in flows (e.g. Johansson and Nilsson 2002), calling for more studies to provide a basis for implementing environmental flows in regulated rivers.

Most reaches in the Ume River are classified as heavily modified (Vattenmyndigheten Bottenviken 2010), and actions to meet good ecological potential according to the Water Framework Directive will be required. The association “Samverkansgruppen 3 regleringsmagasin”studied the Ume River impoundments in the municipality of Vindeln and presented management actions for achieving good ecological potential (Widén et al. 2013). Among the presented management actions were suggestions of environmental flows. My study was built on these environmental flow suggestions and presents predictions on how the riparian vegetation distribution would change with implementation of the different flow suggestions. Previous studies have shown that the distribution of different vegetation belts and the presence of individual species can be predicted with hydrological data (e.g. Jansson et al. 2000; Ström et al. 2011 A). The general methodology used in my study has previously been used to predict effects of dams (Auble et al. 1994) and climate-change scenarios on riparian vegetation distribution (Primack 2000; Ström et al. 2011 B).

1.4 Aim

The aim of this study was to assess potential responses of riparian plants to changes in water-level regulation in run-of-river impoundments to better match natural flow regimes. I surveyed the elevational extent of plant species in the riparian zones and calculated their probability of occurrence along the gradient of flooding duration in two run-of-river impoundments in the Ume River. The tolerance to flooding of the plant species in the Ume River was compared to the tolerance of flooding of plant species found in the free-flowing Vindel River. Knowing the flood tolerance limits for each species with present flow regimes in the impoundments, the relationships between plant occurrence and hydrology were used to predict the effect of three scenarios for more natural flow regimes on the spatial extent of the species.

The following questions were addressed in this study:

(1) How does the tolerance of plant species to flooding differ between riparian zones in run-of-river impoundments and free-flowing rivers?

(2) Assuming that the distributions of plant species in riparian zones are limited by flood frequency and duration, how would they respond to changes in water-level regimes altered to mimic water-level changes in free-flowing water bodies? Scaling up to entire impoundments, what is the potential area of riparian habitat where riparian vegetation can be predicted to establish with changes in water-level regimes?

5

2 Materials and methods

2.1 Study areas

Two run-of-river impoundments in the Ume River, regulated for hydropower production, were selected for this study, Bjurfors nedre and Harrsele in the municipality of Vindeln, Sweden (figure 1). The Ume River starts in the Scandinavian mountain range and empties into the Gulf of Bothnia east of the City of Umeå, after a length of 470 km and with a catchment area of 26814 km2_{. The river has produced hydroelectricity since the nineteenth} century, but large-scale transformation of the river with dams and power stations started in the 1950-ies, and the river is the third largest hydropower producer in Sweden (Vattenkraft.info 2014). Before regulation of the river it used to follow natural water-level fluctuations similar to the last large rivers still remaining free-flowing in Sweden (Torne, Kalix, Pite and Vindel Rivers). The upper part of the river now constitutes of high-capacity storage reservoirs with artificial water-level changes, with lowest level in spring, the highest in late summer and with falling water levels during autumn and winter. The middle and lower parts of the river are transformed into run-of-river impoundments with low storage capacity. Water levels fluctuate daily or weekly between the statutory high and low levels throughout the year. The riparian vegetation is often similar to the vegetation around lakes, dominated by species such as Carex rostrata, as the water-level amplitude is small (usually about 0.5 m) and with low current velocity. Close to the high-water level, vegetation with no clear dominant species is found, whereas lower elevations host amphibious species such as

Ranunculus reptans (Jansson et al. 2000).

Figure 1. Map of the Ume River. The black box shows the location of the study areas.

The Bjurfors nedre and Harrsele impoundments (figure 2) are owned by Statkraft Sverige AB. They have fall heights of 20 m and 54.5 m and a normal annual production of 348 and 950 GWh/year, respectively (vattenkraft.info, 2014).

6

Three sites in Bjurfors nedre and Harrsele, respectively, were selected for the survey (each site was between 50-150 m in length), based on two criteria: (1) The riverbank slope should not be steeper than 30%, and (2) the riparian substrate should primarily constitute of finer material such as silt or sand. This was done to ensure that there was potential for vegetation establishment at all levels in the riparian zone. Based on this, I made the assumption that the distribution of individual species at different elevations is mostly dependent on the water-level regime. This assumption is supported by experimental evidence from Ström et al. (2011 A) where turfs of riparian vegetation were transplanted to other elevations on the riverbank, resulting in species composition and abundance adjusting to the new hydrological conditions. The coordinates for each site are presented in table 1.

Figure 2. Map with the three sites in Bjurfors nedre (Bn1; Bn2 and Bn3) and Harrsele (H1; H2 and H3), respectively. (Map modified from eniro.se).

Table 1. Coordinates (WGS 84) for the three sites in Bjurfors nedre (Bn1; Bn2 and Bn3) and Harrsele (H1; H2 and H3), respectively. Coordinates (WGS 84) Bn1 N 64° 7.240', E 19° 32.121' Bn2 N 64° 7.137', E 19° 32.378' Bn3 N 64° 7.142', E 19° 33.155' H1 N 64° 4.858', E 19° 33.988' H2 N 64° 4.630', E 19° 33.938' H3 N 64° 1.613', E 19° 33.490'

7

2.2 Survey

The vegetation survey was done during August and September 2013. Six transects were placed perpendicular to the river channel at each site. The transects were evenly distributed along the sites. The average annual high-water level at each site was estimated visually by inspecting the transition of riparian vegetation to forest vegetation, accumulation of debris along the high-water level, and evidence of sedimentation and erosion indicating flooding. Each transect was constituted of seven 50 x 50 cm large plots. Plot 3 was placed with the upper end of the frame at the high-water level. From the centre point of plot 3 the other plots were distributed with 20 cm difference in elevation, plot 1 and 2 upwards and plot 4, 5, 6 and 7 downwards, using a clinometer and measuring rod. The total cover of vascular plants was estimated visually in percent for each plot. All species present in each plot were noted. The definition of a species followed the taxonomy in Mossberg and Stenberg (2003). In the following cases two or more species were treated as one taxon each: Agrostis spp., Callitriche spp., Carex spp., Hieracium spp., Luzula spp., Poa spp., Rumex spp., Salix spp. and

Taraxacum spp.

The substrate down to about 10 cm depth, approximately equivalent to the rooting depth of plants, was recorded for each transect into the percentage composition of the grain-size classes of Wentworth (1922) (clay, silt, sand, gravel, pebbles, cobbles and boulders). The elevational position of the centre point of each plot was measured with a total station (an integrated electronic theodolite and distance meter) with 1 cm accuracy (Geodolite 506, Trimble Navigation Limited, Sunnyvale, California, USA) in October 2013. To record the position of the high-water level, I measured several positions at all sites, and then used the one deemed to be most reliable at each site. In most cases, the high-water level positions from the same site were similar. A few plots had lost their markers when measuring with the total station. These plots were assumed to be located 20 cm in elevation away from adjacent ones, or in between the adjacent upper and lower plots.

2.3 Water-level data

I used the height data from the total station and calculated the difference in elevation between every plot and the high-water level of the site.

Water-level data for Bjurfors nedre and Harrsele was received from Statkraft Sverige AB. I used data on water levels per hour for the years 2007-2010 during the period May 1 to September 30 for Bjurfors nedre and Harrsele, respectively. The time period from May to September was selected since it encompasses the growing season, when plants are physiologically active and assumed to be more sensitive to water-level variation (Johansson and Nilsson 2002; van Eck et al. 2006). By plotting the water-level data (m above sea level) per hour for each year I calculated the average monthly high- and low-water levels for each impoundment (figure 3) and approximated it to a mean high- and low-water level for the entire period. High-water level was 164.9 and 144.9 m above sea level and low-water level was 163.7 and 143.7 m above sea level in Bjurfors nedre and Harrsele, respectively.

8

b.

Figure 3. Water-level variation for Bjurfors nedre (a) and Harrsele (b) from the growing season 2007. The upper continuous, black line marks the average high-water level (2007-2010) and the lower continuous, black line denotes the average low-water level (2007-2010).

Hydrological data for each plot was obtained by assuming that the average high-water level of each impoundment, as measured at the gauge of the power station, corresponded to the high-water level at each site (i.e. there were no thresholds affecting the high-water-level variation in between sites and gauges). I calculated the actual elevation (m above sea level) of all the plots by correlating the high-water level from the water-level data with the measured elevation data from the total station, i.e. by adding the earlier calculated differences in elevation (from plot to high-water level for the site) to the known high-water level for each impoundment. Then I calculated the duration of inundation of each individual plot. This was done by calculating the duration of flooding for each elevation from the water-level data at the gauge, and then assigning the inundation duration for each plot based on the difference in elevation between the plot and the high-water level. The data is presented as the average number of days of flooding per average growing season (2007-2010).

163.5 164 164.5 165 Elev a ti o n ( mas l)

May June July August September Bjurfors nedre 143.5 144 144.5 145 Elev a ti o n ( mas l )

May June July August September Harrsele

9

2.4 Environmental flow modifications

As a part of a project to identify measures to improve ecological conditions in the Ume River, alternative water-level variation regimes were developed that would better match the requirements of riparian plants while still allowing for hydropower production (Widén et al. 2013). The alternative water-level regimes mimic aspects of natural flow regimes that favour species rich riparian vegetation, in accordance with previous efforts linking aspects of hydrology to ecological conditions (Poff et al. 1997; Richter et al. 1997; Richter and Richter 2000; Arthington et al. 2006; Richter and Thomas 2007). The environmental water-level regimes suggested include three aspects (figure 4) (Nilsson et al. 1993; Johansson and Nilsson 2002; Nilsson and Svedmark 2002): (1) High water levels during the spring flood; (2) Slowly falling water levels after the spring flood; (3) Low water levels during the rest of the growing season. The simulated water levels were applied by modifying historical water-level data for the period May 1 to September 30 for the years 2007-2010. Below the simulated water-level regimes are described in more detail.

(1) High flows during the spring flood. The modified water levels simulated for this modification were a constant high-water level without short-term regulation during the period May 15 to 21 (164.9 and 144.9 m above sea level in Bjurfors nedre and in Harrsele, respectively), followed by a period (May 22 to June 6) during which I simulated water levels being 0.5 m higher than the actual ones, i.e. with short-term regulation of water levels. This “simulated spring flood” would remove plant litter and species intolerant of even short periods of flooding from affected areas (e.g. Picea abies and Vaccinium myrtillus). This would give opportunity for riparian plant species such as forbs and graminoids to establish provided that substrate conditions are suitable (i.e. presence of fine-grained soils).

(2) Slowly falling water levels after the springflood. The simulated water levels went from 0.5 m above to down to actual levels over the period June 7 to 30, with a 2 cm decrease per day (i.e. 0.5, 0.48, 0.46 m etc.), with short-term regulation. This would result in more pronounced zonation of the vegetation in the upper part of the riparian zone, given that species differ in their tolerance to flooding.

(3) Low water levels during the later part of the growing season (July 15 to August 31). The simulated water level used was a constant low-water level equivalent to the average monthly low-water level (163.7 and 143.7 m above sea level for Bjurfors nedre and Harrsele, respectively). This would give riparian species that have established more time to grow and accumulate biomass, which would increase winter survival.

10

Figure 4. Water level variation during a year in a free flowing river (blue line, the Vindel River), in a run-of-river impoundment (red line, the Ume River) and a suggestion where the water level variation of the impoundment has been adapted to favour establishment and distribution of riparian vegetation on the riverbank (black line). The figure illustrates three modifications of the water level regime important for riparian vegetation.

2.5 Data analysis

First, plant cover and species richness per plot were related to elevation in the riparian zone for the impoundments Bjurfors nedre and Harrsele separately. The cumulative number of species was counted in four elevation intervals in the riparian zones of each impoundment. All the species found in the plots at each elevation interval (0.3 m) were counted. Note that the number of plots differed among elevation intervals.

Second, the probability of occurrence of a species was related to duration of flooding. All the species found in at least eight plots per impoundment were used in the analysis. The species were graded as present (1) or absent in a plot (0). Logistic regression was used to analyze the probability that a species would be present depending on the elevation in the riparian zone for each impoundment. A Generalized Linear Model (binary logistic) was done for all species, which adapt a quadratic graph for each species dependent on the probability for the plant to occur at different elevations of the riverbank. Presence/absence of species was used as the response variable, and height and height2 _{were used as predictors. This resulted in a} predicted value of mean response for each species along the elevational gradient. Graphs with each species probability distribution were plotted dependent on height.

The probability curves were used to estimate the elevational extent of each species. To avoid counting species as present at elevations where the calculated probability of presence was low, the tails of the probability curves were cut off at 25 % of the maximum probability, giving the upper and lower limits of the riparian species. This was found to maximize the fit between observed and modeled range of occurrence. Some species (upland species) had hydrological niches that extended above the riverbank, and an upper limit could thus not be predicted.

11

The lower limit for amphibious species with hydrological niches far down in the river, could not be predicted either.

By comparing the water-level history data with the elevational extent (the hydrological niche) for each species the number of flooded hours corresponding to the upper and lower elevational limits were extracted. The simulated water-level data according to three alternatives for environmental flow regimes were compared with the modelled upper and lower limits for each species. Three modified water-level regimes were used: First with only a simulated spring flood (alternative 1 above), then with a spring flood and slowly reduced water levels (alternatives 1 and 2), and the last included all three flow modifications presented in chapter 2.4 (alternatives 1 to 3). I could thus predict new species distribution limits for the three different environmental flow alternatives, based on the assumption that the species would shift their elevational extent in the riparian zone according to changes in inundation duration. The riparian species were divided into two groups: upper riparian species with an upper elevational limit above the high-water level (0 h of inundation) and lower riparian species with an upper limit at a specific number of hours of inundation. As the upper limits for the upper riparian species were missing, 0.5 m (for the spring flood) was added to the original upper elevational limits in all modifications in order to achieve a realistic modification. The species were therefore grouped in four categories; upland, upper riparian, lower riparian and amphibious species. Note that the same species could belong to different groups in the two impoundments.

The difference between present elevational extent and the predicted extent for the three simulated water-level regimes were calculated for each species. For the upland and amphibious species, with only one elevational limit, the changes in elevational extent in meter were used. For the riparian species (i.e. species with both upper and lower limits in the riparian zone) the proportional change in elevational extent with the alternative simulated water-level regimes was calculated. For all four elevational species groups in both the Bjurfors nedre and Harrsele impoundments, except for Bn lower riparian, H upper and H amphibious where the number of modelled species were too few species, statistical tests were made to test if the predicted changes in extent for each simulated water-level regime differed significantly from zero, using one-sample t-tests. To test for differences in the predicted degree of change among the simulated water-level regimes, one-way ANOVAs with Tukey’s post-hoc tests were used.

All statistical tests were perfomed in IBM SPSS Statistics version 21 (IBM Corporation, Armonk, New York, USA).

To compare the inundation tolerance of species found in the run-of-river impoundments with the occurrence data of same species in a free-flowing river, data from the adjacent Vindel River was used. The data from the Vindel River (from the slow-flowing reaches Arvselet, Strycksele and Kronlund) was surveyed with the same method as I have used for the Ume River, but with 0.25 m between each plot as the riverbanks were wider and therefore needed more plots per transect in the Vindel River (Ström 2011). The species found in at least one of the localities in both the Ume River and the Vindel River were compared. The number of days flooded per average growing season (2007-2010) corresponding to the upper or lower limits in the riparian zone were compared between the rivers. In the cases where a species was found in two or more localities per river, the mean number of flooded days per locality was used.

The predictions of changes in elevational extent in the riparian zone are based on the assumption that plant species can adjust occurrence without being limited by e.g. substrate availability. In reality, shorelines along run-of-river impoundments are often heavily eroded as a result of wave action in combination with frequent water-level fluctuations aggravated by ice during winters. Therefore, data on the proportion of riverbanks with conditions suitable

12

for vegetation establishment that would have potential to benefit from changes in water-level regimes were compiled from inventories of riverbanks in the Ume River (Widén et al. 2013). The mapping of suitable riverbanks for vegetation was done for all the run-of-river impoundments between the dam in Harrsele and the dam in Grundfors, except for Hällforsen (lacking short-term regulation of water levels) and the bay Blåvikssjön in the Rusfors impoundment. All riverbanks were surveyed and classified by substrate and steepness. The definition of suitable riverbanks for vegetation establishment was that the substrate should constitute of silt or silt with sand or cobbles in the Harrsele, Bjurfors nedre and Bjurfors övre impoundments. For the remaining impoundments, lying upstream in the river, the definition was at least 70% silt or sand (i.e. less than 30% gravel, stones, boulders or bedrock) and maximum 90% steepness of the riverbank.

3 Results

3.1 Vegetation cover and species richness

In total, 101 species were found in the Bjurfors nedre impoundment and 81 species in the Harrsele impoundment (Appendix I). The substrate on the sites was mostly composed of silt or sand, with elements of cobbles and boulders in especially Bn2 and H2. Figure 5 illustrates the relationship between plant cover per plot and elevation in the riparian zone. The variation in total vegetation cover was considerable along the entire elevational gradient of the riverbanks.

a. b.

Figure 5. Total vegetation cover (%) per plot at different elevations (m above sea level) on the riverbanks in (a) the Bjurfors nedre (high-water level: 164.9 m above sea level) and (b) the Harrsele impoundments (high-water level: 144.9 m above sea level).

0 20 40 60 80 100 164 165 166 C o v er (%) Elevation (masl) Bjurfors nedre total vegetation cover

0 20 40 60 80 100 144 145 146 C o v er (%) Elevation (masl) Harrsele

13

There was no distinct pattern of correlation between species richness and elevation in Bjurfors nedre (figure 6a and c). The species richness in Harrsele was highest close to the high-water level (figure 6b and d).

a. b.

c. d.

Figure 6. Species richness per plot at different elevations in the riparian zone of (a) Bjurfors nedre (high-water level: 164.9) and (b) Harrsele (high-water level: 144.9).The cumulative species richness at different elevation intervals in the riparian zones are displayed in (c) for Bjurfors nedre and in (d) for Harrsele. Bjurfors nedre had 31 plots (n) between 164.3-164.6; 164.6-164.9, n=35; 164.9-165.2, n=25; and 165.2-165.5, n=19. Harrsele: 144.3-144.6, n=24; 144.6-144.9, n=31; 144.9-145.2, n=29; and 145.2-145.5, n=19. 0 5 10 15 20 25 164 165 166 Nu mb er o f s p ec ies Elevation (masl) Bjurfors nedre species richness 0 5 10 15 20 25 144 145 146 Nu mb er o f s p ec ies Elevation (masl) Harrsele species richness 0 20 40 60 Nu mb er o f s p ec ies Elevation (masl) Bjurfors nedre species richness 0 20 40 60 Nu mb er o f s p ec ies Elevation (masl) Harrsele species richness

14

3.2 Species distribution along the elevational gradient

The elevational extent of a subset of plant species as estimated by logistic regression are illustrated in figure 7. The elevational limits and the inundation duration these limits correspond to are presented in Appendix II for all the species present in at least eight plots. a.

b.

Figure 7. Probability curves from logistic regressions of the relationship between presence/absence of species and the elevation (m above sea level) in the riparian zone for a subset of species in Bjurfors nedre (a) and Harrsele (b). Explanations for abbreviations of species names are found in Appendix I.

3.3 Comparison of inundation tolerance between the Ume and the Vindel

Rivers

A list of the modelled upper and lower limits of the species found in both the Ume River and the Vindel River is presented per locality in Appendix III, expressed in inundation duration. For eleven species having lower elevational limits in the riparian zone, the difference in the estimated limit was less than 20 days between the rivers (figure 8), whereas for nine species, the difference between the rivers was larger. Of those nine species, all except Juncus

filiformis had lower limits in the Ume River than in the Vindel River. Thus, the eight

remaining species extended to elevations with longer duration of inundation in the Ume than the Vindel River.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 164 164.5 165 165.5 166 P ro b a b ili ty o f d is tri b u ti o n Elevation (masl) Bjurfors nedre Vacc_vit-i Angel_syl Caltha _p Coma_pa Filip_ulm Ran_rept Cala_pu Car_rost 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 144 144.5 145 145.5 146 P ro b a b ili ty o f d is tri b u ti o n Elevation (masl) Harrsele Vacc_vit-i Angel_syl Caltha _p Coma_pa Filip_ulm Ran_rept Cala_pu Car_rost

15

Figure 8. Comparison of maximum number of days of inundation per average growing season (2007-2010) for species with a lower elevational limit in the riparian zone in both the Ume River and the Vindel River. Note that Agros_cagi and Viola_palep consist of Agrostis capillaris and A. gigantea and of Viola palustris and V. epipsila. Explanations for abbreviations of species names are found in Appendix I.

Six species with upper elevational limits in the riparian zone had limits differing less than 20 days between the rivers (figure 9). Four species (Ranunculus reptans, Eleocharis palustris,

Myosotis laxa and Juncus alpinoarticulatus) had limits differing more than 20 days between

the rivers, all of them having limits equivalent to shorter duration of flooding in the Ume River.

Figure 9.Comparison of minimum number of days of inundation per average growing season (2007-2010) for

riparian/amphibious species with an upper elevational limit in both the Ume River and the Vindel River. Explanations for abbreviations of species names are found in Appendix I.

0 20 40 60 80 100 120 140 M a xi mum d a ys i n u n d a te d p er g ro w in g s ea so

n Ume River and Vindel River

Maximum days Ume

Vindel 0 20 40 60 80 100 M in imum d a ys i n u n d a te d p er g ro w in g sea so n

Ume River and Vindel River

Minimum days Ume

16

3.4 Predicted changes in extent of riparian plant species as a result of

changes in water-level regime

Upland species (defined as species with a lower limit above or in the riparian zone, but no upper limit) in the Bjurfors nedre impoundment were predicted to lose habitat with their lower limits moving upward on the riverbank with simulated water-level regimes according to alternatives 1 and 2 (figure 10a, table 2), assuming similar flood tolerance with modified water-level regimes. The predicted mean distributional change of the upland species was almost significantly different between the three simulated water-level regimes in the Bjurfors nedre impoundment (one-way Anova: F2,5=4.056, P=0.051).

The predicted changes in elevational extent of the upper riparian species in both the Bjurfors nedre and Harrsele impoundments differed significantly between the three alternative simulated water-level regimes (one-way Anova: F2,14=8.57, P<0.001 and F2,14=23.97, P<0.001). All species were predicted to increase in extent with all alternatives, but alternative

3 was predicted to result in the highest habitat gain (figure 10b and d, table 3). The predicted distributional change of the lower riparian species in Harrsele also differed significantly between the modifications (one-way Anova: and F2,7=58.956, P<0.001). The extent was predicted to increase in all cases with significantly larger changes for alternative 3 (figure 10e, table 3). The lower riparian species in Bjurfors nedre exhibited similar changes, but there were too few species to test this statistically (table 3).

The predicted change in the extent of amphibious species in Bjurfors nedre differed significantly between the alternative simulated water-level regimes (F2,14=4.703, P=0.017) but all were predicted to increase the extent (figure 10c). In contrast to the riparian species, the amphibious species are predicted to increase most with alternative 2 and least with alternative 3 (figure 10c). According to table 2, there were three amphibious species (R.

reptans, Tillaea aquatica and Agrostis stolonifera) predicted to lose habitat with alternative

17

a. b. c.

d. e.

Figure 10. Predicted change in elevational extent of species for three simulated water-level regimes (Mod 1, 2 and 3) (mean ± 95% CI). In Bjurfors nedre, the changes in upland species (N=6) are in meter (a), the upper riparian species (N=15) in percent (b) and the amphibious species (N=15) in meter (c). In Harrsele, the changes in upper riparian (N=15) and lower riparian species (N=8) are in percent (d and e). The calculations are based on changes during an average growing season (May-September) for the period 2007-2010. Changes significantly different from zero are indicated: *p<0.05, **p<0.01, ***p<0.001 (one-sample t-tests). Significant differences between the three modified water level regimes are indicated with different letters (a or b), where the letters are in brackets if the difference was only marginally significant (0.05 <p<0.1) (Tukey’s post hoc tests).

18

Table 2. Predicted change in the lower elevational limit of the upland species and the upper elevational limit of the amphibious species according to three alternative simulated water-level regimes (Mod 1, 2 and 3). Explanations for abbreviations of species names are found in Appendix I.

Bjurfors nedre

Mod 1 Mod 2 Mod 3 Mod 1 Mod 2 Mod 3

Upland species (m) (m) (m) Upland species (m) (m) (m)

Filip_ulm -0.06 -0.1 0.5 Alnus_inc -0.36 -0.36 -0.36 Geran_sy -0.06 -0.1 -0.01 Vacc_vit-i -0.41 -0.41 -0.41 Agros_ca -0.05 -0.1 0.06 Desc_flex -0.36 -0.36 -0.36 Desc_ces -0.06 -0.1 0.5 Equi_prat -0.3 -0.3 -0.3 Equi_sylv -0.1 -0.1 -0.1

Mod 1 Mod 2 Mod 3 Mod 1 Mod 2 Mod 3

Amphibious species (m) (m) (m) Amphibious species (m) (m) (m)

Caltha _p 0.1 0.25 0.25 Ran_rept 0.09 0.13 0.11 Epilo_ad 0 0 0 Tillae_aq 0.08 0.13 0.05 Galiu_pa 0.4 0.4 0.4 Junc_al 0.1 0.1 0.1 Lysim_th 0.1 0.1 0.1 Myos_lax 0.06 0.1 0.01 Ran_rept 0.05 0.1 -0.14 Rum_lon 0.1 0.25 0.25 Tillae_aq 0.05 0.1 -0.14 Agros_st 0.05 0.1 -0.06 Car_rost 0.06 0.1 0.01 Car_spp 0 0 0 Car_vesi 0.4 0.4 0.4 Equi_fluv 0.4 0.4 0.4 Junc_al 0.06 0.1 0.01 Junc_fil 0.06 0.1 0.07

19

Table 3. Predicted change in elevational extent of upper riparian species and lower riparian species according to three alternative simulated water-level regimes (Mod 1, 2 and 3). Explanations for abbreviations of species names are found in Appendix I.

Bjurfors nedre

Mod 1 Mod 2 Mod 3 Mod 1 Mod 2 Mod 3

Upper riparian species (%) (%) (%) Upper riparian species (%) (%) (%)

Alnus_inc 35.5 32.3 64.6 Angel_syl 70.0 61.7 75.0 Vacc_vit-i 50.0 31.3 31.3 Bistor_viv 72.4 63.8 77.6 Angel_syl 62.9 57.1 61.4 Coma_pa 41.2 38.2 88.2 Coma_pa 48.9 44.4 111.1 Epilo_ad 65.6 59.4 89.1 Gymn_dr 50.0 31.3 31.3 Filip_ulm 43.8 39.6 57.3 Maian_bi 44.4 44.4 44.4 Leont_au 52.5 47.5 71.3 Oxalis_ac 57.1 35.7 35.7 Viola_pal 68.3 63.3 83.3 Ran_acri 75.0 66.7 93.3 Agros_spp 83.7 75.5 83.7 Rub_sax 48.9 44.4 47.8 Cala_pu 48.8 44.2 54.7 Trient_eu 50.0 50.0 50.0 Car_aqu 47.8 43.3 74.4 Tuss_far 44.0 40.0 90.0 Car_rost 42.9 39.0 87.7 Valeria_s 32.1 28.6 45.7 Car_spp 35.0 30.8 37.5 Cala_pu 33.8 30.8 61.5 Desc_ces 38.2 34.5 51.8 Equi_arv 27.5 25.0 62.5 Equi_arv 60.0 54.3 81.4 Poa_spp 48.9 44.4 88.9 Equi_sylv 51.4 42.9 42.9

Mod 1 Mod 2 Mod 3 Mod 1 Mod 2 Mod 3

Lower riparian species (%) (%) (%) Lower riparian species (%) (%) (%)

Carda_pr 5.0 18.8 81.3 Caltha _p 40.2 36.8 94.3 Prunel_v 8.0 0.0 80.0 Carda_pr 25.0 21.7 65.0 Galiu_pa 42.7 37.8 95.1 Myos_lax 5.6 5.6 107.4 Prunel_v -1.6 7.8 75.0 Agros_st 6.3 6.3 68.9 Eleoc_pal 4.0 6.0 90.0 Junc_fil 59.4 53.1 87.5

The length of riverbanks in run-of-river impoundments of the Ume River that could be suitable as habitat for establishment of riparian vegetation and thus potentially could benefit from changes in the water-level regime are presented in table 4. Seventyone percent of the area in the Harrsele impoundment and 45% of the area in the Bjurfors nedre impoundment were found to be suitable for vegetation establishment (i.e. having fine grained substrates and steepness less than 90%). In total, 47% of the length (km) of the riverbanks along the run-of-river impoundments (Harrsele - Rusfors) in the Ume River is suitable for vegetation establishment.

20

Table 4. Run-of-river impoundments in the Ume River, from the Harrsele to the Rusfors impoundment. Note that the Rusfors impoundment is displayed as three separate parts. Total riverbank length and riverbank length and area with conditions suitable for establishment of vegetation that would potentially benefit from changes in water-level regime.

Harrsele Bjurfors nedre Bjurfors övre Tuggen Betsele Bålforsen Rusfors nedre Rusfors övre Rusfors Juktån

Riverbank length (km) 25 18 134 57 21 46 57 73 73 Suitable riverbanks (proportion of length, %) 11 24 18 72 64 45 65 61 67 Suitable riverbanks (proportion of area, %) 71 45 54 86 61 22 81 69 83

4 Discussion

4.1 Comparison of species distribution in riparian zones between

run-of-river impoundments and a free-flowing run-of-river.

The comparison between the Ume River and the Vindel River (figures 8 and 9) indicate that the method used was based on realistic assumptions, since more than half of the species exhibited similar flooding tolerances between the rivers. The differences in elevation extent observed were probably an effect of river regulation. The species in the Ume River that occurred at longer periods of inundation (at the lower elevational limit) were often found only as small plants further down on the riverbank. Such occurrences are not as common on riverbanks of the Vindel River since the free-flowing river host more plant species and riparian zones have a higher vegetation cover, especially at lower riparian elevations (Jansson et al. 2000), making it harder for seedlings to survive and establish outside their preferred habitat. In contrast, lack of competition from other species in the Ume River probably allowed these species to use more of their fundamental niches than in the Vindel River. This is supported of an experimental study by Grace and Wetzel (1981) which showed that removal of a dominant species may result in subordinate species to extend their distribution in riparian zones.

The elevational extent of the species with large differences between rivers in their upper elevational limit also showed the trend that the species in the Ume River had a wider elevational extent, i.e. the amphibious species could tolerate more days of drought in the Ume River. The reason for this could be similar as for the lower limit, that is, less interspecific competition in the Ume river, having a more sparse and species-poor riparian vegetation (Jansson et al. 2000). However, a difference between the rivers is the flood frequency, being higher in the run-of-river impoundments having daily and weekly fluctuations instead of seasonal variation in water level. The amphibious species could tolerate more days in drought in total as the fluctuation frequency were higher than in the Vindel River, and the opposite, probably caused by the same reason, were true for the riparian species that could tolerate more days of flooding than in the Vindel River. The higher rate of erosion of fine-grained substrates in regulated rivers that reduce habitat availability (Jansson et al. 2000) could also be a factor that reduced the competition from other species that otherwise would have limited the elevational extent of the species growing in the regulated river. However, sheltered parts of the riverbanks, often with boulders preventing erosion, were selected for the survey so that the otherwise increased erosion risk in regulated rivers would not affect the vegetation.

Carex rostrata was one of few species that tolerated fewer days with drought in the upper

elevational limit in the Ume River than in the Vindel River (figure 9). This suggest that other factors than competition might control its upper elevational limit, as it can extend further up in the Vindel River despite higher competition from more species in the free-flowing river. C.

21

rostrata prefer really wet sites with high groundwater supply (Wierda et al. 1997). The

species only occurs in free-flowing rivers in riparian zones with permanent groundwater supply, being sensitive to large water-level fluctuations (R. Jansson, personal observations), suggesting that the elevational extent was independent of inundation duration. In contrast, C.

rostrata is common on lakeshores and shorelines along run-of-river impoundments. Filipendula ulmaria was found to tolerate more days of inundation in the Ume River than in

the Vindel River (figure 8), but this was probably because of absence of competition (as discussed above) since Johansson and Nilsson (2002) found F. ulmaria to be able to establish below its normal elevational extent, but without positive growth in the absence of competition. They found that the growth rate of F. ulmaria was significantly lower in regulated rivers than in free-flowing rivers and that the plant performance was positively correlated to riverbank elevation while flood duration affected negatively. The experiment by Johansson and Nilsson (2002) showed that the plant performance would increase if the flood duration and frequency were reduced. So, even if a number of species had a wider elevational extent in the Ume River, plant performance was probably poor, and occurrences represented by short-lived individuals.

4.2 Predicted plant responses to simulated water-level regimes

I predict that all the three alternative water-level regimes would result in significant increases in the average extent of riparian and amphibious plant species along both impoundments (figure 1o). The purpose with the alternative water-level regimes is to extend the area in the riparian zone being flooded short periods. This would be achieved by a simulated spring flood, adding riparian habitat with short inundation duration at the top, and keeping water levels low in the latter part of the growing season. Upland species are predicted to be removed upward and leave place for the riparian vegetation, while riparian and amphibious species are predicted to increase their realized niches with the modified water level regimes (figure 10). The combined effect (modification 3) is predicted to result in the largest increases in elevational extent, but such a management action is also the one that would have the highest negative impact on power production (Renöfält et al. 2010; Widén et al. 2013). The predictions for alternative 3 is in line with the findings of Auble et al. (1994) and Johansson and Nilsson (2002), who predicted that reduced daily variation in flow would increase the abundance of riparian species. The first alternative with an introduced spring flood would also increase the average extent of riparian plant species, but not to the same degree as modification 3. Modification 2 (spring flood with slowly decreasing water level) would result in smaller increases in elevational extent since flood duration is slightly longer compared to modification 1 (spring flood only). However, the purpose of modification 2 would be to enhance zonation rather than increase habitat availability.

The modifications toward a more natural flow regime, with introduced spring flood and a lower water level in late summer are predicted to favour several riparian species with as much as 70-80% expansion in elevational extent (table 3). Scaling up to entire impoundments, less than half (47%) of the length of the riverbanks along the Ume River are predicted to benefit from re-regulation since large parts of the riverbank length are eroded with only coarse-grained substrate and a steep slope. The length of potential habitat for riparian vegetation to establish, summarized in table 4, is much lower than the proportion of the area suitable, since the riverbanks were wider where the conditions were suitable for vegetation establishment. The potential habitat (in area) the species are predicted to gain with changed flow regime could still be large, but concentrated to few localities. It is important to find where this kind of mitigating measure would have largest effects, as the riverbanks in many rivers are so changed because of regulation or timber floating that there are almost no potential habitat left to colonize even if the flow regime would be re-regulated (Renöfält et al. 2010). Other mitigation measures, like erosion protection, could perhaps be more effective where the riverbanks are highly degraded. For this survey, riverbanks with fine-grained substrate, often with boulders protecting the riverbank from erosion, were used. As long as fine-grained substrate was a major part of the riverbank, the vegetation cover was

22

rather well established, which indicates that erosion protection could be as important in increasing riparian vegetation cover as changing the flow regime.

I assumed that the distribution of plant species in riparian zones is limited by tolerance to inundation. Two results indicate that this assumption might have been violated. First, many plots had low plant cover (figure 5), indicating that few or no species were present in some patches. According to Jansson et al. (2000), plant cover on run-of-river impoundment shorelines are high only along the high-water level. Since several plots had low plant cover, other factors than inundation duration might have influenced the species distribution. However, the plots with low cover were distributed across all elevations in both impoundments. This increases the sampling effort needed to model plant distribution, but does not mean that there is any bias detecting upper and lower limits. Second, although both the upper and lower distribution limits were rather similar between the Ume and Vindel Rivers, a few species extended to lower elevations in the Ume River (figure 8 and 9). As discussed in chapter 4.1, these species presumably have wider elevations in the Ume River due to competitive release as some species present in the free-flowing river are missing and plant cover is lower in the regulated river.

The highest species richness in Harrsele (figure 6), found close to the high-water level, corresponds to the distribution noted in the free-flowing Vindel River (Ström 2011). However, the number of species at higher elevation on the riverbanks in the Vindel River does not show the same falling trend as in Harrsele, but the fact remains that it were fewer plots surveyed at higher elevations in Harrsele. Several plots in Harrsele had almost no species over the entire elevational gradient, which not was the case in the Vindel River, where such species poor plots only were noted in the amphibious species zone. The fact that the species richness and plant cover are lower in regulated rivers compared to free-flowing rivers is in line with the findings of Jansson et al. (2000) and Nilsson et al. (1997).

Studies like mine, providing quantitative predictions of vegetation responses to mitigation measures are needed to be used as a basis for water management decisions (Auble et al. 1994). A more comprehensive study, or new studies with more parameters compared between the regulated river and the free-flowing river would be valuable, providing even better grounds for management decisions. The predictions from my study should be relevant also to other run-of-river impoundments in northern Sweden, as long as the assumption that the elevational extent of species in riparian zones is mostly controlled by river hydrology can be presumed. The assumption is supported by Auble et al. (1994); Naiman and Décamps (1997); Nilsson and Svedmark (2002) and Ström et al. (2011 A).

To test or verify the predictions of this study different experiments can be done. This could be large-scale experiments with re-regulation of an impoundment in collaboration with a hydropower company, or lab experiments testing responses of plant species individually and in combinations to variation in inundation (Vervuren et al. 2003). A third way is to plant individual plants or transplant turfs of vegetation among levels in riparian zone, with different hydrological conditions. Such experiments with potted plants or vegetation being transplanted along an inundation gradient in riparian zones has already been done by Johansson and Nilsson (2002) and Ström et al. (2011 A).

One may ask if it is realistic to expect that riparian vegetation communities would benefit from changed water-level regimes. First, there have to be suitable habitats available for plant establishment, and large proportions of the riverbanks along the impoundments are too steep or lack suitable substrate for vegetation establishment (Renöfält et al. 2010). Second, for new species to colonize, increasing species richness to levels of free-flowing rivers (Nilsson et al. 1997; Jansson et al. 2000), there also need to be dispersal pathways and suitable conditions for establishment. As dispersal of species is disrupted by dams (Nilsson et al. 1991; Andersson et al. 2000), the restoration myth “Field of Dreams” (Hilderbrand et al. 2005) might apply in some extent, i.e. that even if the water-level regime is favourable for the