Recovery rates of benthic communities following physical disturbance

Authors


K.M. Dernie. E-mail: k.dernie@ccw.gov.uk

Summary

  • 1Despite numerous studies that have investigated the effects of physical disturbance on marine benthic communities, deducing the ecological significance of such events has been hampered by the specificity of individual studies. Less stable habitats (coarse, clean sands) are thought to recover more quickly than stable (muddy sands and mud) habitats but so far an empirical test of this paradigm has been lacking.
  • 2We describe a large-scale field experiment that investigated the response of marine benthic communities within a variety of sediment types (clean sand, silty sand, muddy sand and mud) to physical disturbance. Sites were sampled for macrofauna and habitat characteristics (sediment organic content and water content, depth of water within disturbed pits) following disturbance, in order to examine any relationship between the rate of physical and biological recovery.
  • 3There were no detectable changes to the measured physical parameters following the disturbance with the exception of the water depth within disturbed areas. It was not possible to derive a linear relationship between the measured habitat characteristics (percentage silt and clay content) and the recovery trajectory of the associated community. However, the rate of sediment infilling of disturbed plots was strongly correlated to the recovery rate of the numbers of individuals within disturbed areas.
  • 4Clean sand communities had the most rapid recovery rate following disturbance, whereas communities from muddy sand habitats had the slowest physical and biological recovery rates. These findings concur with the predictions of a meta analysis of fishing disturbances for similar habitats.
  • 5We suggest that physical and biological recovery rates are mediated by a combination of physical, chemical and biological factors that differ in their relative importance in different habitats. Monitoring the infilling rate of physically disturbed patches of sediment has the potential to become a useful tool to predict the recovery rate of associated communities.

Introduction

Physical disturbance of habitats is an important factor in the maintenance of diversity within an ecosystem (Grassle & Saunders 1973; Connell 1978). The intermediate disturbance hypothesis proposes a predictable relationship between the frequency and intensity of disturbance and the resultant measure of species richness (Connell 1978). While this theory provides the basis for predicting the response to increasing levels of disturbance, it also implies that the subsequent recovery rate of the disturbed habitat is linked to the background disturbance regime and the stability of the habitat. Intertidal and shallow subtidal soft sediment benthic communities are subject to a range of natural disturbance regimes that are dominated by physical processes (Hall 1994) and are therefore amenable to testing the relationship between the recovery rate of the habitat and restoration of community attributes. Overall sediment composition is largely controlled by hydrodynamic forces over the substratum (Snelgrove & Butman 1994) such that clean, coarse sandy bottoms predominate in high-energy environments, whereas silty, muddy sediments develop in very low-energy environments. Presumably, the communities that inhabit such different sediment types have adapted to very different environmental disturbance regimes (Hall 1994). Many species that are typical of wave-exposed sandy environments exhibit behaviours that enable them to survive daily tidal scouring events (Gorzelany & Nelson 1987). Conversely, species found in low-energy muddy habitats are adapted to life in relatively low-oxygen environments (Forbes, Forbes & Depledge 1994), where organic loading of the sediment is high and oxygen mixing at the sediment/water interface is reduced due to stratification. Macrobenthic species are relatively unselective in their food requirements and depend on spatial partitioning of the habitat to maintain diversity. Spatial partitioning in turn is dependent on the stability of the sediment, and will break down in situations of frequent disturbance. It is generally assumed that communities found in dynamic sandy habitats will recover more quickly following physical disturbance than those found in less energetic muddy environments based on the adaptive strategies of the differing assemblages (Kaiser 1998; Ferns, Rostron & Siman 2000) and this seems to be supported by microcosm studies, e.g. (Schratzberger & Warwick 1998).

Physical disturbance affects not only the infaunal community but also the structure of the sediment matrix itself. Natural sediment deposits often exhibit vertical variation in particle sizes (Rhoads & Stanley 1965; Rhoads 1967) in addition to aggregations of faecal pellets and mats of bacteria/microalgae that affect sediment stability and erodability (Grant & Gust 1987; Miller, Geider & McIntyre 1996). The complexity of the structure of the sediment matrix may be increased due to the presence of biota, e.g. dense aggregations of tube building worms (Thrush et al. 1996). Physical disturbance resuspends and homogenizes the upper layers of sediment of the seabed (Palanques, Guileen & Puig 2001). Fine particles will remain in suspension for longer than larger particles resulting in the deposition of a surficial layer of very fine particles on the substratum, or the displacement of this fraction of the sediment as it is carried away on tidal currents. Disturbance of microbial and microalgal communities may have a concomitant effect on sediment stability, allowing surface sediments to be eroded more easily (Underwood & Paterson 1993; Paterson & Black 1999). The formation of surface features such as pits and mounds will affect the local hydrodynamics (Nowell & Jumars 1984). Where the upper layers of sediments are removed (i.e. in aggregate extraction) the remaining substratum may be comprised of a totally different sediment type that is unsuitable for recolonization by the species that previously inhabited the area (Kenny & Rees 1996).

At large scales empirical evidence suggests that there is an important relationship between gross physical parameters of the benthic environment and the composition of the associated infaunal soft sediment communities (Warwick & Uncles 1980; Yates et al. 1993). Following a disturbance event both the sediment habitat and the associated infaunal community are amenable to restoration over a period of time. Presumably some level of habitat recovery is required before the eventual recovery of the associated benthic assemblage.

Currently, there is considerable debate regarding the ecological significance of physical disturbance of benthic communities as a result of bottom fishing using towed fishing gears on continental shelf areas throughout the world (e.g. Dayton et al. 1995; Watling & Norse 1998; Kaiser et al. 2002). Empirical studies that have manipulated fishing disturbance in situ have reported a variety of responses that could lead to conflicting conclusions (see Jennings & Kaiser 1998; Kaiser et al. 2002). This has hindered the formulation of general conclusions regarding disturbance effects and subsequent recovery rates of communities that would be of use in an ecosystem management context. The variation in the responses reported in the studies undertaken to date almost certainly arises from the specific characteristics of each individual study, for example, location, fishing gear, season and habitat (Collie et al. 2000).

Collie et al. (2000) overcame the problems of study-specificity by undertaking a meta-analysis of published fishing impact studies to quantify general responses of community variables to this form of disturbance. Meta-analysis of independent studies is a useful tool (Gurevitch & Chester 1986; Gurevitch et al. 1992) for exposing general patterns in community responses to different treatments, especially if care is taken to avoid confounding biases in the data (Gates 2002). Collie et al.'s (2000) analysis revealed trends in community responses that followed broadly theoretical expectations for a habitat effect such that initial responses and rates of recovery from trawling impacts would be related to, and could be predicted from, the physical stability of the sea bed. The authors were able to infer that benthic assemblage recovery occurred most rapidly in sand habitats, followed by mud communities, while muddy sand communities had the slowest recovery trajectories following fishing disturbance. Nevertheless, the interpretation of these findings was flagged with caveats that highlighted the potential danger of broad categorization of habitats. The relationship between recovery rate and sediment type may be more complex than we intuitively assume and highlights the need for directly comparable experimental studies over a range of habitat types that test the predictions of Collie et al.'s (2000) meta-analysis.

We designed an experiment to investigate and quantify those changes to habitat parameters (organic content, water content and depth of disturbed pit) and community composition that occur in a range of soft sediment habitats following a physical disturbance treatment. This approach allowed a direct comparison of the recovery rate of benthic assemblages and habitat parameters in different sediment types at the same time of year in order to establish any pattern between the rate of biological and physical recovery and habitat type. The experiment was restricted to the intertidal zone to allow access to sites and accurate manipulation of our treatment areas. The scale of disturbance was chosen to be relevant to fishing impacts that occur both intertidally and subtidally. Such activities would include bait digging, hand-raking, suction-dredging and some forms of trawling. We wished to establish whether the recovery rate of intertidal benthic assemblages and sediment parameters could be predicted by habitat type and whether there was any predictive relationship between recovery of physical and biological components following a disturbance event in line with the expectations of Collie et al. (2000).

Methods

Field sites of varying sediment characteristics were studied along the Menai Strait, North Wales, UK (Fig. 1). Preliminary testing of the sites for physical (granulometric) and biological (community) characteristics resulted in the choice of four replicate sites for each of four sediment types. For granulometric analysis, sediment samples were first washed to remove salt over a filter paper using a Buchner funnel and vacuum system. The washed sediment was wet-sieved through a 63-µm sieve to separate the silt/clay fraction, which was subsequently dried under hot lights, weighed and used for the calculation of percentage of silt and clay. The remaining sediment was dry-sieved through a series of Wentworth sieves (2 mm, 1 mm, 0·5 mm, 250 µm, 125 µm, 63 µm and < 63 µm fractions measured) on a mechanical sieve shaker for 20 min. The four sediment types were defined broadly as clean sand (CS: < 3% silt and clay), silty sand (SS: 5–20% silt and clay), sandy mud (MS: 35–45% silt and clay), and mud (M: > 55% silt and clay) (Table 1), where the silt and clay fraction was defined as the sum of the 63 µm and < 63 µm sieve fractions. All sites were located between the low to mid-shore level.

Figure 1.

Position of experimental sites along the Menai Strait, North Wales. Sites are grouped according to the percentage silt and clay content of sediment: clean sand sites (< 3% silt and clay), silty sand (5–20% silt and clay), muddy sand (35–45% silt and clay) and muds (> 55% silt and clay).

Table 1.  Numbers of individuals of six most common species and taxa present within each replicate treatment and control plot on the first and last sampling occasions (n.d. = no data, treatment plot lost)
Site % silt/clayorganicsdayP. eleganT. benediiM. balthicaCirratulidaeCapitellidaeCorophium spp.
cdcdcdcdcdcd
W1CS 1·61·5 1520 6  6  1 61 12  5  2 0 16 11
    213 6 5  1  5 70  1  0  0 0  6  4
W2CS 2·21·3 1562 1 17  2 00  0  1  1 0220 47
    21310n.d.  5n.d. 1n.d.  0n.d.  1n.d. 71n.d.
W3CS 2·191·9 15 1 0  1  3 00 20 16  3 1  0  0
    213 1 3  2  2 00  3  7  2 2  0  0
W4CS 1·461·2 15 5 0  1  1 00  3  3  0 0 19  5
    213 9 0  0 12 00  3  5  3 2  6  4
X1SS19·682·0 1517 1778150303548128 1412  1  0
    2134923653170116325359  1 3 86 17
X2SS11·311·6 1558 0  9  0 10  3  0  3 0 19  1
    21319 9 14 88 00  0  0  0 5 72 38
X3SS21·132·3 15 0 0 36  1 00276  9 32 0  0  0
    213 914 32 25 03119171  2 0  0  0
X4SS10·881·7 15 9 16  1 00 56  0  2 0 20  3
    21321373 14 40 22 43  1 6  1  5
Y1MS44·022·6 15 0 1 37 42100 34  3 13 2  0  1
    21314 5482 75 22128 15 10 1150 62
Y2MS45·872·6 15 0 0392  2 00 47  0  7 2  0  0
    213 0 1421120 40 40 54  3 1  3  0
Y3MS56·164·6 15 9 0 42  0 24 86  3  7 1  2  0
    213 4 2104 28 01406260  3 7  7  4
Y4MS34·961·6 15 0 0 10  0 10 86  3  2 0  1  1
    213 0 28  5 15 31 28  1 0  2  4
Z1M60·893·0 15 0 0395 34 11109 12  3 4  7  0
    213 2 1629 95 30117 24  3 0354485
Z2M65·485·0 15 0 0  1  1 00  1  0  0 0  0  0
    213 0 0 32 25 00 21 22  3 5  0  0
Z3M59·254·4 15 0 0 23  6 00218  9 35 2  0  0
    213 0 0 52 31 00275257  0 0  0  6
Z4M63·133·4 15 0 0194 93 0092216432348 23  7
    213 0 0745574 00660701  2 0 51 48

At each site, one control plot and one ‘disturbed’ plot of 1 m × 4 m were marked out using steel poles. In a previous experiment in a clean sand habitat (Dernie et al. 2003) 2 × 2-m plots were used. The dimensions of plots chosen for this study allowed samples to be taken while minimizing any trampling effect within the plots, as this is a particular problem when sampling muddy sediments. Throughout the experimental period, efforts were made to reduce the effects of trampling both in the areas within and adjacent to the experimental plots. Plots were always approached from the same direction and samples taken over one side of the plot only to avoid trampling the entire circumference of the area, as this would have effectively increased the size of the experimental treatment plots. Plots were approximately 5 m away from one another parallel to an incoming tide. On 4 July 2001 each site was visited and the disturbed plot dug out to a depth of 10 cm, marking the beginning of the experiment. For the muddy sand and mud sites, planks of wood were used to delineate the plots. This prevented excessive trampling of the surrounding sediments and acted as a platform to aid the process of digging out the sediments. In each case, sediment was removed from the plot and transported at least 50 m away from the experimental area, in order to prevent the immediate recolonization of displaced organisms back into the disturbed areas. Thus the effect of the disturbance treatment was to create a pit in the sediment surface, removing most of the existing infaunal organisms and uncovering a previously uninhabited layer of sediment. Each of the disturbed plots at the 16 sites was dug out within one tide by a team of workers operating simultaneously at different sites.

Sampling occurred during spring tides through necessity, in order to allow each site to be visited and sampled within one tidal exposure. Nevertheless, on certain occasions sampling had to occur over two consecutive tides due to bad weather and/or light conditions. Sampling occurred approximately 15, 35, 63, 105 and 213 days (from July to January) following the experimental disturbance. The experimental plots were not sampled prior to the application of the disturbance treatment since the chances of treatment and control plots differing systematically from one another was thought to be minimal. The treatment plot at one site in the Foryd Estuary was lost at some point after day 63, which coincided with storms that occurred in September.

sampling procedure

On each sampling date the following samples and measurements were taken at each of the 16 sites.

Macrofauna

Four 10-cm-diameter cores were taken to a depth of 10 cm within both the control and disturbed plots. Samples were sieved through 0·5-mm sieves with sea water and the contents were fixed with 4% buffered formalin solution. Samples were subsequently washed and identified to the lowest possible taxonomic level in the laboratory.

Sediment

Depth of disturbed pits: the depth of water remaining in the disturbed plots was measured with a ruler to ±1 mm from the sediment surface to the air/water interface. Although the exact levels of water remaining in pits would have been affected by the length of time of tidal emersion, as the depth measurements were taken consistently at the same stage of the tide for all sites, variability as a result of this factor was considered to be minimal. Three transects of four measurements were taken across the length of each of the disturbed plots.

Organic content and water content: one 10-cm-diameter core of sediment was removed from each plot for further analysis in the laboratory. Three 5-cm3 cores were taken from each of these samples with a syringe corer and used for the determination of water and organic content. Samples were placed in preweighed crucibles and wet weight recorded. Samples were then dried to constant weight at 80 °C, re-weighed and organic material was combusted in a muffle furnace at 550 °C for 6 h. Wet, dry and ash-free dry weight values were used to calculate water content and organic content of the sediment.

statistical analysis

Physical parameters

Differences in physical parameters between treatments were tested using ancova, with percentage silt and clay content (our main habitat descriptor) assigned as a covariate. The rate of change (slope) of log10 depth values over time for each disturbed plot was calculated and regressed on the percentage silt and clay content values. These rates were based on the time taken for the pits to reach a depth of zero (i.e. the rate of infilling). Simple linear regression was undertaken to examine the relationship between infilling rate and sediment type.

Community parameters

Data from the four cores collected from control and disturbed plots at every site and sampling occasion were amalgamated prior to any statistical analyses as within-treatment variability was not the focus of this experiment. Diversity indices were calculated using the primer software (Clarke & Warwick 1994). Analyses of covariance were performed on the number of individuals, number of species and Pielou's evenness on each sampling date. These indices were chosen following preliminary graphical analyses in which clear treatment differences could be seen for individuals, species and evenness. The raw data revealed that recovery occurred gradually through the recolonization of treatment plots with species commonly found in ambient sediments (i.e. there was no evidence of an influx of differing species into disturbed plots), and further analyses were restricted to the total numbers of individuals as these data showed the clearest response to the disturbance treatment (i.e. the clearest recovery trajectory).

Any changes in the total numbers of individuals in the disturbed treatments occurred concomitantly with natural fluctuations in the numbers of individuals in control plots over time. These fluctuations can complicate the interpretation of the response of infauna in disturbed treatments relative to control plots. Therefore, we also expressed our data as the relative response (negative or positive) of the numbers of individuals in treatment plots compared with control plots with the aim of facilitating the interpretation of our results.

In order to investigate the use of physical habitat parameters in the prediction of the response of biological communities following the disturbance treatment, we undertook a series of regression analyses. We regressed the relative numbers of individuals in disturbed plots compared to control plots over time for each site to obtain the rate of change (slope) of the relative difference between treatment and control plots. These slope data provided an estimate of the recovery rate of the numbers of individuals − which we will refer to hereafter as the biological recovery rate (number of individuals day−1) − in the disturbed plots at different sites. We also used the resulting regression equations to calculate the predicted number of days to ‘recovery’ of numbers of individuals in the disturbed communities, which we defined as having occurred when the relative number of individuals in each disturbed plot compared to the associated control plot was equal to zero.

Following these initial analyses, clean sand sites were omitted from further univariate analysis since the data for these sites did not conform to simple linear regressions. ‘Recovery rate’ and the predicted number of days to recovery at each site were regressed on the percentage silt and clay values. The categorization of sites using their percentage silt and clay content may not represent adequately the ecologically important characteristics of the sediment habitat. In order to remove this effect we additionally examined the relationship between the rate of infilling of disturbed plots and the ‘recovery’ rate of the total number of individuals at each site.

analysis of multivariate data

Physical parameters

Multivariate analyses were carried out on environmental data sets in order to investigate differences between disturbed and control treatments over the duration of the experiment. Multivariate analysis should be more sensitive to overall changes in the environmental data as all the measured environmental parameters influence the similarity matrix upon which further analysis is based. A similarity matrix of the data from each sampling occasion was produced using normalized euclidean distance. Non-metric multidimensional scaling (Kruskall & Wish 1978) was carried out to produce a two-dimensional ordination plot of the similarity between treatment plots in terms of their physical characteristics. On such ordination plots, the distance between samples corresponds to the relative similarity of the multivariate data that defines them (in this case, a number of environmental parameters), such that samples that are far apart are more dissimilar than those that lie near to one another on the plot. Tests for significant differences between the physical habitat variables for disturbed and control plots on each date were performed using one-way analysis of similarities (anosim) tests (Clarke 1993). anosim assesses significant differences between groups of replicates against a series of random simulations, resulting in the calculation of a test statistic (R), which can range from −1 to 1. R will be close to 1 when replicates are very dissimilar and approach 0 as they become more similar. A very negative R-value is unusual and indicates that between-group similarity is higher than within-group similarity. Both the MDS plots and the anosim tests were performed on the data including and excluding depth data. This comparison was undertaken to investigate the influence of the zero values recorded for depth in control plots that would have a large influence on the outcome of any multivariate analysis.

Community parameters

Similarity matrices were produced using the Bray–Curtis index of similarity on fourth root-transformed data for communities in disturbed and control plots on each sampling occasion. Significant differences in community composition between control and disturbed treatments were determined using one-way anosim tests for each sampling date.

The relationship between measured habitat parameters and benthic community characteristics during the experiment was investigated using the bioenv procedure. This is an exploratory test that performs a rank correlation of two similarity matrices (biotic and environmental), and successively tests every possible combination of environmental parameters to indicate which set of habitat attributes best explains the observed multivariate community patterns. bioenv was run both on the whole data set, and then on control and disturbed community and environmental data separately.

Results

Percentage silt and clay values ranged from 1·6% to 63·13% and organic content varied between 1·2% and 5% (Table 1). A number of species and taxa were present at the majority of sites, including Pygospio elegans, Tubificoides benedii, Macoma balthica, Corophium spp., Cirratulidae and Capitellidae. Abundance of these common species and taxa varied over different sediment types. However, all showed a decrease in abundance in disturbed plots on day 15 that was less pronounced on the last sampling occasion, 213 following the disturbance treatment (Table 1).

Analysis of covariance (ancova) revealed that there were no significant differences between control and disturbed treatments for percentage organic content (F1,29 = 2·83, P= 0·103) or water content (F1,29 = 2·88, P= 0·101) on the first sampling date (2 weeks after the initial disturbance treatment). No significant differences were found between treatment and control plots for any of these parameters for the remainder of the experiment.

ancova tests were not suitable to consider differences between depth of pits of the treatment and control plots, as control values were always zero. Water depth in all disturbed plots decreased gradually over the course of the experiment (Fig. 2). All pits began to refill immediately, but disturbed plots on sandy sites were consistently shallower than those in the muddier sediments and took less time to refill. All of the clean sand pits had completely refilled by day 105, whereas pools of water were clearly distinguishable in four of the eight disturbed muddy sand and mud sites after day 213. The regression analysis of infilling rate against percentage silt and clay content of sediments (Fig. 3) revealed that as the percentage silt and clay content of the substratum increased, the rate of infill decreased (r2 = 0·37, F1,14 = 9·88, P = 0·007).

Figure 2.

Change in the mean log10 depth (mm) of disturbed plots for the four sediment groups over the course of the experiment. Filled squares = clean sand, open diamonds = silty sand, open triangles = muddy sand, filled circles = mud. Error bars represent SE values.

Figure 3.

Sediment infilling rates for individual sites plotted against the percentage silt and clay content at each site.

The influence of the zero values recorded for depth in control plots had a large influence on the outcome of the multivariate analyses investigating differences in physical parameters in control and disturbed plots (Fig. 4a,b, Table 2). When the depth data were included in the analyses, anosim revealed significant differences on the first three sampling occasions, and a differentiation of control and disturbed plots was discernible on the MDS plot. However, when depth data were excluded no significant differences were found at any time during the course of the experiment (Table 2) and no discernible pattern was found on the MDS plot.

Figure 4.

Two-dimensional non-metric multidimensional scaling plots for physical parameters at all sites on day 15: (a) all environmental variables, (b) all environmental variables excluding depth. Plots within the ordination are coded as (c) = control and (d) = disturbed treatments. The stress values indicates the accuracy with which the 2-d plot represents the high-dimensional relationship between the samples. A value of below 0·1 indicates a very accurate representation.

Table 2.  Results of anosim tests for significant differences between physical parameters in disturbed and control treatments on each sampling occasion. a = test run for all environmental variables (organic content, water content and depth of plot); b= test run excluding depth data
Day R-statisticP
15a 0·2180·003
b−0·0380·92
35a 0·1690·006
b−0·0340·87
63a 0·1210·011
b−0·0210·67
105a 0·0420·11
b−0·050·99
213a−0·0340·89
b−0·0360·86

Analysis of covariance revealed that there were significant differences between disturbed and control communities in terms of the total number of species (Table 3), the total number of individuals (Table 4) and Pielou's evenness during the experiment. The total number of species and number of individuals were significantly different between treatments on day 15 (Fig. 5a,b), 35 and 105, whereas significant differences between the values of Pielou's evenness were found only on days 15 (Fig. 5c) and 35. There were no significant differences found between control and disturbed treatments on day 63, even though there were significant differences following this date. In 11 of the 16 sites there was a decrease in numbers of individuals within control plots between days 35 and 63, while the numbers of individuals increased within disturbed plots over the same period. However, following this date (between days 63 and 105) the numbers of individuals increased much more rapidly within control plots than in the disturbed plots. Because this trend is apparent at the majority of sites (Fig. 6a–d) it suggests that some aspect of the disturbed plots has prevented the level of faunal colonization evident in control plots between days 63 and 105.

Table 3. ancova test results for differences between the total numbers of species in disturbed and control plots on each sampling occasion using the percentage silt and clay content at each site as a covariate (no significant interactions existed for any of these tests)
DaySourced.f.SSMSFP
15% silt and clay 1 94·95 94·95 8·41  0·007
Treatment 1175·78175·7815·57< 0·001
Residual29327·49 11·29  
35% silt and clay 1 58·58 58·58 7·8  0·009
Treatment 1 84·5 84·511·25  0·002
Residual29217·79  7·51  
63% silt and clay 1 70·26 70·26 5·17  0·06
Treatment 1 16·53 16·53 1·22  0·28
Residual29393·93 13·58  
105% silt and clay 1107·22 98·78 8·12  0·008
Treatment 1105·57105·57 8·68  0·006
Residual28340·75340·7512·17 
213% silt and clay 1 20·18 20·15 1·72  0·2
Treatment 1  0·00  0·00 0·00  0·99
Residual28328·79 11·74  
Table 4. ancova test results for differences between the total numbers of individuals in disturbed and control plots on each sampling occasion using the percentage silt and clay content at each site as a covariate (no significant interactions existed for these tests)
DateSourced.f.SSMSFP
15% silts/clay 1 0·55 0·55 0·41  0·53
Treatment 127·927·921·02< 0·001
Residual2938·49 1·33  
35% silts/clay 1 2·14 2·14 1·97  0·17
Treatment 114·8414·8413·63  0·001
Residual2931·57 1·09  
63% silts/clay 1 0·27 0·27 1·12  0·30
Treatment 1 0·47 0·47 1·95  0·17
Residual29 6·98 0·24  
105% silts/clay 1 0·97 1·07 6·21  0·02
Treatment 1 1·49 1·49 8·67  0·006
Residual28 4·82 0·17  
213% silts/clay 1 2·86 2·923·22< 0·001
Treatment 1 0·1 0·1 0·83  0·37
Residual28 3·49 0·12  
Figure 5.

Community variables for control (filled squares, solid trend line) and disturbed (open squares, dotted trend line) plots at different sites on day 15; (a) total number of species, (b) total number of individuals, (c) Pielou's evenness (J′).

Figure 6.

(a–d) Changes in mean total numbers of individuals in control (filled circles) and disturbed (open circles) plots for the four sediment groups: (a) clean sand, (b) silty sand, (c) muddy sand, (d) mud. (e, f) Changes of numbers of individuals in each of the disturbed plots relative to control plots for the four sediment groups: (e) clean sand, (f) silty sand, (g) muddy sand, (h) mud. Different symbols represent individual sites.

The total number of individuals in disturbed plots relative to control plots increased gradually over the duration of the experiment (Fig. 6e–h). For clarity, data have been plotted for the four sediment type groupings (clean sand, silty sand, muddy sand and mud). Patterns of recovery of the number of individuals in the clean sand sediments are difficult to elucidate and fluctuate around the control levels from day 35 onwards. More consistent patterns of increasing numbers of individuals are apparent in the other sediment groupings, although certain sites showed only marginal signs of recovery in their numbers of individuals over the duration of the experiment.

The rate of change in the relative numbers of individuals over time at all sites (excluding clean sand) against the percentage silt and clay content values suggested a higher rate of recovery in the sediments with a higher proportion of sand (Fig. 7a), but the relationship was not significant (r2 = 0·22, F1,8 = 2·21, P = 0·175). Similarly, although the predicted time to recovery of communities from different sediments followed a trend of increasing recovery time with increasing silt and clay content (Fig. 7b), this was not significant (r2 = 0·12, F1,9 = 1·03, P = 0·34). However, there was a significant relationship between the rate of infilling and the rate of ‘recovery’ for each site (Fig. 8) (r2 = 0·65, F1,10 = 21·39, P < 0·001). Pits that filled in more rapidly were associated with a higher rate of community recovery. Hence, single measures of physical parameters do not adequately correlate with biological recovery rate, while direct measures of habitat recovery (infilling rate) represent the combination of habitat parameters that truly represent the nature of the environment.

Figure 7.

(a) Biological recovery rate, no. individuals day−1 (derived from a regression of the numbers of individuals at each site and sampling occasion) at the different experimental sites. (b) Predicted number of days to recovery at the different sites (derived from the regression of the numbers of individuals at each site and sampling occasion). Recovery is defined as the point in time when numbers of individuals within disturbed plots are equal to the numbers of individuals in control plots.

Figure 8.

Biological recovery rate (no. individuals day−1) and physical recovery rate (infilling rate in mm day−1) at all sites. Data points are coded for the defined sediment groups: Open squares = silty sand, open circles = muddy sand, open triangles = mud. The linear relationship between physical and biological recovery rates is not reflected by the defined sediment groups.

A priori anosim tests revealed significant differences between treatments on days 15 and 35 when all sites were analysed together for each sampling date (Table 5). We then re-ran tests on sediment subgrouping (CS, SS, MS and M). Significant differences between treatments in MS sediment types were found on days 15 and 35. On day 35 significant differences were also found between treatments for SS and M groups. When the analysis was repeated and the data were left untransformed, significant differences were also apparent between treatments in CS sites on day 35, which suggests that this effect was due to differences in the dominant fauna within these communities.

Table 5.  Results of anosim tests for significant differences between disturbed and control communities on each sampling occasion
DayR-valueP-value
15 0·1860·001
35 0·4370·001
63−0·0390·82
105 0·0240·24
213−0·0480·91

A bioenv analysis did not produce a very high level of correlation between any suite of environmental variables. The highest value of rho obtained was 0·444 (organic content and penetration) on the first sampling occasion (Table 6). Water depth within disturbed plots was not identified as an important variable that explained community composition at any time during the experiment. Re-running bioenv on control and disturbed communities separately produced better correlations for control communities (Table 7). The percentage of organic content consistently produced the highest correlation value with the exception of the last date when penetrometer values and water content gave the best correlation. Nevertheless, values varied from a fairly high rho of 0·622 to only 0·392 over the course of the experiment. Correlation values between disturbed communities and physical parameters were generally lower than those found for the control communities (although there was no significant difference between treatment and control values overall) and tended to include a selection of the physical parameters measured (Table 7). However, there was no overall significant difference between rho values for separate treatments (anova, F1,8 = 2·73, P= 0·136). Depth of pits was important on days 15, 35 and 63, and organic content was important on days 105 and 213. However, on these last dates correlation values were so low (0·242 and 0·075) that no relationship between any of the parameters and community composition of the disturbed communities can be inferred.

Table 6.  Results of the bioenv analyses for each sampling occasion. Rho (ρ) values indicate the degree of correlation shown by those environmental parameters that best explain the multivariate community patterns observed in control and disturbed community data
DayEnvironmental parameter(s)ρ-value
15Organic content, penetration0·444
35Organic content0·275
63Organic content0·421
105Organic content0·385
213Penetration0·246
Table 7.  The results of bioenv analyses for separate treatment groups on each sampling occasion. Rho (ρ) values indicate the degree of correlation shown by those environmental parameters that best explain the multivariate community patterns observed in control and disturbed community data. Environmental parameters 1 = percentage organic content, 2 = penetration values, 3 = percentage water content, 4 = water depth.
DayTreatmentEnvironmental parameter(s)ρ-value
15Control10·606
Disturbed2,3,40·602
35Control10·464
Disturbed40·381
63Control10·403
Disturbed1,2,3,40·362
105Control10·622
Disturbed1,30·242
213Control2,30·392
Disturbed1,20·075

Discussion

We were unable to detect any significant changes in the measured environmental parameters following the disturbance treatment. The methods we applied to measure habitat characteristics were techniques used commonly by ecologists to characterize soft sediment habitats. Other studies have detected elevated levels of organic material within the depressions of disturbed areas as a result of increased deposition of fine particles following the alteration of flow over the seabed (e.g. Thistle 1980; Van Blaricom 1982; Oliver & Slattery 1985; Savidge & Taghon 1988). Despite the persistence of the pits in this experiment we could detect no increase in the organic matter within the depressions. In previous work in the Menai Strait we found no elevation in the proportion of fine particles or organic material within depressions following physical disturbance in a sand flat habitat (Dernie et al. 2003). As the scale of our disturbance is similar to that used by Van Blaricom (1982) and Savidge & Taghon (1988), this would seem to indicate that there was either a lack of supply of organic/fine particles to the seabed at our sites or that the hydrodynamic regime was unsuitable for their deposition.

The fact that we did not detect any changes to the measured structural parameters of the sediment is more likely to be a reflection of the lack of sensitivity of standard granulometric methods used by benthic ecologists rather than an indication that there were no changes to habitat parameters following physical disturbance. In addition, as sampling could not begin until 15 days after the initial disturbance, it is possible that the sediment had undergone partial restoration within that time, and thus any differences between control and disturbed plots may have been undetectable. Nevertheless, the present study questions the value of quantifying gross sediment characteristics in studies of physical disturbance. Relevant changes in sediment structure may be difficult to detect using currently available techniques, and may occur in only the top 10 mm of the sediment or at the sediment water interface. Changes to sediment structural properties may not be detected through granulometric analysis (Snelgrove & Butman 1994). In addition, as the depth to which many sediment cores are taken is at least 5 cm and beyond this is likely to mask changes occurring in the surficial sediments.

Although our bioenv analysis indicated a fairly high correlation between organic content values and the community composition of control sites, it did not reveal any environmental parameters that might explain the long-term biological differences between disturbed and control plots. Because depth was always zero in control plots, this indicates that there was not a consistent directional change in community composition following disturbance, as depth values should have correlated to such a consistent pattern. The increase in the importance of organic content in correlation values for disturbed plots as the experiment progressed might reflect the gradual restoration of communities to control levels. However, overall this analysis highlights the variability in community responses following disturbance in comparison to that of the majority of the environmental parameters, which remained constant over the duration of the experiment.

The rate of infilling of disturbed plots had a positive relationship with the increasing percentage silt and clay content, such that pits at muddier sites generally took longer to infill (and thus had less negative infilling rates) than those in sandier sites. Presumably this is related largely to the hydrodynamic regime at a site, for example resulting in increased transport of material in sandier sediments where local hydrodynamic forces over the sea bed are stronger and sediments are less cohesive. The two sites that had the lowest rates of infilling were both muddy sand sediments. These sites showed little sign of physical recovery over the duration of the experiment. In addition, the rate of increase of numbers of individuals was slower than at any other sites. We suggest that in these intermediate sites (i.e. where both sand and silt/clay particles significantly contribute to the sediment composition), a reduced infilling rate in comparison to muddier sediments might be related to the physicochemical properties of the substratum. Sediments with a high proportion of clay particles can act as fluids (when water content is reasonably high) such that pit refilling may have occurred largely through the slumping of the sediments into the depression. In sites with a relatively higher sand content, a mixture of chemical attraction between the clay fraction of the sediment in addition to the drainage provided by the sandier fraction could result in a more stable sediment structure and thus a slower rate of infilling in comparison to muddier sites.

Community composition showed clear differences following the disturbance treatment. The total numbers of individuals and species were reduced significantly immediately post-disturbance. Pielou's index of evenness initially increased in the disturbed communities, indicating that there was reduced dominance of communities by certain species. Communities within the control plots consistently showed a trend of decreasing evenness with increasing percentage silt and clay content as a result of the very high abundance of certain species found within the muddier substrata. This is common in soft sediment habitats in which the higher organic content of muddy sediments can support large numbers of small opportunistic species that can tolerate the conditions of low oxygen in these habitats (Pearson & Rosenberg 1978).

Significant differences in the total numbers of individuals and the number of species were found on days 15, 35 and 105. All the observed significant differences were attributable to lower numbers of species and individuals in disturbed treatments. There was no indication of an influx of opportunistic species into disturbed areas at any of the 16 sites. The sudden lack of a significant difference in community composition between treatments on day 63 is interesting and suggests that long-term monitoring of communities following disturbance events may be important in order to assess accurately the recovery trajectories of communities. We suggest that the drop in the total numbers of individuals in control plots between days 35 and 63 at the majority of sites was related to natural disturbance through persistent stormy weather over this period. This may have led to a redistribution of benthic organisms such that while numbers decreased within control areas, disturbed plots received an influx of infauna so that overall abundance increased within the pits. If this is the case, it is interesting that community composition in control and disturbed plots did not follow a similar trajectory after day 63. The numbers of individuals increased much more rapidly within control areas than in the disturbed plots. This suggests a longer-term effect of disturbance on the treatment plots that prevented the increase in numbers of individuals that occurred naturally in adjacent control areas over this time period, despite both treatments subsequently experiencing physical disturbance as a result of stormy weather.

Although it was not possible to predict the recovery rates of assemblages based on percentage silt and clay content of the sediment, there was a good relationship between recovery rate and infilling rate. This suggests that although biological recovery rate is linked to the restoration of the sedimentological characteristics of the habitat, this process is not well described by physical characteristics such as grain size distribution alone. The relationship between infilling rate and percentage silt and clay content (r2 = 0·37), although statistically significant, was weaker than the relationship between infilling rate and biological recovery rate (r2 = 0·65). The two sites that had the slowest rates of both physical and biological recovery consisted of muddy sand sediments. This may seem counter-intuitive, as it has often been assumed that recovery will be most rapid in sandy sediments (Ferns et al. 2000), and that there may be some linear decrease in recovery rates with increasing silt and clay content. However, our results support the work of Collie et al. (2000) whose meta-analysis revealed that muddy sand habitats had the slowest rate of community restoration following fishing disturbances (in soft sediment habitats).

Physical aspects of the habitat include local hydrodynamics, particle sizes and sediment structure. The presence of clay particles within sediments results in chemical cohesion between grains. In addition to these factors the presence of organisms may enhance stability (e.g. Grant & Gust (1987), Underwood & Paterson (1993), Thrush et al. (1996)) or act to destabilize sediments (e.g. Widdows et al. (2000)). We suggest that recovery rates are related to a variety of physical, chemical and biological factors that differ in relative importance at different sites (Fig. 9). These parameters cannot be encapsulated by granulometric descriptions alone; indeed, we question the utility of such measurements in all but the coarsest of large scale habitat descriptions.

Figure 9.

Schematic diagram of the recovery rates within disturbed depressions. In habitats with a low silt and clay content, physical processes dominate recovery, which may be rapid. With increasing silt and clay content, biological recovery rate is reduced as chemical and biological factors act to increase sediment stability. In silt- and clay-dominated sediments the recovery rate increases slightly as the substratum acts as a fluid and aids recovery of the biota through ‘slumping’ of adjacent sediments and organisms into disturbed pits.

It was particularly difficult to elucidate any consistent patterns of recovery that occurred in clean sand sediments. Infilling rates were particularly rapid, having occurred at one site by the first sampling occasion. Biological recovery also appeared to be rapid at these sites, but the lower numbers of individuals in this sediment type meant that the statistical power to detect any changes was low, and hence we were unable to fit a relationship that predicted recovery rates for these sites. Nevertheless, we postulate that infilling rates (and presumably biological recovery) in sediments that are made up of unconsolidated clean sands are dominated by physical processes (i.e. the local hydrodynamic regime). Recovery in such habitats may be rapid, as many characteristic sandy species are relatively mobile (Posey et al. 1996; Collie et al. 2000). Additionally, bed load transport may be an important mechanism for recolonization of disturbed areas (Hall & Harding 1997). However, if this is the case recovery rates may be extremely variable due to their dependence on local hydrodynamics, which will be very strongly affected by changing weather conditions (e.g. Hall, Basford & Robertson 1990). Thus, where chemical and biological factors are less important components of the sediment system, it may be very difficult to accurately predict recovery trajectories of the benthic assemblage. Nevertheless, it is clear that physical recovery of these habitats is rapid and occurs within 100 days post-disturbance at the scale of the present study and concurs with the findings of Collie et al. (2000).

This experiment was performed at a scale relevant to various anthropogenic activities including bait-digging, hand collection of cockles and hydraulic dredging. Biological recovery over such scales is likely to occur largely through the migration of adults (i.e. post-larval stages) into disturbed areas, either by active migration or passive transport from adjacent undisturbed areas (Savidge & Taghon 1988). However, even studies performed at a larger scale have highlighted the importance of bed load transport as a recovery mechanism for certain species (Beukema et al. 1999; Peck et al. 1999). The time of year at which a disturbance occurs may have a strong influence on the mode of recovery and recovery rate of a disturbed community. This will relate both to the difference in recruitment supply of larvae and adult infauna at different times of year and to changes in the physical regime at a specific location, which may be significantly affected by, for example, prevailing wind conditions that control levels of wave disturbance. The importance of larval recruitment as a mechanism for recolonization increases in importance with the size of a disturbed patch (Smith & Brumsickle 1989). Thus, care must be taken in applying this relationship between biological and habitat recovery rates to physical disturbance events operating at much larger scales (see Thrush et al. 1995, 1997a,b). Nevertheless, Collie et al.'s (2000) study indicated that communities in muddy sand habitats exhibited slower recovery rates than those in muddier sediments based on those experiments (with scales of disturbance varying between 1 and 5000 m2) included in their meta-analysis. Clearly, at all scales the hydrodynamic regime will affect the recolonization rate of a disturbed area by influencing the deposition of both the adults and juvenile/larval stages of infaunal species regardless of which life stages dominate the recolonization process. Further studies of comparable large-scale disturbances in a variety of sublittoral habitats are required to verify the applicability of our findings to the management of large-scale disturbance impacts. Nevertheless, using the infilling rate of physically disturbed patches of sediment has the potential to become a useful management tool in assessing the impacts of a disturbance in a given habitat for a variety of perturbative activities of interest to managers of coastal habitats.

Acknowledgements

This work was funded by the Natural Environment Research Council Grant GT04/99/MS/291. The authors would like to thank Helen Beadman for assistance during fieldwork.

Ancillary