- Top of page
1. Shellfish of marketable size can be harvested much more quickly and efficiently using mechanical methods such as tractor-powered harvesters and suction dredgers than by traditional methods. The adverse effects of such machines on non-target organisms need to be considered carefully before licensing such activities.
2. A tractor-towed cockle harvester was used to extract cockles from intertidal plots of muddy sand and clean sand in order to investigate the effects on other benthic invertebrates and their predators.
3. Harvesting resulted in the loss of a significant proportion of the most common invertebrates from both areas, ranging in the muddy sand from 31% of Scoloplos armiger (Polychaeta) (initial density 120 m−2) to 83% of Pygospio elegans (Polychaeta) (initial density 1850 m−2). Significant effects could not be detected in most populations with a density of less than 100 m−2.
4. Populations of Pygospio elegans and Hydrobia ulvae (Gastropoda) remained significantly depleted in the area of muddy sand for more than 100 days after harvesting, and Nephtys hombergi (Polychaeta), Scoloplos armiger and Bathyporeia pilosa (Amphipoda) for more than 50 days.
5. Invertebrate populations in clean sand with relatively few cockles Cerastoderma edule (Pelecypoda) recovered more quickly than those in muddy sand with a more structured community, which included several tube-dwelling species such as Pygospio elegans and Lanice conchilega (Polychaeta).
6. Bird feeding activity increased at first on the harvested areas, with gulls and waders taking advantage of invertebrates made available by harvesting. Subsequently, in the area of muddy sand, the level of bird activity declined compared with control areas. It remained significantly reduced in curlews Numenius arquata and gulls for more than 80 days after harvesting and in oystercatchers Haematopus ostralegus for more than 50 days.
7. It is concluded from this study that tractor dredging for cockles in high density areas causes a sufficiently large mortality of non-target invertebrates that harvesters should be excluded from areas of conservation importance for intertidal communities such as invertebrates, fish and birds.
- Top of page
Traditional manual methods of harvesting intertidal invertebrates such as cockles Cerastoderma edule L. are relatively inefficient and time-consuming. It has been assumed that the scale of the damage such methods cause to intertidal biota is small, provided that the numbers of gatherers involved is controlled. As mechanical methods become increasingly widespread, concern has been expressed that these may be damaging to non-target infauna (Cox 1991) and may lead to depletion of food supplies for predators such as oystercatchers Haematopus ostralegus L. and knots Calidris canutus L. (Lambeck, Goss-Custard & Triplet 1996; Piersma & Koolhaas 1997). One of the reasons for this is that mechanical harvesting disrupts large areas of substratum very quickly. This has led to the strict regulation of some fisheries in order to avoid adverse secondary effects upon birds, e.g. in Holland (Lambeck, Goss-Custard & Triplet 1996).
Since suction dredgers and tractor-towed harvesters have been employed, British cockle landings have risen from 10 000 tonnes per annum to over 20 000 tonnes per annum (Rees 1996). The evidence available from studies of these new harvesting methods in Britain, including the results of an intensive study in the Solway Firth (Hall & Harding 1997), has suggested that the effects on populations of non-target infauna are small. We present evidence from a study of tractor-towed cockle harvesting in the Burry Inlet, South Wales, which indicates otherwise.
The cockle fishery in the Burry Inlet has existed at least since mediaeval times. Manual harvesting in the early part of this century meant that daily yields per person were limited to 2–3 cwt (100–250 kg). The introduction of horse-drawn carts in the 1920s allowed this to rise to about 10 cwt (500 kg) per person per day (Coates 1995). The fishery currently removes less than 25% of the available stock of cockles, but even this results in a decrease in the numbers of oystercatchers feeding in the Inlet in some years (Norris, Bannister & Walker 1998). The introduction of mechanical harvesting methods would allow more efficient exploitation of the stock, and a proposal to use a tractor-towed machine within the Burry Inlet site of special scientific interest (SSSI) provided the impetus for the present trial of the effects of such a harvester upon non-target invertebrates and their dependent bird populations.
- Top of page
The study protocol was established by a working group of the South Wales Sea Fisheries Committee, the North-western and North Wales Sea Fisheries Committee, the Joint Nature Conservation Committee, the Countryside Council for Wales, Scottish Natural Heritage and the Ministry of Agriculture, Fisheries and Food. Two experimental areas (each roughly hexagonal in shape) in contrasting substratum types, muddy sand (area A) and clean sand (area B), were chosen. These two areas were located in the central parts of the Burry Inlet, in regions of high and low cockle density, respectively (Fig. 1). Low density areas were considered more likely candidates for licensed mechanical extraction because of their unsuitability for manual harvesting. Another key feature of the sampling was to allow measurements of the efficiency with which particular size classes of cockles could be selectively removed from the sediment, and smaller size classes returned with minimal damage. This aspect of the study, and more detailed information about the sampling areas, has already been published (Cotter et al. 1997).
Figure 1. Location of the study areas in the Burry Inlet, South Wales, and (right inset) plan of one area to show sampling plots for invertebrates with adjacent harvested (h) and control (c) sectors and bird footprints (dotted lines).
Download figure to PowerPoint
The tractor-towed cockle harvester employed in the trials is illustrated in Cotter et al. (1997). Sediment was removed from an adjustable depth and transported up a ramp and into a horizontally mounted rotating drum with a slight rearward tilt. Sediment and other material below a certain size (determined mainly by the size of the holes in the screens of the drum) fell between the vehicle tracks and formed a ridge, while objects above this size (mostly cockles) moved to the back of the drum and were collected in sacks. The harvested strips were intended to be immediately adjacent to one another, but in practice small gaps meant that about 18% of the area was left unharvested (Cotter et al. 1997).
Within each trial area, 0·98 ha of which was harvested, there were six plots each containing adjacent 30 × 20 m harvested and control sectors (Fig. 1). The corners of the sampling plots were marked with coded nylon whip aerials attached to wooden plates buried in the sediment and placed in position during harvesting. These were inconspicuous and resistant to storm damage.
The invertebrate sampling was conducted as follows in each area. On the day before harvesting and immediately after harvesting, which took place on 29 October 1992, 10 core samples (0·01 m2 in area and 7 cm deep) were taken from control and harvested sectors of each plot. Fifteen and 86 days after harvesting, a similar set of five samples was taken from each sector. A final set of 10 samples was taken from each sector 174 days after harvesting in area A. The positions of the cores within the sectors were determined using randomized co-ordinates on each occasion. The samples were preserved in 10% formalin and invertebrates were removed subsequently by manual rinsing through a 0·5-mm mesh sieve. An additional sediment sample was taken from each sector for granulometric assay using a Malvern 2600L laser size particle analyser (Malvern Instruments Ltd, Malvern, UK).
The species richness of the pre- and post-harvesting invertebrate communities was measured by α in the log series (rank/abundance plots were reasonably linear), dominance by Simpson's index and equitability by Shannon evenness. These indices were jack-knifed and values were compared using Student's t-tests (Magurran 1988). The raw results of the invertebrate sampling did not satisfy the assumptions necessary for parametric analysis, usually because of the large number of samples with zero counts. Power analysis was used to eliminate those species that were too infrequent or variable in abundance to enable significant effects of change to be detected. Box–Cox transformations were then used to achieve normality for the remaining species. The effects of harvesting, date and plot on the abundance of each of these species was determined by anova after checking for homogeneity of variance. All interactions were included in these analyses but only those between harvesting and date were reported because these were important in interpretation. Least significant differences were calculated in order to determine for how long the effects of harvesting remained significant. Although some plots recovered from the effects of harvesting faster than others, the overall plot means were used in these tests. Invertebrate population recovery times were estimated from the point at which the harvested density reached the control density (e.g. Fig. 2 for Bathyporeia pilosa (Lindström)). If the density in the harvested sectors remained lower than the control density, recovery was assumed to occur on the day halfway between the sampling date on which the densities ceased to differ significantly and the preceding sampling date.
Figure 2. Mean density (± SE) of the most abundant species of invertebrates in the control and harvested sectors of each area. SE estimated by bootstrapping the untransformed counts. Black squares = control; white squares = before cockle harvesting; circles = after cockle harvesting. Differences exceeding the least significant difference for the transformed counts on the relevant date are indicated by an asterisk. *P < 0·05.
Download figure to PowerPoint
The relatively small size of the experimental areas and their remote location made it difficult to observe bird foraging behaviour directly. This was mainly because of the difficulty of locating birds accurately in the foreshortened views of the areas under suboptimal viewing conditions. After initial trials it was therefore decided to use bird footprints as an index of activity (Wilson 1988). This has several disadvantages, including the fact that the species responsible for making the footprints cannot be identified with certainty, and that non-feeding birds make tracks that are virtually indistinguishable from feeding ones. The major advantage of footprints in the present context is that they can be located precisely. They were quantified as follows. Circular plastic hoops were laid on the sediment in lines running in eight compass directions from the centre of each area. This meant that 25 immediately adjacent circular samples of 0·44 m2 were examined in the harvested sector, and at least 30 identical circles in the non-harvested sector (Fig. 1). The total area sampled was thus about 190 m2 in each experimental area. The presence or absence of the footprints of oystercatchers Haematopus ostralegus, curlews Numenius arquata (L.), gulls, medium-sized waders and small waders was recorded in each circular sample. The number of footprints visible depended greatly upon the weather, with prints disappearing quickly in windy and rainy conditions. Wind caused prints in and around the fringes of any standing water to disappear, while rain caused footprints in drier areas to erode. For this reason, the presence or absence of footprints in each sample area, rather than the number, was analysed.
Oystercatcher footprints (total length, including hind toe, c. 80 mm) were distinguished from those of the only other wader with long digits, the curlew (total footprint length c. 76 mm), by the greater width of each toe and the much wider spread of the outer toes. Medium-sized footprints, potentially made by grey plovers Pluvialis squatarola (L.), golden plovers Pluvialis apricaria (L.), lapwings Vanellus vanellus (L.), knots Calidris canutus, redshanks Tringa totanus (L.) and bar-tailed godwits Limosa lapponica (L.) (Howells 1995), were recorded on only two dates, and brent goose Branta bernicla (L.) footprints on only four dates; so none of these was analysed. The footprints of the two small waders present, dunlins Calidris alpina (L.) and ringed plovers Charadrius hiaticula (L.), could in theory have been distinguished by the absence of the hind toe in the latter. In practice, the hind toe was only visible in clear unweathered prints, and it had disappeared from most weathered ones. However, ringed plovers were never seen in the area, even though they regularly fed on a muscle scar near area A. It was therefore assumed that all small footprints (total length c. 28 mm) were made by dunlins. Gull footprints were readily distinguishable from those of waders by the webbing, but black-headed gulls Larus ridibundus L. and common gulls Larus canus L. were too similar in size to be reliably distinguished. The feeding marks made by flatfish (Summers 1980) were also recorded. These consisted of approximately circular depressions in the sediment, about 2–6 cm in diameter.
The validity of the footprint technique was tested on two occasions with optimal viewing conditions and low winds (26 November 1992 and 23 January 1993). The number of birds of each species within control and harvested sectors of both areas was counted at 15-min intervals following the exposure of the area by the falling tide, until they were disturbed by the arrival of the person recording the footprints. All birds were actively feeding at this stage of the tidal cycle. The agreement between the level of bird activity directly observed (number of feeding birds times the number of hours they were present) and the proportion of samples with footprints was reasonably close (r = 0·759, P < 0·0001). On average, a doubling of bird activity (i.e. numbers feeding per unit time) resulted in a 1·6-fold increase in the proportion of samples occupied by footprints.
Bootstrapping was used to determine whether the difference in the number of samples with and without birds footprints was significant. This was done by combining all samples, drawing two sets of surrogates with replacement (representing control and harvested sectors) and counting the difference in the number with and without prints. The proportion of 10 000 such resamplings with differences as large or larger than that actually observed was calculated. This procedure, being based only on the occurrences of footprints actually observed, is free from all distributional assumptions (Simon 1995).
- Top of page
The poorly sorted fine muddy sand in area A (modal particle size peaks at 3 and 7 φ) had a rather different invertebrate fauna to the coarser and better sorted clean sand in area B (modal particle size peak at 3 φ) (Table 1). The former had more tube-dwelling and sedentary species such as Pygospio elegans Claparède and Corophium arenarium Crawford, and the latter more mobile species such as Bathyporeia pilosa. The initial density of cockles was about an order of magnitude higher in the muddy sand (Table 1). A dark anoxic mud layer was brought to the surface by the action of the harvester in the muddy sand, but no such layer was revealed in the clean sand. The harvested strips remained visible on part of the area of muddy sand until the end of the study in April 1993, but took only about 2 weeks to disappear completely from the clean sand.
Table 1. Macroinvertebrate community composition before and after harvesting in the two study areas. Average abundance (per m2) is given for each species; taxonomy in accordance with Howson & Picton (1997). Differences in community parameters before and after harvesting were tested by jack-knifing and using Student's t-tests (two-tailed) on the pseudo-values
|Time in relation to harvesting|
|Area A (muddy sand)||Area B (clean sand)|
|Species||Before||Immediately after||After 86 days||Before||Immediately after||After 86 days|
|Tetrastemma sp. Ehrenberg|| || ||192||87||47|| |
|Eteone longa (Fab.)||18||10||13||47||43||57|
|Anaitides mucosa (Oersted)||7||7|| || ||2|| |
|Eumida sanguinea (Oersted)||3||3||3|| || || |
|Hediste diversicolor (O.F.Müller)|| || || ||3||2|| |
|Nephtys hombergii Savigny||153||73||53||13||5||13|
|Scoloplos armiger O.F.Müller||120||83||367||3||8|| |
|Pygospio elegans Claparède||1852|| ||323||733||95||57|
|Spio martinensis Mesnil||7||3||3||3||7||13|
|Spiophanes bombyx (Claparède)||5||2|| ||12||5||20|
|Tharyx sp. Webster & Benedict||2|| || || || || |
|Psammodrilus sp. Swedmark|| || ||122||113||60|| |
|Capitella sp. de Blainville||3||2||3||22||27||60|
|Lanice conchilega (Pallas)||50||35||27||5||2|| |
|Halicyclops magniceps (Liljeborg)|| ||2|| ||12||3|| |
|Gastrosaccus spinifer (Goës)||5|| ||3|| || || |
|Urothoe poseidonis Reibisch||45||52||17||7||5|| |
|Bathyporeia pilosa Lindström||45||12||157||1920||343||3943|
|Bathyporeia pelagica (Bate)|| || ||103|| || ||227|
|Corophium arenarium Crawford||480||228||107||3||3|| |
|Eurydice pulchra Leach|| || || || ||3|| |
|Sphaeroma monodi Bocquet||3|| || || || ||10|
|Crangon crangon (L.)||20|| ||10||8||5|| |
|Carcinus maenas (L.)||7||5|| ||3|| ||3|
|Hydrobia ulvae (Pennant)||5143||2045||2476||423||185|| |
|Retusa obtusa (Montagu)||5|| ||3|| || || |
|Mytilus edulis L.||2||7||17||5||2|| |
|Cerastoderma edule (L.)||1127||737||780||115||98||103|
|Angulus tenuis (da Costa)||13||17||13|| ||2|| |
|Macoma balthica (L.)||18||28||23||30||28||30|
|Scrobicularia plana (da Costa)||2||2|| || || || |
|Species richness (α)||3·13||2·95***||2·94***||3·06||4·26***||1·36***|
|Dominance (Simpson's index)||0·38||0·36||0·28***||0·46||0·20***||0·78***|
|Equitability (Shannon evenness)||0·42||0·48*||0·57***||0·44||0·66***||0·24**|
Soft-bodied invertebrates were subjected to grinding by cockle shells and other debris inside the drum of the harvester. Not surprisingly, a great many annelids and thin-shelled molluscs, including some of the smaller cockles (Cotter et al. 1997), were damaged. These were deposited on the surface of the sediment and afforded an immediate source of food for predators. The first to take advantage of this were common and black-headed gulls, which gathered to feed on the most recent ridges while the machine was working. The peak numbers involved were 200 black-headed and 55 common gulls on area A and 43 black-headed and 50 common gulls on area B. These birds concentrated on the most recently harvested strip, feeding on the worms, crustaceans and molluscs left lying on the sand. Reasonably intact invertebrates buried themselves in the sediment within a few minutes, leaving only moribund ones on the surface. There was thus a relatively short-lived food supply available to scavengers and predators on the surface, and a longer-lasting one that remained shallowly buried.
With the exception of Bathyporeia pilosa, and one or two other species, the initial density of invertebrates was far higher in the area of muddy sand (A) than in the area of clean sand (B) (Table 1). The initial densities of cockles were 10–20% higher than those recorded by Cotter et al. (1997) on area A, perhaps because there were some cockles too small to be retained by their 4-mm sieve. Species richness declined significantly after harvesting in area A, but increased at first in area B (Table 1). In the former case this was due to the loss of five species, and in the latter to the gaining of three. In all cases, except perhaps Crangon crangon (L.) in area A, the species involved were very uncommon. Dominance declined significantly in area B, largely as a consequence of the marked reduction in the numbers of Bathyporeia pilosa following harvesting. Equitability increased significantly in both areas as the population densities of the dominant species crashed following harvesting. By 86 days after harvesting, species richness was significantly lower in both areas than at the outset. Equitability increased further in area A, but declined in area B compared with its initial values.
Annelids declined by 74% on area A and 32% on area B, and the differences in the proportion remaining alive in the two areas was significant (contingency χ2 = 141·7, P < 0·0001). The difference in the decline of molluscs in the two areas was much less marked (55% in area A, 45% in area B), but this too was significant (contingency χ2 = 7·49, P = 0·006). Annelids were vulnerable to being crushed in the harvester drum by large numbers of hard-shelled molluscs. Crustaceans declined by 81% in area B, compared with only 56% in area A, and this difference too was significant (contingency χ2 = 87·0, P < 0·0001). The change in this case was dominated by Bathyporeia pilosa, and much of the decline in this species was probably due to emigration (see below). There was no evidence of any increase in the abundance of invertebrate scavengers, such as Carcinus maenas (L.) and Crangon crangon, in any of the harvested areas, either individually or as a whole. This suggests that their ability to locate and move into small areas of increased food abundance was limited, or that local populations were small. This is in marked contrast to invertebrate scavengers in offshore areas which have been found to aggregate on trawled areas to take advantage of the increased availability of prey within 1–7 h (Kaiser & Spencer 1994; 1996a).
anova revealed marked effects of harvesting, date and plot on the abundance of all the common invertebrate species (Table 2). All harvesting effects involved decreases in density, and the most pronounced reductions occurred in the densest populations. Thus the losses ranged from 15% in the low density cockle population in area B to over 80% in the high density Pygospio elegans population in area A and Bathyporeia pilosa population in area B (Table 2 and Fig. 2). The significant effects of date in the control sectors mostly involved reductions in population size during the course of the winter (six reductions and two increases among the 13 populations in Table 2). However, there were three increases and two reductions in the harvested sectors, perhaps as a result of some immigration from the surrounding areas. Differing trajectories of population change were responsible for the significant interaction effects in Table 2. The many significant plot effects demonstrated the considerable variation in invertebrate density present in quite small areas. Invertebrate densities remained depleted on area A for longer than on area B, with Hydrobia ulvae (Pennant) and Pygospio elegans having failed to recover 174 days after harvesting. Bathyporeia pilosa is a highly mobile species, capable of making a rapid recovery. The high population density on area B recovered in 39 days, but the much smaller population on area A did not do so until 51 days after harvesting.
Table 2. anovas of the transformed totals for each of the most abundant species of macroinvertebrates. The F-values and significance levels are shown for each factor. The percentage change in population size immediately following harvesting and the time to recovery (see text) are also shown
|Species||Area||Harvesting||Date||Plot||Harvesting × date||% change after harvesting||Time to recovery (days)|
|Pygospio elegans||A||66·1***||1·3||17·7***||1·5||−82·6||> 174|
|Hydrobia ulvae||A||106·4***||7·2***||9·8***||5·5**||−60·2||> 174|
|Cerastoderma edule||A||60·3***||2·4||4·1**||0·4||−34·6||> 174|
The relative differences in bird activity between control and harvested sectors were, in all cases, less marked than the relative differences in invertebrate densities. In the few days following harvesting, there was a large increase in the numbers of waders, gulls and flatfish feeding in the harvested area (Fig. 3). For example, 80 dunlins and seven curlews were observed feeding on the harvested strips 6 days after harvesting. These birds were probing rather than pecking at the surface, and were presumably feeding on buried invertebrates. This elevated level of bird activity also included some of the control sectors, although a much higher proportion of samples within the harvested sector yielded prints of all species and feeding signs of flatfish. The increased level of predator activity in the vicinity of the harvested sectors was more marked in the area of muddy sand (A) than on the area of clean sand (B) and was especially noticeable 6 days after harvesting in the control sectors of area A (Fig. 3).
Figure 3. Mean proportion (± SD) of samples in the control and harvested sectors containing footprints of different bird species. SD and significant differences between sectors (indicated by an asterisk) estimated by bootstrapping. Black squares = control; white circles = harvested sectors. *P < 0·05.
Download figure to PowerPoint
Flatfish feeding marks were present in 23% of the 0·44 m2 harvested samples in area A, but only 9% of the controls, and this difference was significant (contingency χ2 = 4·96, P = 0·026). Very few flatfish feeding marks were present on area B on this date. Less than 5% of samples showed any flatfish feeding activity on both areas subsequently.
Fourteen days after harvesting, no significant differences were detected in bird activity between control and harvested sectors. This was at a time of high winds, when footprints eroded rapidly, explaining the apparent drop in apparent activity on both areas (Fig. 3). Significant reductions in bird activity on the harvested sectors became apparent 21 and 45 days after harvesting in all cases except gulls on area A and dunlin on area B. No differences in bird activity between control and harvested sectors were detectable 115 days after harvesting in all cases except dunlin on area A (Fig. 3).
- Top of page
Mechanical harvesting clearly had a major impact on invertebrate communities but sampling was not continued for long enough to determine how long they took to recover. The movement of the adults of most species, or their passive transport in bulk movements of sediment (Ratcliffe, Jones & Walters 1981; Evans 1981; Ferns 1983), was apparently sufficient to allow recovery of the modest invertebrate populations in area B, but was inadequate to allow recovery of the large populations in area A. Re-establishment of the original populations in area A would probably have been dependent on the recruitment of young. Small numbers of juvenile specimens began to appear in the samples taken in January and significant numbers appeared in April, but there were no significant differences in their densities between control and harvested sectors. The most abundant juveniles were those of Scoloplos armiger, Pygospio elegans and Bathyporeia pilosa, with densities in the range 165–290 m−2 in April.
The ability of invertebrates to recolonize depleted areas is very variable. Bathyporeia pilosa is capable of undertaking mass movements during a single tide and should have been capable of recolonization within a few days. The fact that it did not recolonize area B fully for about 39 days suggests that conditions were initially unsuitable. Conditions remained unattractive to Bathyporeia pilosa for even longer in area A, as full recolonization did not occur there until 111 days had elapsed. The initial unattractiveness may have been due to defensive and other chemicals released into the sediment by dead and injured animals. For example, Arenicola juveniles are stimulated to settle in sand conditioned by healthy Nephtys hombergii Savigny, yet they avoid sand containing defensive secretions produced by stressed individuals (Hardege, Bentley & Snape 1998). Young spionids prefer to settle in fresh sand when compared with that conditioned by arenicolid polychaetes. The smallest numbers settle in sand containing injured worms (Woodin 1985).
The slow recovery of the adult populations in area A may also have been a consequence of the physical disruption caused by the harvester to the complex layered structure of the sediment and the communities it supported. The fact that the anoxic layer was brought to the surface and dispersed, along with the tubes of Lanice conchilega, indicates the severity of this disruption. Sulphides will have been released into the upper layers of the sediment and may have contributed to the longer time needed for recovery (Schöttler & Grieshaber 1988; Rees 1996; Schiedek et al. 1997). The fact that more complex and productive intertidal communities, such as that in area A, take longer to achieve stability after disruption is perhaps not surprising given both theoretical and practical considerations (May 1981; Kikkawa 1986; Kaiser & Spencer 1996b; Lawton et al. 1998). This is one of the factors that needs to be taken into account before allowing disruptive activities to take place. Another significant factor affecting the rate of invertebrate recovery is the longevity of the species involved. Large species such as Arenicola marina, Mya arenaria and Ensis sp. take several years to reach maturity, and therefore take much longer to recover than smaller ones (Conner & Simon 1979; Hall, Basford & Robertson 1990; Beukema 1995). Cockles were the most abundant long-lived species in the Burry Inlet community.
A previously published study of mechanized cockle harvesting conducted during the summer in the Solway Firth, found that non-target benthic infauna had recovered from tractor dredging 56 days after dredging (Hall & Harding 1997). This is quite similar to our findings for area B. The density of cockles at the start of Hall & Harding's (1997) trials was about 127 m−2, which was similar to the density on area B (115 m−2) but an order of magnitude lower than area A (1127 m−2). Small invertebrates were subjected to a more powerful grinding action from the larger number of cockles in the harvester drum on area A. For example, the damage to cockles small enough to escape harvesting was much greater in area A (mean percentage ± SD of surviving individuals damaged = 10·3 ± 2·2) than in B (5·9 ± 1·8) (calculated from data in Cotter et al. 1997).
The Solway study was based on 24 unmarked experimental areas in which a total of 2·52 ha of sediment was harvested. In our study, 1·96 ha was harvested, but divided into only two experimental areas. Clearly, the potential for more rapid colonization of small areas of sediment is greater. The relocation of the smallest of Hall & Harding's (1997) areas (15 × 15 m) using a global positioning system capable of accuracy to only ±15 m must have also been difficult. Another important difference between the two studies is that ours was carried out during the winter months and Hall & Harding's (1997) during the summer. Different species of invertebrates spawn at different times, but the majority do so between May and September. Spawning leads both to greater mobility of adult individuals in the search for mates (e.g. Hediste, Corophium;Fish & Mills 1979; Mettam 1981) and, in due course, to recruitment of young. Spawning takes place mainly in spring (Nephtys hombergi, Scoloplos armiger and Macoma balthica) or spring and autumn (Lanice conchilega, Corophium arenarium, Hydrobia ulvae and Cerastoderma edule). Only in Pygospio elegans and some populations of Bathyporeia pilosa does breeding occur throughout the year (Fish & Fish 1996). Both the latter species were relatively slow to recover in our study, despite the fact that the samples in January contained many Pygospio tubes with egg capsules and small pelagic larvae. Significant numbers of juveniles of all species did not appear in the samples until April. This is in marked contrast to the rapid colonization of borrow pits in the Wash by Pygospio (McGrorty & Reading 1984) and supports the notion that adverse physical and chemical conditioning of the sediment was a factor inhibiting settlement.
Most forms of intertidal invertebrate harvesting provide some initial extra feeding opportunities for intertidal birds, including manual digging, e.g. for ragworms as bait (Aspinall 1992). As cockle harvesting also resulted in an initial increase in food supplies for gulls and waders, it might be argued that its net effects were beneficial to birds. In fact, these benefits were matched by decreased feeding opportunities later in the winter. This does not mean, however, that the effects of harvesting larger areas would be neutral for shorebirds. It is unlikely, for example, that there would be enough birds available locally to fully exploit the greatly increased feeding opportunities during harvesting. More importantly, the subsequently reduced feeding opportunities would extend over a long period of time. Its effects would also impinge on birds in unharvested areas as a consequence of movements of individuals away from harvested zones. When feeding conditions change, birds can be mobile both within estuaries (Davidson 1981; Ferns 1983; Symonds, Langslow & Pienkowski 1984; Rehfisch et al. 1996) and between them (Evans 1981; Evans & Townsend 1988). Such movements would inevitably lead to increased bird densities elsewhere (Sutherland & Goss-Custard 1991; Goss-Custard 1993). In other words, much of the increased availability of food would be unexploited by birds, while the subsequent food reductions would lead to increased densities elsewhere. Similar short- and long-term effects are likely to occur in the case of predatory fish. The former, for example, have been demonstrated in detailed studies of beam trawling (Kaiser & Spencer 1994, 1996a,b).
This study has demonstrated that tractor dredging for cockles can cause significant depletion of non-target invertebrates for several months and consequently can reduce bird feeding activity, especially in more productive areas of intertidal sediment. Such harvesters ought therefore to be excluded from use in those areas in which the conservation of intertidal communities (i.e. invertebrates, fish or birds) is given priority, such as SSSIs, special areas for conservation and marine nature reserves. While it could be argued that little damage would be caused by harvesting small areas with low densities of intertidal invertebrates (such as our area B), the consequences for areas with higher densities (such as area A) would be serious. Commercial harvesting on a larger scale would have even more far-reaching effects.