Mitigation of acidified salmon rivers – effects of liming on young brown trout Salmo trutta
Abstract
In southern Norway, 22 acidified rivers supporting anadromous salmonids were mitigated with lime to improve water quality and restore fish populations. In 13 of these rivers, effects on Salmo trutta and Salmo salar densities were monitored over 10–12 years, grouped into age 0 and age ≥ 1 year fish. These rivers had a mean annual discharge of between 4·9 and 85·5 m3 s−1, and six of them were regulated for hydro-power production. Salmo salar were lost in six of these rivers prior to liming, and highly reduced in the remaining seven rivers. Post-liming, S. salar became re-established in all six rivers with lost populations, and recovered in the seven other rivers. Salmo trutta occurred in all 13 study rivers prior to liming. Despite the improved water quality, both age 0 and age ≥ 1 year S. trutta densities decreased as S. salar density increased, with an average reduction of >50% after 10 years of liming. For age 0 year S. trutta this effect was less strong in rivers where S. salar were present prior to liming. In contrast, densities of S. trutta increased in unlimed streams above the anadromous stretches in two of the rivers following improved water quality due to natural recovery. Density increases of both age 0 and age ≥ 1 year S. salar showed a positive effect of river discharge. The results suggest that the decline in S. trutta density after liming is related to interspecific resource competition due to the recovery of S. salar. Thus, improved water quality through liming may not only sustain susceptible species, but can have a negative effect on species that are more tolerant prior to the treatment, such as S. trutta.
Introduction
Human-induced environmental changes can affect native species negatively. During the 20th century, salmonid abundance decreased in rivers in northern countries because of water acidification. In southernmost Norway, Atlantic salmon Salmo salar L. 1758 disappeared from a total of 25 highly acidified rivers (Hesthagen & Hansen, 1991). These rivers had a pH of 4·5–5·3 and a concentration of inorganic toxic aluminium of 70–160 μg l−1. Brown trout Salmo trutta L. 1758, regularly occurring in these rivers, are less vulnerable to acidification than S. salar. At pH 5·5, however, their reproduction rate is c. half of that above pH 6·0 (Henrikson & Brodin, 1995) and populations are usually extinct at pH below 5·0 (Degerman & Lingdell, 1993; Hesthagen et al., 2008). Therefore, in many acidified rivers S. trutta survived although often in a reduced state despite S. salar being extirpated.
To restore and enhance lost and reduced populations of S. salar and S. trutta in acidified Norwegian rivers, a liming programme was initiated, mainly using limestone powder (Sandøy & Romundstad, 1995; Clair & Hindar, 2005). The mitigation measure improved the survival of juvenile S. salar in acidified rivers (Lacroix, 1996; Staurnes et al., 1996). Full-scale liming of one river was initiated in 1985, the River Audna, and another 21 rivers were included in the programme from 1987 to 2002 (Hesthagen & Larsen, 2003; Hesthagen et al., 2011; Larsen et al., 2015). Ten of these 22 limed rivers had lost their S. salar, while the populations were strongly reduced in the remaining 12 rivers. S. trutta were in a better state than S. salar, and were not extirpated from any of these rivers.
Application of limestone powder increased pH, reduced toxic inorganic levels of Al and the water quality became suitable for S. salar (Kroglund et al., 2007, 2008). Thus, S. salar were re-established in all the limed rivers (Hesthagen et al., 2011). There was a significant effect of time after liming in both formerly lost and damaged stocks, on densities of both fry (age 0 years) and older parr (age ≥ 1 year). The rates of increase in densities of young S. salar in these two status categories were not significantly different in either age group. Regular stocking, which was carried out in seven rivers since liming commenced, and rivers unaltered by hydropower development generally had higher fry densities and faster increase in older parr densities.
Species do not live in isolation, and interactions between species may result in indirect effects on their abundance with liming (Mant et al., 2013). Thus, the recovery of S. salar may have influenced the density of sympatric S. trutta negatively. These two species are among the most abundant freshwater fishes in Scandinavian rivers, where they compete for space and food (Kalleberg, 1958; Stradmeyer et al., 2008; Berg et al., 2014). Salmo salar and S. trutta are often spatially segregated with juveniles of the former typically occurring further offshore and in deeper habitats than S. trutta, presumably due to strong interference interactions (Heggenes et al., 1999). Furthermore, in the River Bran, Scotland, S. salar growth and survival were reduced by the presence of S. trutta (Mills, 1964, 1969). Kennedy & Strange (1986) reported similar results from streams in Northern Ireland, and Skoglund et al. (2012) showed that juvenile S. trutta dominated similar sized S. salar in stream channels.
Here, effects of improved water quality through liming was assessed on S. trutta densities in 13 Norwegian rivers that also contain S. salar, that increased after liming (Hesthagen et al., 2011). Changes in densities of S. trutta were tested against that of young S. salar, river flow, river-regulation for hydro-power production and status of S. salar prior to liming in terms of lost or reduced stocks. It was hypothesized that: densities of S. trutta decreased concurrently with the increase in abundance of young S. salar following liming; the decrease in S. trutta density was higher in rivers where S. salar were lost pre-liming, than where S. salar still remained, because there was no interspecific competition when liming started; the decrease in abundance of S. trutta was stronger in smaller than in larger rivers, because competition should be keener in smaller rivers where these two species cannot segregate spatially to the same extent as they do in larger rivers.
Materials and methods
Study area
The study rivers are located in southernmost and south-western Norway, in the counties of Agder and Rogaland (Fig. 1). In most cases, S. salar, S. trutta and European eel Anguilla anguilla (L. 1758) were the only fish species caught in these rivers (Hesthagen et al., 2011; Larsen et al., 2015). While no stocking of S. trutta was carried out, the increase in S. salar densities was to a small extent, supplemented by stocking (Hesthagen et al., 2011). The increase in S. salar density, however, was to a small extent driven by stocking (Hesthagen et al., 2011). Brook lampreys Lampetra planeri (Block 1784), threespined sticklebacks Gasterosteus aculeatus L. 1758, ninespined sticklebacks Pungitius pungitius (L. 1758), brook trout Salvelinus fontinalis (Mitchill 1814) and European flounders Platichthys flesus (L. 1758) were sporadically found in some of the rivers.
The rearing stretch for anadromous fish in the study rivers ranged between 4 and 50 km (Table I). Mean values for width and annual flow ranged between 21–111 m and 4·9–85·5 m3 s−1, respectively (cf. Anonymous, 2014). Five of the rivers have a mean flow ≥30 m3 s−1, enough to sustain populations of large-sized salmon [3 sea-winter (SW) fish]. Furthermore, eight of the rivers were smaller with a mean flow ≤17 m3 s−1, typically used for spawning by 1 and 2 SW salmon (Jonsson et al., 1991). Six of the rivers were regulated for hydroelectric power generation. Four were heavily regulated due to the creation of obstacles for upstream, adult migration, low water flow, passage through dams and descent of smolts through tunnels with installed turbines. Two rivers, i.e. Tovdalselva and Ogna, are affected by regulations as some of their water is diverted.
| No.a | River | S. salar rearing stretch(km) | Mean width (m) | Mean flow (m3 s−1) | S. salar status prior to limingb | Hydro-power regulationc |
|---|---|---|---|---|---|---|
| 1 | Storelva | 15 | 23 | 13·2 | 1 | 1 |
| 2 | Tovdalselva | 35 | 78 | 65·0 | 0 | 1 |
| 3 | Mandalselva | 48 | 95 | 85·5 | 0 | 1 |
| 4 | Lygna | 20 | 63 | 30·0 | 0 | 0 |
| 5 | Kvina | 14 | 111 | 32·0 | 0 | 1 |
| 6 | Soknedalselva | 12 | 31 | 17·0 | 0 | 0 |
| 7 | Bjerkreimselva | 50 | 49 | 54·4 | 1 | 0 |
| 8 | Ogna | 30 | 21 | 6·6 | 1 | 1 |
| 9 | Frafjordelva | 5 | 35 | 14·3 | 0 | 0 |
| 10 | Espedalselva | 13 | 35 | 12·2 | 1 | 0 |
| 11 | Jørpelandselva | 3 | 25 | 6·2 | 1 | 1 |
| 12 | Vikedalselva | 10 | 21 | 10·3 | 1 | 0 |
| 13 | Rødneelva | 4 | 18 | 4·9 | 1 | 0 |
Liming in 11 of the study rivers was carried out continually with limestone powder (crushed calcium carbonate) from dosers controlled by water flow and pH below the liming sites (Clair & Hindar, 2005). In the Soknedalselva and Jørpelandselva, the liming was performed by liming lakes in their catchments. Liming produced a satisfactory water quality in all the study rivers, with annual mean pH at 6·3 and concentrations of inorganic toxic aluminium at 7 μg l−1 (Hesthagen et al., 2011).
Sampling
Between 1987 and 2005, young S. salar and S. trutta were sampled annually, using a portable backpack electrofishing apparatus (1600 V DC unloaded) in the 13 study rivers (Hesthagen et al., 2011). All rivers were sampled from 1 year before liming to 12 years or more after liming started. With few exceptions, the fish were sampled at low flow in August to provide comparable data. In most cases, the same people were involved in the sampling. Between six and 20 fixed stations were established in each river depending on their size, and sampling areas at each station ranged between 100 and 150 m2. Only stations that were sampled annually were included in the analysis. Each station had a width of 4–8 m from the riverbank with water depths between 10 and 50 cm. Each station was sampled in an upstream direction in three successive runs. After each run, the total length (LT, mm) of each fish was measured and kept in cages for release after the final run. Based on LT, the fish were divided into two age categories, namely fry (age 0 years) and older parr (age ≥ 1 year). LT of the fish in these two age groups ranged between 45 and 60 mm and 75 and 130 mm, respectively. Some individuals were sacrificed for age determination by use of scales and otoliths (Jonsson, 1976). These data were used to verify the grouping into age 0 years and older (age ≥ 1 year), based on length.
Fish densities were estimated by the removal method using data from the three successive runs (Bohlin et al., 1989). In each river, separate probabilities of capture for age 0 and age ≥ 1 year groups were estimated by compiling annual catch data from all stations for each run (Hesthagen et al., 2011). This is recommended when the total catch on each station is fewer than 35 individuals (Bohlin et al., 1989). This was largely the case in the study rivers, especially during the first years of the study. Annual mean densities 100 m−2 were estimated for each river. Furthermore, mean annual density for all rivers were estimated. The rivers were separated in regulated and unregulated, and with formerly lost and damaged S. salar populations at the start of liming (Table I). A total of 56 072 age 0 years and 20 332 age ≥ 1 year S. trutta were caught during the study.
As reference sites, densities of young resident, allopatric S. trutta in some tributary streams and inlets and outlets to lakes in the rivers Vikedalselva and Bjerkreimselva catchments (cf. Fig. 1, river nos 7 and 12), were monitored. The width of these streams generally ranged between 1 and 2 m, and 4 and 6 m, respectively. None of the sites sampled were affected by liming, local water pollution or habitat destruction. These localities do not contain S. salar due to barriers to migration. In the river Vikedalselva catchment, sampling of fish was carried out each year between 1987 and 2005 (Hesthagen et al., 2001, 2016). Fish data from the anadromous stretches were available during the same period, i.e. 19 years. In the river Bjerkreim catchment, data on resident S. trutta were not available in 1988, 1994, 2002, 2003 and 2005. In that river system, sampling on the anadromous stretches was carried out from 1996 to 2005, giving corresponding data for seven years. In most cases, 20–24 sites were sampled annually in each catchment by the same portable back-pack electrofishing apparatus used on the anadromous stretches. The sampling was carried out in late August to early September. Each site was sampled in a single run during the first 6 years of the study (1987–1992), and in three successive runs in later years (1993–2005). From the three catches in this last period, probabilities of capture (p) for age 0 and age ≥ 1 year fish was estimated. Finally, mean annual densities were estimated for sites in each catchment.
Analysis
Change in density of S. trutta at reference sites
Change in the natural logarithm of number of fish 100 m−2 year−1 was estimated in an autoregressive model with river and age class as categorical variables and year as a continuous variable. Residuals were assumed to be normally distributed and independent among rivers and age classes, but autocorrelated over years according to an AR(1) process.
Change in density of S. trutta and S. salar after liming
Density in population i at year j was modelled as a function of number of years after liming, t, using an autoregressive random-regression model: 
where subscript * denotes the mean taken over this level, b is the relative change in density over time for all populations, the random effect c is the deviation of each population for change in density over time assumed to be normally distributed and e is the residual assumed to be normally distributed and autocorrelated within population according to an AR(1) process. This model was fitted independently for the each age group of both species. In six rivers, S. salar was lost prior to liming (Table I). Data with zero density was not included in this analysis, and the parameter b is thus an estimate of relative change in density after the population was established.
This basic model was extended by allowing the parameter b (the change in density over time) to vary systematically with additional explanatory variables. Explanatory variables used in the analyses were: water regulation for hydropower purposes (categorical variable 0 = not regulated, 1 = regulated); S. salar status before liming started, categorical variable 0 = lost, 1 = still present; river size in terms of mean annual water flow. Inclusion of explanatory variables were evaluated based on Akaikes information criterion (AIC; Akaike, 1974) corrected for sample size, AICc (Burnham & Anderson, 2002).
Results
Densities of S. trutta in reference sites
In streams located in the catchments above anadromous stretches of the rivers Bjerkreimselva and Vikedalselva, age 0 years S. trutta increased by 9% and 5% annually respectively, while age ≥ 1 year S. trutta increased by 7% and 3% annually. The increase of age ≥ 1 year S. trutta in the river Vikedalselva was not statistically significant, however (Fig. 2). In contrast, there were small changes in young S. trutta densities on the downstream stretches supporting anadromous S. salar and S. trutta in the river Bjerkreimselva, while the density decline in the river Vikdedalselva was strong.
) and non-anadromous S. trutta (
) in rivers subject to liming in southernmost and south-western Norway. (a) Age 0 years [
: ln y = (0·088 ± 0·025)x − (172·3 ± 50·1), r2 = 0·56, P < 0·001; 
: y = (0·024 ± 0·050)x − (44·4 ± 99·5), r2 = 0·27, P > 0·05] and (b) age ≥ 1 year [
: ln y = (0·071 ± 0·025)x − (139·5 ± 50·1), r2 = 0·61, P < 0·01; 
: ln y = (−0·007 ± 0·050)x + (16·0 ± 99·5), r2 = 0·02, P > 0·05] for Bjerkreimselva (Table I) when liming commenced at the start of the anadromous time series. (c) Age 0 years [
: ln y = (0·053 ± 0·020)x − (101·9 ± 39·5), r2 = 0·53, P < 0·01; 
: ln y = (−0·053 ± 0·020)x + (107·8 ± 39·5), r2 = 0·33, P < 0·01] and (d) age ≥ 1 year [
: ln y = (0·034 ± 0·020)x − (66·7 ± 39·5), r2 = 0·34, P > 0·05; 
: ln y = (−0·114 ± 0·020)x + (228·7 ± 39·5), r2 = 0·56, P < 0·001] in Vikedalselva (Table I) when liming started at the initial year of the time series.Changes in density of S. trutta on anadromous stretches after liming
Salmo trutta occurred in all study rivers prior to liming. During the initial phase of the study, both age 0 and age ≥ 1 year fish were caught at most sampling stations. Figs 3 and 4 show changes in their densities in all rivers compared to S. salar. The model best explaining changes in density of age 0 S. trutta included an effect of S. salar status before liming started. The density age 0 year S. trutta decreased by 7% annually in rivers where S. salar was lost prior to liming, corresponding to a change in density by a factor of 0·49 over 10 years, i.e. 51% reduction (Table II and Fig. 5). In rivers where S. salar was present prior to liming, the decrease of age 0 S. trutta density was estimated at 2% annually. This corresponds to a change in density by a factor of 0·81, or 19% reduction over 10 years, with 95% confidence interval overlapping no change in density (Fig. 5).
) and Salmo salar (
) over time since the start of liming in the 13 study rivers in southernmost and south-western Norway. Regression lines shows the predictions of the autoregressive random-regression model: (a) Storelva: 
, ln y = −0·02x + 3·40, r2 = 0; 
, ln y = 0·16x + 2·99, r2 = 0·43; (b) Tovdalselva: 
, ln y = −0·06x + 3·55, r2 = 0; 
, ln y = 0·33x + 0·76, r2 = 0·51; (c) Mandalselva: 
, ln y = −0·08x + 0·44, r2 = 0·44; 
, ln y = 0·36x + 1·29, r2 = 0·67; (d) Lygna: 
, ln y = −0·05x + 3·54, r2 = 0; 
, ln y = 0·32x − 0·21, r2 = 0·71; (e) Kvina: 
, ln y = −0·10x + 2·91, r2 = 0·55; 
, ln y = 0·26x + 1·28, r2 = 0·68; (f) Soknedalselva: 
, ln y = −0·07x + 2·38, r2 = 0·38; 
, ln y = 0·20x + 1·76, r2 = 0·75; (g) Bjerkreimselva: 
, ln y = −0·01x + 3·16, r2 = 0; 
, ln y = 0·35x + 1·70, r2 = 0·72; (h) Ogna: 
, ln y = −0·03x + 1·87, r2 = 0·14; 
, ln y = 0·09x + 3·37, r2 = 0·40; (i) Frafjordelva: 
, ln y = −0·08x + 2·23, r2 = 0·22; 
, ln y = 0·17x + 2·39, r2 = 0·49; (j) Espedalselva: 
, ln y = −0·02x + 1·50, r2 = 0; 
, ln y = 0·16x + 2·47, r2 = 0·56; (k) Jørpelandselva: 
, ln y = −0·003x + 2·08, r2 = 0; 
, ln y = 0·12x + 1·46, r2 = 0·22; (l) Vikedalselva: 
, ln y = −0·04x + 2·78, r2 = 0·29; 
, ln y = 0·11x + 2·62, r2 = 0·33; (m) Rødneelva: 
, ln y = −0·02x + 3·23, r2 = 0·08; 
, ln y = 0·12x + 2·99, r2 = 0·41.
) and S. salar (
) over time since the start of liming in the 13 study rivers in southernmost and southwestern Norway. Regression lines shows the predictions of the autoregressive random-regression model: (a) Storelva: 
, ln y = −0·15x + 0·69, r2 = 0·43; 
, ln y = 0·09x + 1·48, r2 = 0·61; (b) Tovdalselva: 
, ln y = −0·05x + 1·64, r2 = 0; 
, ln y = 0·30x − 0·67, r2 = 0·61; (c) Mandalselva: 
, ln y = −0·07x + 1·17, r2 = 0·08; 
, ln y = 0·19x + 0·98, r2 = 0·46; (d) Lygna: 
, ln y = −0·03x + 1·70, r2 = 0·03; 
, ln y = 0.27x − 1·28, r2 = 0·74; (e) Kvina: 
, ln y = −0·09x + 1·17, r2 = 0·36; 
, ln y = 0·12x + 0·76, r2 = 0·37; (f) Soknedalselva: 
, ln y = −0·10x + 1·87, r2 = 0·38; 
, ln y = 0·15x + 0·87, r2 = 0·42; (g) Bjerkreimselva: 
, ln y = −0·04x + 1·78, r2 = 0; 
, ln y = 0·29x + 0·96, r2 = 0·87; (h) Ogna: 
, ln y = −0·16x + 1·09, r2 = 0·52; 
, ln y = 0·07x + 2·24, r2 = 0·18; (i) Frafjordelva: 
, ln y = −0·09x + 2·28, r2 = 0·46; 
, ln y = 0·20x + 1·44, r2 = 0·67; (j) Espedalselva: 
, ln y = −0·08x + 1·92, r2 = 0·73; 
, ln y = 0·13x + 2·19, r2 = 0·64; (k) Jørpelandselva: 
, ln y = −0·01x + 1·47, r2 = 0; 
, ln y = 0·05x + 1·01, r2 = 0·11; (l) Vikedalselva: 
, ln y = −0·05x + 1·13, r2 = 0·16; 
, ln y = 0·07x + 1·96, r2 = 0·20; (m) Rødneelva: 
, ln y = −0·07x + 2·77, r2 = 0·08; 
, ln y = 0·20 + 1·85, r2 = 0·45.| Models and parameters | Value | S.E. | Units |
|---|---|---|---|
| Model: S. trutta age 0 years | |||
| b | −0·072 | 0·019 | lnN year−1 |
| Δb S. salar present | 0·051 | 0·026 | lnN year−1 |
| Model: S. trutta age ≥ 1 year | |||
| b | −0·075 | 0·02 | lnN year−1 |
| Model: S. salar age 0 years | |||
| b | 0·211 | 0·020 | lnN year−1 |
| Δb discharge | 0·103 | 0·024 | lnN year−1ln(m3 s−1)−1 |
| Model: S. salar age ≥ 1 year | |||
| b | 0·191 | 0·026 | lnN year−1 |
| Δb regulation | −0·090 | 0·050 | lnN year−1 |
| Δb discharge | 0·068 | 0·027 | lnN year−1ln(m3 s−1)−1 |
- b, Change in log population density (log number of fish 100 m−2, denoted lnN) year−1; Δb, the change in b due to either S. salar present prior to liming, changes in discharge away from the mean value of 2·9 ln (m3 s−1), or hydropower regulation.
, Initial density set to the reference value of unit; 
, river flows c. 50 m3 s−1; 
, river flows c. 5 m3 s−1; 
and 
, 95% c.i.); (d) S. salar age ≥ 1 year in absence of and (e) with hydropower regulation (
, S. salar was absent prior to liming; 
, S. salar was present prior to liming; 
and 
, 95% c.i.). Estimates of model parameters are presented in Table II.The model best explaining changes in density of age ≥ 1 S. trutta included only the main effect of time since liming. Their densities decreased by 7% annually (Table II), corresponding to a change in density by a factor of 0·47 (i.e. 53% reduction) over 10 years (Fig. 5).
Changes in density of S. salar after liming
In six out of the 13 studied rivers, S. salar was not present at the initial year of liming (Table I). However, the first year after liming all rivers were populated by S. salar and the increase in density from this point onwards was large. Changes in S. salar densities in all rivers compared with that of S. trutta are shown in Figs 3 and 4.
The model that best explained changes in density of age 0 years S. salar included an effect of water discharge, i.e. river size. Their densities increased by 23% annually at discharge 18·2 m3 s−1 (Table II), corresponding to a factor of 8·2 over ten years, i.e. 720% increase. There was a strong effect of rivers discharge in non-regulated rivers. At an average discharge of 5·0 and 50·0 m3 s−1 the increase in density was estimated at 8% and 37% annually, respectively (Table II). These values corresponds to an increase in density by factors of 2·1 and 23·4 over 10 years, respectively (Fig. 5).
The best model explaining changes in density of age ≥ 1 year S. salar included effects of regulation and river discharge. This model was only 1·34 AICc units better than a model without effect of regulation (Table S1, Supporting Information), and the effect of regulation had high uncertainty (Table II). The density age ≥ 1 year S. salar increased by 21% annually (Table II), corresponding to a factor of 6·7 over 10 years, in non-regulated rivers at discharge 18·2 m3 s−1. In regulated rivers, the increase was estimated at 11% annually at the same discharge (Table II). There was a strong effect of river flow in non-regulated rivers. At an average flow of 5·0 m3 s−1, the increase in density was estimated at 11% annually while at a discharge of 50 m3 s−1, it was 30% (Table II). These values correspond to an increase in density by a factor of 2·8 and 13·4 over 10 years, respectively (Fig. 5).
Years with good conditions are expected to be beneficial for both S. trutta and S salar densities. This is supported in that the residuals of both species models were positively correlated and estimated at 0·26 and 0·42 for age 0 and age ≥ 1 year, respectively. This was also apparent between the two age classes within each species with correlations estimated at 0·22 and 0·37 for S. trutta and S. salar, respectively.
Discussion
To conclusively demonstrate the effect of environmental perturbation on an ecosystem or populations, a proper before-after-control-impact design of the investigation is required (Smith, 2013). Nevertheless, long-term data series such as presented here, are useful for showing and separating factors causing environmental effects (Elliott, 1994). Furthermore, because no encroachment or other environmental changes have taken place in the study rivers during recent years, the present study functions as a natural experiment revealing differences in environmental effects on the species.
The results confirm that liming is well suited for enhancing abundances of some freshwater fishes in acidified rivers, such as young S. salar (Hesthagen et al., 2011). It also works well for other fish species, such as A. anguilla (Larsen et al., 2015). This was not the case, however, for S. trutta in sympatry with S. salar in the same rivers. Their densities decreased with increasing total S. salar densities as well as S. salar status when liming started, i.e. lost or reduced populations. These findings lend support to the first hypothesis that densities of S. trutta are negatively affected by increased abundance of young S. salar following liming. Density regulation of S. trutta chiefly occurs early in life (Elliott, 1994), and they typically expand their use of fast-flowing river stretches in absence of S. salar (Karlström, 1977; Gibson & Cunjak, 1986).
Spatial segregation of S. trutta and young S. salar in rivers is used as proof of interspecific competition between the two species (Jonsson & Jonsson, 2011). Salmo trutta are territorial and very aggressive, and through interference, they can constrain S. salar from shallow, slow flowing water. If S. trutta are removed, young S. salar expand their habitat into slow-flowing areas, i.e. competitive release (Kennedy & Strange, 1986; Hearn, 1987). In stream tank experiments, Gibson & Erkinaro (2009) showed that S. trutta were four times more aggressive than corresponding S. salar. This strategy gives S. trutta a competitive advantage in quiet, shallow areas along stream banks where they more easily can monopolize food resources relative to similar sized S. salar. Thus, in the presence of S. trutta, S. salar make more use of cover, deep pools and fast-flowing riffles where food is more difficult to defend. In addition, the more streamlined body and larger pectoral fins are assumed to give S. salar a selective advantage in rapidly flowing areas of rivers (Karlström, 1977). The high aggressiveness of S. trutta, however, is energetically costly, and their high energy use is assumed to be the main reason why they are outcompeted by Arctic charr Salvelinus alpinus (L. 1758) in food limited habitats (Finstad et al., 2011). The present study indicates that S. trutta pay a higher cost of competing with S. salar after liming, than the cost they paid when living in the acidified water prior to this mitigation measure. This is assumed because S. trutta density decreased when S. salar densities increased following improved water quality due to liming.
Furthermore, the decrease in S. trutta density was stronger in rivers where S. salar were lost pre-liming, than in rivers where S. salar were present. This is probably because there was no interspecific competition between these two species when the liming started, and the effect was consequently stronger when S. salar returned after liming commenced. Salmo salar responded rapidly to the improved water quality after liming, and presumably imposed immediate and strong competition on S. trutta. These findings lend support to the second hypothesis.
In two of 22 investigated rivers in Sweden, S. trutta abundances in sympatry with S. salar decreased after liming (Degerman & Appelberg, 1992). Furthermore, data from limed rivers in U.K., Norway, Sweden, U.S.A. and Canada reviewed by Mant et al. (2013), indicated that S. salar increased to a greater extent than S. trutta in all except one of 19 limed rivers. Thus, the present results may be general and a possible consequence of interspecific resource competition between the two salmonid species.
There was little or no effect of river size on S. trutta density. Thus, no support was found for the third hypothesis. However, S. salar increased more in larger than in smaller rivers. This indicates that interspecific competition is keener for this species in small rivers, possibly because of more frequent encounters and salmon is the weaker competitor where resources easily can be defended, such as in small streams. In large rivers, young S. salar avoid the strong competition from territorial S. trutta by moving offshore. Thus, the outcome of the competition appeared habitat dependent, being more severe for S. salar in smaller than in larger rivers.
The effect of river regulation was weak on S. salar if present at all, and no effect was observed on S. trutta. Almodóvar & Nicola (1999), however, found that the density of S. trutta decreased after regulation of the River Hoz Seca, Spain. The water level in regulated rivers may at times be very low, and such conditions may make the habitat less suitable for older and larger individuals (cf. Heggenes et al., 1999).
In the unlimed control stretches in two catchments with no S. salar, densities of young S. trutta either remained unchanged or increased during the study period. This is probably related to a general improved water quality in southern Norwegian rivers during recent years as a consequence of major reductions in sulphur and nitrogen depositions in Europe (Tørseth et al., 2012; Garmo et al., 2014; Hesthagen et al., 2016).
Liming does not influence fish only, but entire ecosystems (Rundle et al., 1995). In the past, concerns about effects of liming have been raised with respect to effects on conservation value of ecosystems that may be naturally low in cations (Mant et al., 2013). This mitigation measure can increase abundance and richness of acid sensitive invertebrates, e.g. used by salmonids for food. This is reported from U.K. (Ormerod et al., 1990; Bradley & Ormerod, 2002), Sweden (McKie et al., 2006) and Norway (Raddum & Fjellheim, 1992, 2003). Already, two years after liming of the River Audna in Norway, several acidity sensitive invertebrates responded positively and in the following five years they recolonized the whole limed reach of the river (Raddum & Fjellheim, 2003). A recent meta-analysis, however, found that benefits for invertebrates may be variable and not guaranteed in all rivers (Mant et al., 2013). In some cases, liming can even decrease the overall invertebrate abundance and biodiversity and change assemblage structure (D'Amico et al., 2004). For instance, liming of north Swedish rivers reduced abundance of large, acid-tolerant shredding caddisfly larvae, while the abundance of small shredding stoneflies increased (McKie et al., 2006), possibly because of interspecific competition.
The present study exhibits how water mitigation can affect salmonid fish populations in rivers. The liming allows re-establishment of S. salar, but at the same time, abundance of S. trutta decreased probably because of interspecific resource competition with invading S. salar. Possibly, densities of S. trutta in limed rivers will with time become similar to what they were before acidification started. Although the water quality of south Norwegian rivers has improved in recent years, it is still inadequate for survival of young S. salar in unlimed stretches of the rivers that lost their populations towards the end of the 20th century (Kroglund et al., 2008). Thus, to sustain S. salar, liming is still continued to maintain a suitable water quality for this species.
The study was financed by the former Directorate for Nature Management (now a part of the Norwegian Environment Agency), and the Norwegian Institute for Nature Research (NINA). We are grateful to H. M. Berger for his indispensable contribution during the electrofishing surveys. Three anonymous referees made valuable comments to the manuscript.
