ANIMAL BEHAVIOUR, 2001, 61, 645–654 doi:10.1006/anbe.2000.1628, available online at http://www.idealibrary.com on
Behavioural responses of juvenile barbel in an artificial channel: distribution and velocity use LORENZO VILIZZI & GORDON HOWARD COPP
Landscape & Ecology Research Group, Department of Environmental Sciences, University of Hertfordshire (Received 20 March 2000; initial acceptance 2 May 2000; final acceptance 30 October 2000; MS. number: 6525R)
To determine the combined effects of depth, discharge and time of day on the distribution and velocity use of juvenile barbel, Barbus barbus, we observed groups of 20 fish over replicate 24-h periods at dawn, midday, dusk and midnight in an artificial channel. Four combinations of depth and discharge were examined (low depth–low discharge: LDLS; high depth–low discharge: HDLS; low depth–high discharge: LDHS; high depth–high discharge: HDHS). At dawn and midday, the fish were more aggregated than at dusk and midnight, regardless of depth and discharge. However, the latter variables did have a subtle effect on distribution, but only through a complex interaction with time of day, which we detected by a combination of a modified index of distribution and a wider reference lattice. In particular, at dusk and midnight juvenile barbel were dispersed under lentic conditions of HDLS, were reaggregated when velocities increased at LDLS and HDHS, and were dispersed again under the more lotic conditions of LDHS. Furthermore, the weighted used area for velocity decreased dramatically under LDHS and was highest at HDLS. The ecological significance of these results is discussed, with special attention to their likely consequences for foraging, growth and, ultimately, recruitment success in barbel.
The increased intensity of discharge variations caused by river regulation, combined with the loss of marginal and off-channel annexes, has contributed to the declines in fish stocks in medium-size rivers such as the Great Ouse (Copp 1990). The same may also be applicable to smaller streams such as the River Lee, Hertfordshire, U.K., in which the discharge regime has been made more variable by the combined effects of groundwater abstraction and land and road run-off drainage infrastructure. In particular, treated sewage effluent discharges play an important role in the Lee, which receives a large proportion of its flow (40–80%) from a sewage treatment plant near its source (Pilcher & Copp 1997). These hydrobiological variations imposed by the effluent discharges are linked to human domestic-related patterns, resulting in river discharge increases that coincide with dusk decreases in light intensity (H. Faulkner & G. H. Copp, unpublished data). Fish are known to disperse as light levels decrease, but little is known of the concomitant behavioural response of fish to combinations of fluctuating variables. There is evidence that fish forage more efficiently when dispersed (Pitcher & Parrish 1993), so any behavioural response to environmental fluctuation that diminishes foraging efficiency and/or increases predation risk may subsequently decrease growth and survivorship. Comprehensive baseline data on the behavioural response of riverine fish to
Unlike many marine fish species, which sustain higher mortalities as larvae, freshwater fish are expected to be more vulnerable as juveniles (Houde 1994). In river systems, the dispersion of most fish species is greatest as larvae and juveniles, which are less able than older, larger conspecifics to maintain station in elevated water velocities (Mann & Mills 1986; Lightfoot & Jones 1996). Predation risk is also relatively high for juveniles and thus behavioural mechanisms of aggregation and dispersion play an important role in species-specific responses to variations in environmental conditions and pressures (Pitcher & Parrish 1993). Human alteration to river ecosystems can increase or diminish these natural variations, with potential effects on recruitment and year-class strength. Such detrimental effects are most pronounced in fish larvae and juveniles, with changes in water velocity and discharge being particularly important in regulated rivers. For example, Mann & Bass (1997) found the usable area of a side channel in the River Great Ouse, U.K., to decrease dramatically during elevated discharge, with juvenile dace, Leuciscus leuciscus, unable to hold station except for relatively short periods. *Correspondence and present address: L. Vilizzi, Allsystem, Via Florinas 6, 07100 Sassari, Italy (email:
[email protected]). G. H. Copp is at the Department of Environmental Sciences, University of Hertfordshire, Hatfield AL10 9AB, U.K. 0003–3472/01/030645+10 $35.00/0
2001 The Association for the Study of Animal Behaviour
645
2001 The Association for the Study of Animal Behaviour
646
ANIMAL BEHAVIOUR, 61, 3
variations in discharge, water level and time of day are therefore required as a precursor to foraging or predation risk studies. Our aim was to examine how variations in water level, discharge and time of day affect distribution and velocity use in juvenile barbel, Barbus barbus. The barbel is a European cyprinid sensitive to pollution and to physical alteration of the stream ecosystem, and, as such, risks extinction in some areas of its range (Watkins et al. 1997). The present study, therefore, contributes to an understanding of the factors that influence foraging, growth and, ultimately, recruitment success in the barbel. METHODS
Experimental Set-up Hatchery-reared juvenile barbel of age 13 weeks and mean standard length (SL: measured from the tip of the snout to the base of the caudal fin) SE of 28.40.6 mm were obtained from Calverton Fish Farm, Nottinghamshire, U.K., in August 1998. In the laboratory, the fish were transferred to 60-litre glass tanks (water temperature 18C; natural photoperiod; stocking density 0.5 fish/litre) and were fed 800-m pellets (Rock Pauls Ltd) ad libitum twice daily. All fish belonged to the same cohort and were in their juvenile period of development throughout the study (August–November 1998). Experiments were undertaken in an artificial channel made of Perspex (18050 cm and 25 cm high) with the lower end fitted with an adjustable panel (weir) to regulate water level. Two removable partitions, consisting of Perspex frames supporting plastic mesh, divided the channel into an upstream inflow compartment 20 cm long and a downstream preweir compartment 10 cm long, leaving in between an experimental arena 150 cm long (7500 cm2; further details in Copp et al. 1998). Water recirculated through the system by way of two independent drainage pumps (Oase Aquadex A8), providing a turnover rate of 3.7 litres. Water temperature was kept at 211C, similar to summer temperatures in the River Lee (L. Vilizzi, unpublished data). A layer about 4 cm thick of gravel from the River Lee (particle size 0.5–2.0 cm) was laid on the bottom of the experimental arena to simulate a substratum composition found in natural riffle areas. Two overlapping reference lattices made of plastic strings were then laid on top of the arena, which was thus subdivided into two grids of 75 smaller (1010 cm, 155 quadrat grid) and 12 larger (2525 cm, 62 quadrat grid) contiguous quadrats, numbered sequentially with labels from the downstream upper corner to the upstream lower corner. Two water levels (14 cm and 20 cm) and two discharge regimes, provided by one pump (Q=0.0037 m3/s) or both pumps running at the same time (Q=0.0072 m3/s), were available for the study. For each of the four resulting combinations of depth (=water level) and discharge, water velocities were measured with a velocity meter at the centre of each of the 75 smaller quadrats and in the middle of the water column. There were no artificial lighting sources in the room at midday, whereas a red light lamp (100 W)
was used at night, immediately before dawn and soon after dusk. On each experimental occasion, 20 juvenile barbel, haphazardly selected from the tanks, were acclimatized to the channel for ca. 12 h prior to the experiment. The fish were not fed during the acclimation period nor throughout the experiment. We visually recorded the position of fish in the stream relative to the reference grids at each time of observation (hereafter ‘trial’). A fish was considered to occupy a specific quadrat when more than half of its body length was within a quadrat’s boundaries. There were 10 replicate trials at 2-min intervals, starting 10 min before and ending 10 min after each sampling time (dawn, midday, dusk, midnight), always at solar time. Upon completion of each experiment, the fish were measured for SL and returned to another set of holding tanks. Our experimental procedure was discussed in detail with the U.K. Home Office, which considered that the conditions of the experiments would not subject the fish to any undue stress or harm, and thus no Home Office licence was deemed necessary. No aggressive behaviour of any kind was observed when the fish were reassorted into new groups, and the fish did not show any stress in the face of changes in water level or discharge.
Experimental Design Depth, discharge and time were three predictor variables (factors) in an analysis of variance for repeated measures (ANOVAR, Potvin et al. 1990), which was based on a split-plot factorial design (SPF-224: notation after Kirk 1995). This consisted of two between-blocks factors (depth: D; discharge: S) and one within-blocks (repeatedmeasures) subjects-by-trials factor (time). There were two levels (low: L; high: H) of both depth (low depth: LD=14 cm; high depth: HD=20 cm) and discharge (low discharge: LS=0.009 m3/s; high discharge: HS= 0.030 m3/s), and four ‘levels’ of time (dawn, midday, dusk, midnight). All factors were fixed, calling for a model I design (Bennington & Thayne 1994). The experimental units, each consisting of a group of 20 fish, were assigned randomly to each of the four combinations (blocks) of depth and discharge levels (i.e. low depth–low discharge: LDLS; high depth–low discharge: HDLS; low depth–high discharge: LDHS; high depth–high discharge: HDHS). Replication within each combination was four-fold (i.e. four groups of 20 fish per combination), for a total of 16 replicate experimental units. Each replicate, therefore, consisted of a separate experiment, which took place over a 24-h period lasting from dawn to midnight, always in this order. Replicated experiments were systematically interspersed over time (i.e. LDLS, HDLS, LDHS, HDHS, LDLS, . . ., HDHS), for a total of 640 trials (10 trials 4 time levels4 depth–discharge combinations4 replicated experiments per combination).
Response Variables The response variables in the design were: (1) the index of dispersion ID (Upton & Fingleton 1985); (2) a new
VILIZZI & COPP: BARBEL BEHAVIOUR
index of distribution, here referred to as the partition index PI; and (3) the weighted used area for velocity WUAv (slightly adapted after Orth & Maughan 1982). ID was calculated as variance/mean ratio (s2/x¯), based on the number of fish counted in each (non-null) quadrat and including all null quadrats (scored as 0), where no fish were present. Departures from randomness towards contagion (terminology follows Elliott 1971) were assessed by chi-square (2) or d (a normal variable with zero mean and unit standard deviation), depending on the number of sampling units (quadrats=n). When n>31, the statistic d is calculated as:
where is the number of degrees of freedom (n1). A regular distribution will be present when d< 1.96, a random distribution when 1.96cd≤1.96, a contagious distribution when d>1.96. As the total number of sampling units, the sum of counts (fish: x=20) and the arithmetic mean of the sample (mean number of fish/ quadrat) were always constant, ID could be used reliably as a comparative index of dispersion (Elliott 1971). PI was computed based on the number of all possible arrangements of fish in the stream relative to the reference lattice in use. Thus, the number of ways an integer n can be written as a sum of positive integers cn is the number of solutions of n=ai1 +ai2 +. . . where aix is the xth positive integer a of the ith partition. The corresponding partition function p(n) can be obtained, for values of n≤14 031, by Euler’s recursion formula (Apostol 1986). Denoting the total number of partitions p(n)=N, calculation of PI is as follows. (1) Each of the N partitions (P1, P2, . . ., Pi, . . ., PN) is computed and, after sorting in ascending order, assigned a sequential number (identifier: Ii), based on the ascending order, from 1 (partition P1 consisting of all 1’s) to N (‘partition’ PN equivalent to N). (2) Each of the N identifiers is divided by the total number of partitions and converted to a percentage value, giving the partition index:
Whenever the number of quadrats q of the reference lattice in use is smaller than the number of fish under study (i.e. q
