Hydrobiologia

, Volume 652, Issue 1, pp 299–310

Mesohabitat use by bullhead (Cottus gobio)

Authors

    • School of Geography, Earth and Environmental SciencesUniversity of Birmingham
    • Institute of Science and the EnvironmentUniversity of Worcester, Henwick Grove
    • Department of BiologyKarlstad University
  • G. E. Petts
    • University of Westminster
  • I. P. Maddock
    • Institute of Science and the EnvironmentUniversity of Worcester, Henwick Grove
Primary research paper

DOI: 10.1007/s10750-010-0363-z

Cite this article as:
Gosselin, M., Petts, G.E. & Maddock, I.P. Hydrobiologia (2010) 652: 299. doi:10.1007/s10750-010-0363-z

Abstract

Habitat composition and connectivity within a stream vary with changing flows but the influence of changing flow on habitat use by fish is not well understood. Meso- and microhabitat surveys were used to investigate habitat use by bullhead (Cottus gobio Linnaeus) in response to discharge variation in a small tributary of the Upper Severn, England. Mesohabitat mapping surveys were carried out over a range of summer flows (0.016–0.216 m3 s−1) and were coupled with direct underwater observations (snorkelling) of fish location. Five mesohabitat types—glides, runs, riffles, chutes and pools—were present in the reach at all flows surveyed and ‘backwaters’ were found at three flows. The macro-morphology of the reach comprised six riffle–pool sequences divided into 27 mesohabitats with the maximum diversity (23 mesohabitats) at intermediate flows (Q43) and only 15 mesohabitats at Q95. Despite low numbers of fish (N = 78), bullhead displayed a strong association (51% of the fish) with glides—relatively deep habitats having high rates of velocity increase with flow. However, 54% of the fish were observed in two large, persistent mesohabitats, a glide (34%) and a pool (20%), both located below a faster flowing mesohabitat. Habitat use curves based upon micro-habitat data showed bullhead favoured low velocities (<0.30 m s−1), depths less than 0.30 m and a cobble substratum. This study illustrates the value of cross-scale investigations in linking fish ecology, flow and physical habitat variability and suggests mesohabitat size, persistence and arrangement may influence fish distribution.

Keywords

Flow variabilityHabitat compositionMesohabitat surveysCottus gobioHabitat use curves

Introduction

Many species are adapted to the natural flow regime (Poff et al., 1997; Lytle & Poff, 2004) and the temporal dynamics of habitat quantity may be a major factor determining fish population responses in riverine environments (Stalnaker et al., 1996). Over the past three decades there has been a rapid growth of research on environmental flows but the challenge to relate habitat use to changing flows remains elusive (Petts, 2009). There is limited evidence of different patterns of habitat use and the large number of empirical case studies have been unable to develop general relationships (Poff & Zimmerman, 2010). Progress in developing models that link physical habitat dynamics and population biology of large organisms such as fish may has been constrained by the difficulty in merging the space- and time-scales appropriate to both physical and biological sciences (Petts et al., 2006). Yet, progress is particularly necessary in the context of the European Community’s Water Framework Directive, which requires monitoring of water bodies to achieve good ecological status by 2015.

A template for examining habitat use by biota that has been widely employed over the past decade is mesohabitat classification. The terms used to describe these habitat units have varied between authors and include ‘mesohabitats’ (e.g. Tickner et al., 2000), ‘channel geomorphic units’ (CGUs) (e.g. Hawkins et al., 1993), ‘physical biotopes’ (e.g. Padmore, 1997) and ‘hydraulic biotopes’ (e.g. Wadeson, 1994). Each mesohabitat is a definable area such as a pool, riffle or run that can be inferred by visual observation of surface flow character and verified by hydraulic measurements and qualitative or quantitative substratum types (Armitage et al., 1995; Newson & Newson, 2000). Attempts to argue the biological significance of meso-scale hydraulic habitat surveys appear premature (Petts, 2009) but the attractiveness of the mesohabitat approach for managers is its practicality (Newson et al., 1998). The suitability of the meso-scale for the study of fish ecology was emphasised by Fausch et al. (2002), who stated that important features, such as obstacles to fish movements, were best seen at this scale. They also argued that river habitat assessment should concentrate on assessing complete reaches at the meso-spatial scale in order to recognise the river landscape as a spatially continuous longitudinal and lateral mosaic of habitats. In response to this, habitat mapping over continuous reaches is becoming more common, but examining the dynamics of mesohabitats with changes in discharge along continuous reaches is still rare (Maddock et al., 2008).

Most studies on fish–habitat relationships have focused on salmonids because of their economic importance and ubiquity. This study focuses on the bullhead (Cottus gobio), a small bottom-dwelling fish that is widespread in the rivers and streams of England and Wales. Bullhead occurrence is considered to be a useful indicator of the health, integrity and naturalness of running waters (Tomlinson & Perrow, 2003) and the species is endangered in several countries of continental Europe (e.g. Belgium, Knaepkens et al., 2004a, b) as a result of habitat degradation. Bullhead life cycle, and in particular the development of young bullhead have been described by Fox (1978). Several studies have focused on different aspects of bullhead ecology such as movement behaviour (Downhower et al., 1990; Fischer & Kummer, 2000; Knaepkens et al., 2004a, b) and habitat preferences (Perrow et al., 1997; Knaepkens et al., 2002; Carter et al., 2004; Legalle et al., 2005). These show that habitat use and preference by bullhead differ between sites and studies. For example, depth preferences have been found to vary from 0.05 m (Legalle et al., 2004) to 0.40 m (Roussel & Bardonnet, 1996); velocities range from 0.10 m s−1 (Carter et al., 2004) to 1 m s−1 (Knaepkens et al., 2002). Most studies agree that bullhead prefer gravel, cobble, pebble and boulder beds. Mesohabitat use by bullhead has not been considered in previous studies, although riffles have been mentioned as the preferred habitat with low depth, high velocity and coarse substrate (Langford & Hawkins, 1997; Perrow et al., 1997). The aim of this paper is to gain further insight into mesohabitat dynamics with flow and its relationship with bullhead distribution. Three questions are addressed: (i) How does mesohabitat composition vary with flow in the Dowles Brook? (ii) How do flow variations and the consequent mesohabitat dynamics influence bullhead distribution? (iii) What are bullhead habitat preferences as defined by depth, velocity and substrate?

Methods

Study area

The Dowles Brook, a 40 km2 catchment within the Wyre forest in Worcestershire (Fig. 1) and a tributary of the Upper Severn, was selected for study because this clean stream flows through a Special Site of Scientific Interest and presents a population of bullhead. The catchment is underlain by carboniferous sandstone and marls. Average annual rainfall is 728 mm. The study site was a 200 m reach (about 40 channel widths) located 500 m above a gauging station and within a Nature Reserve owned by the Worcestershire Wildlife Trust. The mean flow is 0.39 m3 s−1, peak flow is 21.60 m3 s−1, Q95 is 0.03 m3 s−1 and the Q10/Q95 ratio is 33 reflecting the flashy flow regime. The channel has a natural form with an average width through the study reach of 5.5 m and a gradient of 1.56 m km−1. The riparian zone comprises woodland and meadow.
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0363-z/MediaObjects/10750_2010_363_Fig1_HTML.gif
Fig. 1

Map of the Dowles Brook study reach at the lowest flow surveyed showing location of CGUs and, insert, location of the study reach

Scale of study and mesohabitat surveys

Mesohabitats, or CGUs, were mapped over a range of eight flows between April and October 2006 (Table 1), following the bullhead spawning season (March–April). The CGUs were identified using the association between geomorphology and surface flow types (Newson et al., 1998) as described by Parasiewicz (2001). The range of mesohabitats in this study included: chute, riffle, run, glide, pool and backwater on a scale from rapid flow to imperceptible flow (Table 2). Using a scale from deep water to shallow water, the sequence would be: pool, glide, backwater, chute, run and riffle. Depth, velocity and substrate composition were recorded for each CGU. Substrate categories were identified visually according to particle size as: silt (up to 64 μm in diameter), sand (between 64 μm and 2 mm), gravel (2 mm to 4 cm), pebble (4–6 cm), cobble (6–25 cm), boulder (above 25 cm). Within each CGU, depth and velocity (0.6 depth in metre) measurements were taken at five points arranged in a cross pattern within the core of each CGU, estimated visually, to avoid transitional effects. Most prior studies have used transects across a stream to obtain hydraulic data, with the number of transects in each mesohabitat determined by habitat availability. However, this supposes that mesohabitats form a simple sequence, occupying the whole cross-section with boundaries perpendicular to the flow. In reality, mesohabitats may vary both along and across a channel. Thus, a cross-pattern approach was adopted here and it also allowed rapid and easily replicable habitat mapping across the range of flows. Spacing between the measurement points depended on the size of the CGU considered, from 10 cm apart for a very small mesohabitat to several metres for the largest units. The five points of measurement constituted an appropriate trade-off between the need for accuracy and representation of the mesohabitat conditions and the replication of this method during surveys. In addition, at each fish location micro-habitat (point) data (water depth, velocity and substrate type) were recorded to allow the construction of habitat use curves for comparison with mesohabitat data and other studies.
Table 1

Distribution of CGUs in the Dowles Brook with changing flow (expressed as flow percentiles)

Q (%’ile)

Q99 (August)

Q96 (September)

Q95 (July)

Q72 (October)

Q56 (June)

Q43 (May)

Q38 (April)

Q35 (May)

Q (m3 s−1)

0.016

0.021

0.030

0.054

0.101

0.143

0.198

0.216

1

Riffle

Riffle

Riffle

Riffle

Run

Riffle

Riffle

Riffle

2

      

Run

Run

3

Glide

Glide

Glide

Glide

Glide

Glide

Glide

Glide

4

Riffle

Run

Run

Run

Riffle

Run

Run

Run

5

Run

   

Run

Glide

  

6

Riffle

   

Riffle

Riffle

Riffle

Riffle

7

    

Run

Run

  

8

Glide

Glide

Glide

Glide

Glide

Riffle

Glide

Glide

9

    

Run

Run

 

Run

10

    

Chute

Glide

 

Run

11

Pool

Pool

Backwater

Pool

Pool

Pool

Pool

Pool

12

Pool

  

Backwater

  

Pool

 

13

Run

Riffle

Riffle

Run

Run

Run

Run

Glide

14

Pool

Pool

Pool

Pool

Pool

 

Backwater

 

15

Chute

Chute

Chute

Chute

Chute

Chute

Chute

 

16

Glide

Glide

Glide

Glide

Run

Run

Run

 

17

      

Glide

 

18

Riffle

Riffle

Riffle

Riffle

 

Chute

Riffle

Riffle

19

Glide

Glide

Glide

Run

Glide

Glide

Glide

Glide

20

Run

Riffle

Riffle

 

Run

Riffle

Run

Riffle

21

     

Glide

  

22

     

Riffle

  

23

     

Run

  

24

Glide

Pool

Pool

Glide

Glide

Glide

Glide

Glide

25

Chute

   

Chute

Chute

Chute

Chute

26

Run

Riffle

Riffle

Run

Run

Run

Riffle

Riffle

27

Pool

Pool

Pool

Pool

Pool

Pool

Pool

Pool

NCGU

19

15

15

15

20

23

20

17

The units are numbered from the downstream end of the reach onwards (Fig. 1). Locations with fish observations are given in italics

Table 2

Criteria used to identify the mesohabitats encountered in the Dowles Brook, according to the MesoHABSIM method (Northeast Instream Habitat Program, 2007)

Mesohabitat (CGU)

Associated flow type

Turbulence

Brief description

Riffle

Unbroken standing waves

Turbulent and moderately fast

The most common type of turbulent fast water mesohabitats in low gradient alluvial channels. Substrate is finer (usually gravel) than other fast water turbulent mesohabitats, and there is less white water, with some substrate breaking the surface

Run

Rippled

Non-turbulent and Moderately fast

Moderately fast and shallow gradient with ripples on the surface of the water. Deeper than riffles with little if any substrate breaking the surface

Glide

Smooth boundary turbulent

Non turbulent and moderately slow

Smooth ‘glass-like’ surface with visible flow movement along the surface, relatively shallow (compared to pools) depths

Pool

Scarcely perceptible flow

Non turbulent and slow

Relatively deep and slow flowing, with fine substrate. Usually little surface water movement visible. Can be bounded by shallows (riffles, runs) at the upstream and downstream ends

Backwater

Scarcely perceptible flow

Non-turbulent and slow

Water is ponded back upstream by an obstruction, e.g. weir, dam, sluice gate, etc.

Chute

Chute/broken standing waves

Turbulent and fast

Water passes over a break or step in the substrate

The method and nomenclature were simplified to be used in this study

Fish observations

Data on fish occurrence throughout the reach were recorded during five of the mesohabitat surveys using direct underwater observations (snorkelling) as recommended in Heggenes & Saltveit (1990). Flows during the three other surveys were too turbid to allow visual observation of fish. Snorkelling as a fish survey method has often been criticised because it underestimates fish numbers. Nonetheless the authors believe it was the most appropriate technique for this study as it allowed fish distribution to be related to both meso- and microhabitats. Starting at the downstream end of the reach, the surveyor would snorkel upstream in a zig-zag manner so that the probability of fish observation was even throughout the reach. Since the bullhead is a benthic species and is known to hide under coarse substrate particles, cobbles, gravel and pebbles were disturbed to look for fish as the surveyor progressed upstream. Once a fish was observed, a weighted float marked with a number was left at the location of the observation. This allowed the subsequent recording of the microhabitat variables: depth, velocity and substrate. Fish length was estimated visually.

Data analyses

Mesohabitat maps of the reach were produced for each survey and these were compared in order to determine the dynamics of the mesohabitats with flow. Hydraulic geometry relationships between depth and velocity and flow were described by linear regressions, following log10 transformation to meet assumptions of regression analyses, for each mesohabitat. The associations of mesohabitats with bullhead distributions were explored using a range of non-parametric tests (Kruskal–Wallis, Chi Square and Mann–Whitney) as appropriate. Point measurements of depth, velocity and substrate at each bullhead location were used to build normalised microhabitat use curves following Harby et al. (2004). These were compared with microhabitat availability determined by normalising the frequency of depth, velocity and substrate occurrence among all the mesohabitats surveyed.

Results

Mesohabitat structure and dynamics

Mesohabitat surveys were carried out during eight different flows ranging from 0.216 m3 s−1, the 35th percentile flow (Q35) to 0.016 m3 s−1 (Q99) to observe the changing pattern of CGUs with flow. The 200 m reach shows a macro-scale geomorphological structure with six dominant riffle–pool sequences having an average spacing of six times the channel width, typical of alluvial rivers. At this scale, the number of CGUs is 12. However, at the meso-scale and under low to medium flows, a total of 27 CGUs were identified along the reach (Fig. 1), with greatest differentiation, i.e. the largest number of CGUs, at Q43 (Table 1). The CGUs were classified and ranked by channel area as: glides (44%), riffles (21%), runs (18%), pools (13%) and chutes (3%). ‘Backwater’ was recorded in three surveys (Q38, Q72 and Q95) and was located in CGU 14 and 11/12, respectively.

Two large CGUs (3 and 27) persisted throughout the range of flows. Others (1, 5, 8, 11, 15 and 19) varied in type only once across the eight surveys. The main changes in CGUs between surveys were riffle–run (9 CGUs) and run–glide (5 CGUs). At Q72, unit 4 (run) extended to include units 3 (riffle), 5 (riffle) and 6 (glide) increasing the area of run within the reach to 30% and reducing the areas of glide and riffle to 37% and 17%, respectively. At flows above Q43, the pattern of CGUs simplifies and approaches the macro-scale structure of the reach. For example, the complex sequence of small units between CGU 11 and CGU 17 at low flow is drowned at about 0.20 m3 s−1 (Q35) becoming pool–glide CGUs.

The dominant CGUs comprise three groups: (i) pools and glides that are relatively deep and slow flowing, (ii) riffles and runs that are shallow with moderate flow and (iii) chutes with shallow flow and high velocities. These groups have distinctive hydraulic characteristics not only at low flow but also with increasing flows (Table 3). Significant hydraulic geometry relationships (ANOVA, P < 0.001) were found for depth and velocity in all CGUs except chutes where no significant relationship was found for depth and for velocity (ANOVA, P = 0.071 and P = 0.399, respectively, for a critical P value of 0.05).
Table 3

Changing patterns of velocities and depths within the CGUs

CGU

N

Substrate

Hydraulic variable

Mean (std. dev.)

Regression exponent

Regression constant

R2

Chutes

25

Bedrock

Velocity

0.65 (0.280)

0.109

−0.097

0.19

Depth

0.14 (0.087)

0.266

−0.683

0.37

Riffles

126

Gravel, cobble, bedrock

Velocity

0.29 (0.175)

0.319

−0.266

0.79

Depth

0.11 (0.046)

0.288

−0.701

0.54

Runs

162

Gravel, cobble, bedrock

Velocity

0.26 (0.202)

0.244

−0.439

0.27

Depth

0.15 (0.073)

0.256

−0.628

0.43

Glides

226

Cobble, silt, bedrock

Velocity

0.09 (0.091)

0.461

−0.646

0.94

Depth

0.27 (0.101)

0.143

−0.456

0.85

Pools

83

Silt, cobble, bedrock

Velocity

0.02 (0.036)

0.188

−0.389

0.91

Depth

0.30 (0.160)

0.169

−1.573

0.64

Hydraulic geometry relationships were based on log10 transformed data of the hydraulic variable and discharge

As discharge increased, glides showed a rapid increase in velocity with mean velocities exceeding about 0.10 m s−1 for approximately 50% of the time compared with velocities of less than about 0.05 m s−1 during the lowest 10% of flows. At Q35, at some points within glides, mean velocity exceeded 0.25 m s−1. Average depths within glides were always above 0.20 m and increased slowly to ca. 0.32 m at Q35. In contrast, mean velocities through pools were below 0.05 m s−1 across the range of flows and depths increased only very slowly with discharge from about 0.25 m at the lowest flows to deeper than 0.30 m at flows above Q40. Shallow mesohabitats showed rates of velocity change with discharge that were intermediate between glides and pools but depths increased rapidly. At riffles, mean velocity increased from about 0.20 to 0.35 m s−1 and depths from 0.05 to 0.16 m over the range of flows surveyed. In runs, mean velocities increased from less than 0.10 m s−1 to more than 0.20 m s−1 and depths from less than 0.15 m to about 0.30 m.

Bullhead distribution and population structure

Five fish surveys of the whole reach were carried out over five flows and coupled with mesohabitat surveys between May and October 2006 (Table 4), when flows ranged from Q43 (May) to Q99 (August). Bullhead were observed on every occasion and were the only species observed in the stream. Seventy-eight fish were recorded over the five surveys but the number of observations during each survey varied from 4 fish in May (Q43) to 22 fish in September (Q96), an average of 15.8 per survey, or one fish per 13.9 m2.
Table 4

Bullhead occurrences in relation to flow, CGU and micro-habitat

Flow

Month

CGU (see Fig. 1)

CGU type

Bullhead observations

Mean velocity (m s−1)

Mean depth (m)

Substrate

Q43

May

1–2

Riffle

3

0.15

0.05

Cobble, gravel

5

Glide

1

0.06

0.10

Cobble

Q95

July

3

Glide

5

0.04

0.06

Cobble, gravel

4–7

Run

1

0.17

0.10

Cobble

8–10

Glide

1

0.07

0.06

Sand, pebble

11–12

Backwater

1

0.00

0.08

Cobble, sand

27

Pool

8

0.20

0.22

Cobble, sand

Q99

August

3

Glide

5

0.03

0.17

Cobble, gravel

5

Run

3

0.03

0.10

Cobble

19

Glide

7

0.06

0.20

Pebble, cobble

20–23

Run

1

0.00

0.13

Cobble, gravel

27

Pool

2

0.01

0.21

Cobble, gravel

Q96

September

3

Glide

10

0.04

0.17

Cobble, boulder

4–7

Run

1

0.27

0.15

Cobble

12

Glide

1

0.055

0.15

Cobble

13

Riffle

1

0.06

0.2

Cobble

14

Pool

1

0.02

0.18

Cobble, gravel

16–17

Glide

1

0.06

0.15

Cobble

26

Riffle

3

0.13

0.18

Cobble

27

Pool

4

0.00

0.26

Cobble, gravel

Q72

October

3

Glide

6

0.11

0.16

Cobble, gravel

4–7

Run

4

0.40

0.14

Gravel, cobble

8–10

Glide

2

0.06

0.11

Cobble

13

Run

2

0.14

0.13

Cobble, gravel

16–17

Glide

1

0.09

0.04

Cobble, gravel

19–23

Run

1

0.00

0.03

Cobble

27

Pool

2

0.06

0.23

Cobble, gravel

In all surveys, 51% of the bullhead (N = 78) were recorded in glides with the use of this CGU reaching a maximum of 66% in August (N = 18). Overall, 22% of the fish were observed in pools and 17% in runs. Less than 10% (only seven fish) were found in riffles. None were observed in chutes.

Over the 5-month survey, bullheads were observed in 12 of the units (Table 4). In 10 of the units the number of observations was less than 10. However, in 2 units numbers were much higher: 16 in the pool unit at the upstream end of the reach (unit 27) and 26 in the glide at the downstream end of the reach (unit 3). They are both large units (ca. 10% of the reach area) and are persistent across the range of flows. They are both deep areas compared to other parts of the reach. In the downstream glide, depth varied between 0.17 and 0.28 m and in the upstream pool, depth varied from 0.29 to 0.45 m. They are also slow flowing environments: velocity in the downstream glide constantly remained below 0.30 m s−1 and, in the pool, remained under 0.09 m s−1 across the range of flows. The data suggest a strong association between bullhead distribution and glides across the range of low flows (Kruskal–Wallis, P < 0.001). The Chi Square test on bullhead abundance and glide frequency of occurrence was not significant (Chi-square, P = 0.322), indicating that the strong association between bullhead and glides is not related to the higher frequency of glides in the reach compared to other habitat types. Below Q95, when flow depth may begin to become limiting for some lotic species, glides offer relatively deep habitats with low but detectable velocities.

Bullhead were divided into three classes according to fish size based on information gathered from the literature (Fox, 1978; Cowx & Harvey, 2003). Fish ranged from 2 to 15 cm in length. Hence the three classes were:
  • less than 5 cm: juvenile and adult-but-not-mature individuals (N = 35).

  • from 5 to 10 cm: adults of average size (N = 36).

  • greater than 10 cm: large adults (N = 7).

Figure 2 shows the change in length frequency distribution of bullhead throughout the survey season. From May, the number of small sized bullhead (length less than 5 cm) increased to a maximum of 65% of the observations in August and then steadily decreased. At the same time the proportion of average sized bullhead decreased from May to a minimum in August (35% of the observed population). Large bullhead were observed in small numbers in July, September and October.
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0363-z/MediaObjects/10750_2010_363_Fig2_HTML.gif
Fig. 2

Variation of the length frequency distribution of observed bullheads from May to October

Microhabitat analyses

Records of depth, velocity and substrate at each bullhead location (N = 78) allowed habitat use (or frequency of use) curves (Harby et al., 2004) to be constructed according to habitat availability in the stream (Fig. 3). Depths most frequently used by bullhead were those between 0.05 and 0.20 m. Depths above 0.30 m were not used at all except at Q99 when one large individual was found in depths of 0.40 m. Velocities below 0.10 m s−1 were the most frequently used and few fish were observed where velocities exceeded 0.30 m s−1. Comparison of the frequency distributions of habitat availability and habitat use showed that while depth use by bullhead followed the distribution of depths in the stream (Mann–Whitney test, P = 0.341), the fish displayed choice in their use of velocity as the frequency distribution of velocity use was significantly different from the frequency of occurrence of velocities (Mann–Whitney test, P = 0.036). With respect to the highest flow surveyed (Q43), all four individuals were observed in shallow water at depths less than 0.10 m and two were associated with relatively high velocities of about 0.40 m s−1. From the substrate use curve it can be seen that bullhead displayed a strong association with cobbles, which are coarse enough to provide shelter. Underwater observations showed that the presence of finer sediment, such as sand and silt, with cobbles did not prevent the fish from using these mesohabitats.
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0363-z/MediaObjects/10750_2010_363_Fig3_HTML.gif
Fig. 3

Habitat use curves for bullhead in the Dowles Brook. (A depth, B velocity and C substrate)

Discussion

The data presented herein give an insight into the dynamic character of the mesohabitats within the Dowles Brook. The analysis of mesohabitat physical characteristics agrees with the description made of these CGUs in MesoHabsim (Parasiewicz, 2007) and also the River Habitat Survey (Environment Agency, 2003). Riffles and runs (shallower and faster flowing mesohabitats) are clearly differentiated from pools and glides (slow flowing and deeper mesohabitats). Each CGU is also shown to have a distinctive hydraulic geometry, describing the rate of change of depth and velocity with flow. Thus, glides differentiate from pools and to a lesser degree, riffles differentiate from runs, by the relatively rapid rise in velocity with increasing discharge. The range of depths and velocities is similar to that measured in the Leigh Brook, Worcestershire (Maddock & Lander, 2002), another low-gradient stream within the Severn catchment that has a similar flow regime to the Dowles Brook. However, pools in the latter are relatively shallow compared to the Leigh Brook, where pool depth reached 0.94 m at Q82 (0.517 m3 s−1). In the Dowles Brook, the five main types of mesohabitats identified (riffle, run, glide, pool and chute) were present at all flows but their persistence varied. Glides and pools were most persistent over the range of flows below the median and this agrees with the findings of Maddock et al. (2008) on the Soča River, Slovenia.

Snorkelling led to observations of 16–22 fish on each of four surveys (July to October) but of only four in the first survey (May) when turbid water made the observations more difficult and may have prevented more fish being observed, particularly in deeper areas. The rise in the number of small bullhead in July and August could be the result of the larval stages becoming sedentary (Fox, 1978), spawning taking place usually in March–April. The rise could also have resulted from the migration into the reach, either passive or active, of young bullhead. The decrease in the number of small bullhead in September and October may have resulted either from the growth of these individuals so that they became accounted for in the ‘average size’ class, or from migration of these individuals to other parts of the river outside the study reach.

The number of bullhead observations and their density in the reach (0.07 fish m−2) are low compared to what could be expected from a headwater population. Indeed, Perrow et al. (1997) discussed the densities of bullhead in the headwaters of some Norfolk rivers and defined low densities as <0.15 individuals m−2 and high densities as >0.6 individuals m−2. The extremely low density of bullhead in this study could be the consequence of two main factors. First is the effect of siltation smothering macroinvertebrates and limiting food resources. Indeed silt was present throughout the study reach and more so in the upstream pool where, however, fish were observed. Gammarus sp., bullhead principal source of food (Tomlinson & Perrow, 2003), was also abundant under coarse substrate particles. Second, Perrow et al. (1997) noted the importance of woody debris as a chosen habitat by bullhead and the lack of woody debris habitats in the Dowles Brook may be significant in limiting the availability of persistent pools/glides fed by chute flow below debris jams.

Bullhead distributions in the Dowles Brook showed a strong association with glides and their use increased with decreasing discharge but was not related simply to mesohabitat availability. The physical characteristics of glides (deep, slow flowing environment) and the presence of cobble substratum may explain why this type of mesohabitat is the most used by bullhead (Knaepkens et al., 2004a, b). Water depth and the cobble substratum provide shelter from predators and the rough bed and slow velocities provide food retention. Cobble has already been showed to be a key predictor of bullhead occurrence in rivers (Knaepkens et al., 2002). Indeed, deeper habitats and coarse substrate particles facilitate organic matter retention (Lamberti et al., 1989; Hoover et al., 2006), which provides a reliable source of food for the macroinvertebrates (Rempel et al., 2000) on which bullhead feed, particularly Gammarus sp. (Dahl & Greenberg, 1996).

From the hydraulic geometry characteristics the contrast between pools and glides is evident. Glides showed stable moderate depths across the range of flows observed and relatively high rates of velocity variation with discharge, although mean velocities were very low. Pools were the least variable mesohabitats with very low velocities and moderate depths. Riffles and runs had significantly lower depths and faster velocities, and at riffles velocities increased relatively rapidly with increasing discharge. Newson et al. (1998) showed that pools, backwaters and to lesser extent glides are habitats influenced by depositional processes, whereas riffles and runs are erosional units. Thus, it may be proposed that, in response to high flow and mesohabitat variability, bullhead tend to choose those habitats that are relatively deep across the range of low flows with cobble substrate providing cover from sight-feeding predators (Knaepkens et al., 2002; Cowx & Harvey, 2003) and sites that are relatively stable to minimise the energy expenditure associated with the stress of a constantly varying environment (Webb et al., 1996). Mesohabitat use by bullhead may be more influenced by flow than by season-dependent factors, such as temperature, or life stage. Territoriality may have played a role in determining the locations at which bullhead were found (Roni, 2002; Davey et al., 2005; Petty & Grossman, 2007) with large individuals always in ‘low energy’ mesohabitats and smaller individuals using both low and high energy areas, where they are able to seek refuge from the current in the lee of cobbles.

The general habitat use curves showed a clear preference for depths in the range from 0.10 to 0.30 m and for velocities between 0 and 0.20 m s−1, which agrees with the observations by Carter et al. (2004) in the River Avon (depth use between 0.10 and 0.20 m). The micro-habitat substrate use curve showed a clear preference for cobbles. These results agree with those of Knaepkens et al. (2004a, b) in that cobbles and coarse substrate particles in general can be used as a predictor of bullhead occurrence. Bullhead were not found in all the pools and glides present in the stream but in those containing large substrate particles, particularly cobble, as a dominant substrate.

The lack of association of bullhead with riffles conflicts with observations by Roussel & Bardonnet (1996), Langford & Hawkins (1997) and Legalle et al. (2005), who found bullhead associated with the low depth, high velocity environments of riffles. However, these rivers were larger than the Dowles Brook and riffle habitats had depths ranging from 0.15 to 0.40 m. Moreover, Perrow et al. (1997) observed a strong association between bullhead distribution, water depth and leaf litter, which correlates with our results showing a strong association between bullhead and higher depths and slow velocities. Similarly, Knaepkens et al. (2004a) observed a positive correlation between bullhead location, coarse substrate and increased depth in natural parts of lowland rivers. The nature of mesohabitat is important but so too is micro-habitat, which explains why even in riffles the velocities at which bullhead were found in this study were low (Table 4). Velocity values at bullhead locations show that by sheltering in the lee of cobbles bullhead can find appropriate velocity conditions.

In this study, most bullhead were located in two large persistent mesohabitats: a large glide (unit 3) and a large pool (unit 27) both with a faster flowing mesohabitat immediately upstream. The high rate of velocity increase with rising flow within glides could prevent siltation. Further, the mesohabitat pairing may be significant as low energy refugia with the higher velocity run providing a reliable food source and surface turbulence (cover). This suggests that not only the size and persistence but also the longitudinal sequence of mesohabitats may be important in determining species distribution. This supports the need expressed by Fausch et al. (2002) to look at the riverscape as continuous mosaic of river habitats and examine their spatial distribution in relation to one another rather than taking a more traditional approach of sampling biota at separate ‘representative’ points and not acknowledging the features in between survey locations. This study illustrates the importance of cross scale investigation in linking fish ecology, flow and physical habitat variability and suggests the need to further elucidate the role of habitat dynamics and spatial arrangement in developing models of fish distribution by investigating longer reaches over a wider range of flows.

Copyright information

© Springer Science+Business Media B.V. 2010