Testing hypotheses about management to enhance habitat for feeding birds in a freshwater wetland

Journal of Environmental Management (2001) 62, 375–388 doi:10.1006/jema.2001.0441, available online at http://www.idealibrary.com on

Testing hypotheses about management to enhance habitat for feeding birds in a freshwater wetland M. Lindegarth†‡* and M. G. Chapman†

The level of water was manipulated in a freshwater wetland, with the aim of enhancing abundances of benthic animals and, ultimately, improving habitat for feeding birds (Japanese Snipe, Gallinago hardwickii). We tested whether these actions had the predicted and desired effects on benthic animals, by contrasting changes in two managed locations to one control location which was left unmanipulated. The number of taxa and abundances of chironomids decreased strongly and significantly in the manipulated locations, while the abundance of oligochaetes appeared to vary in a seasonal manner. Temporal variability of the structure and composition of assemblages was also increased in manipulated locations. Such effects have previously been suggested to indicate stress in benthic assemblages. Therefore, in contrast to what was predicted, managerial actions made benthic fauna less abundant and thus, less suitable as habitat for feeding birds. Several general lessons can be learned from these results. (1) Effects of managerial actions like these are difficult to predict a priori and can only be reliably evaluated within an experimental framework. (2) Because abundances of animals vary naturally, evaluations of managerial actions must include appropriate spatial replication. (3) Sampling at hierarchical temporal scales is important, because abundances of animals may vary in an unpredictable manner at short temporal scales and because changes in temporal variability may be a symptom of stress. (4) Combined use of uni- and multivariate techniques provides a comprehensive set of tools to assess the effects of restoration and creation of new habitats. Finally, these results emphasise the need for clear predictions about desired outcomes and specific experimental plans about how to test whether the desired results were achieved, before managerial actions are taken. Although this is often very difficult to achieve in real situations, it is necessary for practices of management to evolve on the basis of sound empirical experience.  2001 Academic Press

Keywords: wetlands, management, benthic fauna, sampling design, large-scale experiment, freshwater, water-level, Japanese Snipe, Australia.

Introduction Modification and destruction of natural habitats with their unique values of biodiversity and important ecological functions is an ongoing process. There is an increasing need for efficient procedures to protect and conserve natural habitats, in addition to those to restore or rehabilitate damaged habitats and create new habitat to replace what is lost (Anderson, 1995; Bradshaw, 1993, 1996; Gilbert and Anderson, 1998; McMahon, 1998). Such activities often require manipuEmail of corresponding author: mats.lindegarth@tmbl. gu.se 0301–4797/01/080375C14 $35.00/0

lation of areas of habitat or processes at large spatial and temporal scales. This has two important consequences for practical environmental management: (1) management becomes expensive because of the large scales involved, and (2) there are large uncertainties as to whether management will have the predicted effects because of the lack of clear understanding between results from small-scale ecological experiments and largescale processes (e.g. Havens and Aumen, 2000). Ecologists are, therefore, increasingly aware that effects of different options for management must be rigorously evaluated, i.e. management need to be done within an experimental framework (e.g. Carpenter, 1990; Underwood, 1990; Walters and Holling,

Ł Corresponding author † Centre for Research on Ecological Impacts of Coastal Cities Marine Ecology Laboratories, A11 University of Sydney, NSW 2006 Australia ¨ o¨ ‡ Present address: Tjarn Marine Biological Laboratory, S-452 96 ¨ Stromstad, Sweden Received 27 March 2000; accepted 20 February 2001  2001 Academic Press

376

M. Lindegarth and M. G. Chapman

1990; Lubchenco et al., 1991). Only by testing rigorous hypotheses can managers (and the rest of the community) evaluate causal links between management and changes in abundances of animals and plants and whether time and money spent on managerial actions are worthwhile and improve future practices. This contrasts with routine monitoring where the data are not necessarily collected in a way that allow such hypotheses to be tested (Underwood, 1991a). Like any manipulative experiment, inferences about causality can only be justified logically if natural sources of spatial and temporal variability can be separated from the variability caused by the experimental treatment, i.e. the managerial action. This usually requires complex experimental designs, which include replication at appropriate spatial and temporal scales, preferably before and after the managerial changes and simultaneous estimates of natural variability which are not affected by managerial actions (e.g. Green, 1979; Underwood, 1997). Analysis of variance (ANOVA) provides a flexible framework within which hypotheses can be tested for such complex experimental designs, sizes of effects can be estimated and the power associated with Type II errors (i.e. not finding a significant effect when there was one) can be calculated. Furthermore, these analytical methods have been developed for situations where there is potential spatial confounding because there is only one potential impacted location (Underwood, 1991b, 1992, 1993, 1994), and when there are no data from before the impact or management (Glasby, 1997) and when it is not known whether the disturbance may cause a press or pulse perturbation (Glasby & Underwood, 1996). Such methods have been shown to be very useful for detecting impacts. In a similar way, there has been recent development in the use of various multivariate methodologies to detect impacts on assemblages, e.g. ANOSIM and MDS plots (Clarke, 1993), taxonomic distinctness (Warwick and Clarke, 1995), abundance and biomass curves (Warwick, 1986). Although these may be particularly useful in that they examine changes to the entire assemblage, they cannot yet be used in complex, multifactorial experimental designs to identify the relative magnitude of effects of different factors and their interactions.

Restoration has lagged behind many other forms of management when it comes to a rigorous scientific framework (e.g. Bradshaw, 1993; Hobbs and Norton, 1996; Pastorak et al., 1997; Chapman and Underwood, 2001). The comparison between one impacted and one control site cannot identify impact because of problems of pseudoreplication (Hurlbert, 1984; Underwood, 1994). For the same reasons, neither can such comparisons in programmes of restoration (e.g. Thompson et al., 1995) provide anything but the most superficial evidence for restoration. Neither is it always clear what are the relationships between the measurements taken and the changes expected by restoration. Therefore, Moy and Levin (1991) did not explain how measures of grain-size, diversity of infauna and abundances of one species of fish could be used to assess structural and functional restoration of salt marshes. It is fortunate that these simplified approaches to restoration are changing for the better. For example, Short et al. (2000) develop a rigorous protocol for identifying scientific criteria for measuring success of restored wetlands, emphasising that without defensible scientific methods and statistical tests, restoration remains a ‘seat of the pants’ process (see comments by numerous other authors in the same volume). Nevertheless, neither appropriate multifactorial analyses of variance, nor multivariate techniques, are yet being used widely in studies of restoration or rehabilitation of habitat, although the logic and experimental designs needed to measure restoration are similar to those for detecting environmental impacts (Chapman, 1999). Wetlands have recently received considerable attention from environmental managers because of their widespread loss to development and so-called reclamation (Kusler and Kentula, 1990; Adam, 1995; Zedler, 1996; Zedler et al., 1998). For example, many attempts have been made to manage wetlands in ways that will make them more attractive to birds. Such actions involve techniques for manipulating the extent of vegetative cover, depth profile and flooding patterns (Murkin et al., 1981, 1997, Murkin and Kadlec, 1986a,b; Hill, 1989; Rehfisch, 1994; Falk et al., 1994; H¨otker, 1994; Reitan and Sandvik, 1996). Manipulations of levels of water have been proposed to provide greater areas of shallow feeding-habitat

Managing habitat for wading birds

and to increase the abundances of invertebrate animals, which are important prey for many of these birds (Hill, 1989; Rehfish, 1994). Abundances of benthic, invertebrate animals in soft-sediments of shallow aquatic habitats are characterised by large temporal and spatial variability (e.g. Morrisey et al., 1992a,b; Downes et al., 1993; Thrush et al., ´ and B´ır´o, 1998) which is often 1994; Specziar unpredictable. It has also been pointed out that strategies to provide an attractive habitat for the birds may not correspond to conditions which maximise production of invertebrates (Green and Hilton, 1998). It is clear that the most efficient way to resolve ambiguities about how manipulations of the level of water affect abundances of invertebrates and use of habitat by birds, is to test explicit hypotheses about effects of actual manipulations, i.e. treat and analyse the managerial activity as an ecological experiment (Cairns, 2000). These tests should include the best experimental designs and analytical methodologies available (Chapman, 1999; Cairns, 2000; Kentula, 2000; Thom, 2000). In this study, we test the hypothesis that controlled drainage and subsequent refilling of water in a wetland would cause increases in abundance and diversity of benthic invertebrates. These managerial actions were part of a large program to restore estuarine wetlands on polluted and previously developed sites (Burchett et al., 1999). The general aim of the program was to improve the extent and/or quality of potential habitat for a particular species of bird, the Japanese Snipe (Gallinago hardwickii Gray; effects on the population of birds are not evaluated here). This action was based on anecdotal evidence that abundances of the birds had decreased in response to loss of habitat. Previous observations had indicated that the birds generally avoid flooded spots (Naarding, 1985) and feed during the day in soft earth and mud at the edges of freshwater, wetland habitats (Higgins and Davies, 1996; Todd, 1998). Control of water-levels was intended to create areas of emersed sediment along the edges of a shallow pond which contained appropriate food. Therefore, the success of the manipulation depended on a predicted increase in the area available to feed and/or in the amount of potential food, predominantly chironomid larvae and oligochaetes, in these manipulated areas (Frith et al., 1977).

Because of the logical requirements for unconfounded tests of appropriate hypotheses and the practical constrains of the areas available to serve as control or reference locations, the experimental design was necessarily complex and was based on asymmetrical comparisons. There were only two ponds available in the surrounding area, Wharf Marsh and Bennelong Pond. These were small, shallow (<1 m deep) freshwater marshes, approximately 3 km apart. Because these were the only two ponds available, there were no replicate control ponds, against which to contrast Wharf Marsh (the pond that was being altered). Therefore, any changes to the benthos in Wharf Marsh would not strictly be able to be attributed to the managerial actions (Underwood, 1992; Hurlbert, 1984). Furthermore, when there is no prior information about the time scale at which changes in response to management might be expected, it is necessary to measure changes in abundances or diversity at several different times (Underwood, 1992, 1994). Therefore, because the effect of drainage might be long-term (seasonal) or short-term (weekly), sampling was done in one period before the wetland was being manipulated and three periods after varying manipulations (see Methods). These periods were approximately 6 months apart. Within each of these periods, the fauna were sampled twice at approximately 2-week intervals, to give measures of short-term changes within each period. The situation described in this study demonstrates many problems which are often encountered in practical environmental management. For example, there was a limited number of sites available. Only one control location (i.e. which resembled the location to be restored, but where the level of water was not manipulated) and no reference location (i.e. which represented the desired state or end-point of restoration) was available (Chapman, 1999). Furthermore, it was not possible to sample more than one period before the water-level was manipulated. Nevertheless, this design meets minimal requirements to allow logical tests of hypotheses about the effect of this management strategy on abundances of these benthic invertebrates. It also illustrates the

377

378

M. Lindegarth and M. G. Chapman

flexibility and utility of these types of asymmetrical analyses in practical managerial issues.

Methods Sampling design Two locations (approximately 70 m apart) were available in Wharf Marsh, but there was only one suitable location in Bennelong Pond because of the deep water throughout much of the pond. Therefore, this experiment had two locations in which the level of water was manipulated (Wharf Marsh East and Wharf Marsh West) and one control location in Bennelong Pond where the level of water varied naturally. This allows an asymmetrical comparison between one unaltered location (Bennelong Pond) and two locations subjected to change. Replicate samples from two sites (15 m2 , 1–2 m from shore) in each of the locations provided a representative measure of the assemblages of benthic animals at each location. Each site was sampled on two occasions, two weeks apart in each of four periods. Times 1 and 2 were before Wharf Marsh was fully drained, August and September, 1997. There was approximately 20–30 cm of water covering the sites. As predicted, drainage caused emersed areas of soft sediment around the edges of Wharf Marsh by October, 1997. These persisted for a few months and sites were resampled in January and February, 1998 (Times 3 and 4, respectively). Wharf Marsh was allowed to refill in April. It was then drained again in August, 1998 and resampled in early and late September, 1998 and in January and February, 1999 (Times 5 to 8, respectively). Therefore, Times 3 and 4 were after one period of drainage and Times 5–8 were after two periods of drainage. Sites in Wharf Marsh were sporadically covered by shallow water for short periods after heavy rain. The sites in Bennelong Pond were always covered by 20–30 cm water. Samples were collected in each site using five randomly placed 10-cm diameter and 5-cm deep cores. Samples were fixed in 7% formalin before being sieved. All animals retained on a 1-mm mesh were identified to broad taxonomic groups and counted.

Changes in assemblages of invertebrates: multivariate comparisons This study specifically predicted interactions between time (before and after drainage) and space (the drained and undrained locations). Multivariate analyses which can deal with the complexities of ecological data (e.g. skewed distributions, many zeroes) cannot, however, deal with complex experimental designs with interactive terms (Clarke, 1993). Therefore, the predictions for multivariate changes to the assemblage were necessarily simpler than those for univariate measures of abundance/diversity. Multivariate comparisons of the assemblage were done using non-metric multidimensional scaling (nMDS) and analysis of similarities (ANOSIM) on Bray-Curtis measures of dissimilarity among samples using untransformed data (Bray and Curtis, 1957; Clarke and Warwick, 1994). The relative contributions of different taxa to dissimilarities were evaluated using the similarity percentage breakdown procedure (SIMPER; Clarke and Warwick, 1994). Four predictions were made for the multivariate comparisons. The first was that prior to drainage, assemblages would be similar between sites in each location and among the three locations. This prediction reflects the degree to which the assemblages in the control location resembled those of the locations which were going to be manipulated and the degree of spatial variability before the start of management. The second prediction was that after drainage (Times 3–8), assemblages would be similar between the two sites in each location and between the two locations at Wharf Marsh, but each of these would differ from Bennelong Pond, i.e. assemblages in Wharf Marsh responded to the drainage. Third, it was predicted that the magnitude of differences between Wharf Marsh and Bennelong Pond would increase with repeated emersion (Times 5 and 6 compared to 3 and 4) and with longer periods of emersion (Times 7 and 8 compared to Times 3 and 4), i.e. greater magnitude or frequency of disturbance would cause greater changes to assemblages. Finally, it was predicted that the temporal changes in the assemblages from before to after drainage in each site in Wharf Marsh would be greater than those in Bennelong Pond. This prediction was based

Managing habitat for wading birds

on previous observations and suggestions that human disturbances often cause changes in variability rather than changes in average conditions (Warwick and Clarke, 1993; Chapman et al., 1995). The first two predictions were tested using one-factor ANOSIM (Clarke, 1993) to test for significant differences between sites (using replicate cores; ND5) and among locations (using the data from each site; ND10) at each time. Nested ANOSIM could not be used to test for difference among locations because there were only three possible permutations in each pair-wise comparison. These spatial differences were then illustrated with twodimensional nMDS plots (Clarke, 1993). The third prediction was examined by comparing the size of the average Bray-Curtis measures of dissimilarity among locations between the different periods, using the average measure from each Time as replication. Five randomly selected cores were allocated to each comparison, thus providing independent data for each. The data were analysed using a 2-factor analysis of variance (Factor 1, pairwise comparisons, three levels, fixed; factor 2, periods, three levels (Periods 1, 2 and 3), fixed; N D 2 replicates (Times per Period). Finally, the fourth prediction was tested by examining the temporal trends for each location on the same nMDS plot.

Changes in assemblages of invertebrates: univariate comparisons Beyond-BACI analyses of variance (Underwood, 1991b, 1992, 1993, 1994) were used to test the hypothesis that the abundance of oligochaetes and chironomid larvae and numbers of taxa would increase in the drained locations relative to the control locations. Because data were from two times in each of four periods, but one period was prior to drainage, one was after one period of drainage and two were after two periods of drainage, there was no balanced beforeafter comparison. Nevertheless, it was predicted that there would be either an interaction between Times (Periods)ðLocations or Periods ð Locations depending on the time course of the change.

More specifically, it was predicted that the drained locations would change differently from the control location among Times or Periods. Procedures to calculate appropriate sums of squares, mean square estimates and F-ratios for such asymmetrical (i.e. two drained and one control location) are as described by Underwood (1993). The sequence of preliminary tests of spatial and temporal variability and tests of hypotheses about effects of draining is described in Figure 1. Hypotheses about short-term effects of draining are tested by the interaction T(P)ðDs vs. C (Test C). If components of variability can be eliminated following preliminary tests (Tests A and B), this can be done using a pooled-mean-square (PMS) as error term (Winer et al., 1991; Underwood, 1997). Such procedures will ensure that statistical power is maximised while the rate of Type-1 errors is kept constant. Similar procedures are used for testing the hypothesis that drained and control locations differ consistently among Periods (PðDs vs. C; Test F). Significant effects of draining were further investigated using paired contrasts (Underwood, 1997), with the relevant comparisons decided a priori to test hypotheses about differences in means between the average of the two drained locations and the control location at different times. All data were transformed to natural logarithms prior to analyses because the hypotheses were about relative changes in abundances/diversity (Underwood, 1997). Cochran’s test was used to test for heterogeneity of variances after transformation. Effects of drainage on the small-scale spatial variability in number of taxa and abundances of animals was tested by calculating the variance among replicate cores in each site at each time and comparing these variances (one per site at each time) between the three locations using similar asymmetrical analyses of variance.

Results A total of 37 taxa were identified. There were two species of molluscs identified during sampling, the gastropod, Physa acuta Drapanaud and the bivalve, Lasaea australis Lamarck. Two families of polychaetes,

379

380

M. Lindegarth and M. G. Chapman Times (Periods) × Locations

Periods × Locations

A. Do sites within locations change differently among times?

D. Do sites within locations change differently among periods?

Test A: T(P) × S(L)/Res

if P (Test A) > 0.25 then Test D: P × S(L)/PMS else Test D: P × S(L)/T(P) × S(L)

B. Do drained locations change differently among times? if P (Test A) > 0.25 then Test B: T(P) × Ds/PMS else Test B: T(P) × Ds/T(P) × S(L)

E. Do drained locations change differently among periods? if P (Test D) > 0.25 then Test E: P × Ds/PMS else Test E: P × Ds/P × S(L)

C. Do drained locations change differently among times than do the control location? if P (Test B) > 0.25 then Test C: T(P) × Ds vs. C/PMS else Test C: T(P) × Ds vs. C/T(P) × Ds

Inference: If P (Test C) < 0.05 Draining affects short-term dynamics of benthos significantly. Proceed with comparisons between Ds and C to evaluate the nature of effects If P (Test C) > 0.05 Draining does not affect short-term dynamics of benthos

F. Do drained locations change differently among periods than do the control location? if P (Test C, D and E) > 0.25 then Test F: P × Ds vs. C/PMS if P (Test C) > 0.25 then Test F: P × Ds vs. C/P × Ds if P (Test D and E) > 0.25 then Test F: P × Ds vs. C/Ds vs. C else Test F: not possible

Inference: If P (Test F) < 0.05 Draining has significant effects on benthos. These persist for periods for longer periods of time. Proceed with comparisons between Ds and C to evaluate the nature of effects If P (Test F) > 0.05 Draining does not affect benthos significantly

Figure 1. Sequence of tests necessary for evaluating effects of draining (Tests C and F). Tests A, B, D and E are preliminary tests done to estimate spatial and temporal variability. These are done to evaluate possibilities to eliminate nonsignificant components of variance (at P>0Ð25; Winer et al., 1991; Underwood, 1997). PMS, pooled mean square consisting of eliminated components of variance. See Table 2 for full experimental design.

the Capitellidae and the Sabellidae were identified and oligochaetes, nematodes and nemerteans were identified as single separate taxa. The platyhelminths were represented by a single unidentified species and crustaceans were identified as ostracods or cladocerans. Adult insects consisted of seven species of Hemiptera, one specie of Coleoptera, two species of Hymenoptera and one specie of Diptera. Insect larvae included Chironomids (not distinguished further) and 12 other unidentified morphospecies (Oliver and Beattie, 1996). The oligochaetes and chironomid-larvae made up between 85 and 98% of the total abundances at each time of

sampling. No other taxon made up more than 5% of any sample at any time, except for one morphospecies of insect larvae at one time.

Multivariate changes As predicted, the assemblages were generally similar between replicate sites in each location (Table 1). Taxa that most contributed to differences between sites at Wharf Marsh East were: Time 4—chironomid larvae (48% of dissimilarity), a mix of other insect larvae (52%); Time 5—oligochaetes (85%) and chironomid larvae (10%). Oligochaetes and

Managing habitat for wading birds Table 1. Significance of differences in assemblages between replicate sites in each location and among locations at each time of sampling (from ANOSIM analyses) Between sites

Time Time Time Time Time Time Time Time

1 2 3 4 5 6 7 8

Between locations

WME

WMW

BP

WME vs. WMW

WME vs. BP

ns ns ns

ns ns ns ns ns ns n/a

ns n/a ns ns ns ns n/a

ns

ŁŁŁ

ŁŁ

ŁŁ

ŁŁŁ

ns

ŁŁ

ŁŁŁ

Ł

ns

ŁŁ

ns

ŁŁ

ŁŁ

Ł

ns ns

ŁŁ

ŁŁŁ

Ł

Ł

Ł

Ł

Ł

ŁŁ

ŁŁ

Ł Ł

ns n/a ns

WMW vs. BP

n/a-comparison not available because of excessive zeroes in the data; WME, Wharf Marsh East; WMW, Wharf Marsh West; BP, Bennelong Pond. ns, P>0Ð05; Ł P<0Ð05; ŁŁ P<0Ð01; ŁŁŁ P<0Ð001.

chironomids similarly contributed most to between-site differences in Wharf Marsh West (11% and 89%, respectively) and Bennelong Pond (19% and 77%, respectively) at Time 8. ANOSIM also illustrated significant differences in the assemblages between some pairs of locations, but these varied between Times 1 and 2 (Table 1). Therefore, in contrast to what was predicted, the locations differed before drainage started. The control location and the two potentially impacted locations were showing short-term interactive changes in the assemblages before drainage started. Nevertheless, assemblages were very variable among cores within each site, between sites in each location and there was considerable overlap among the three locations (Figure 2(a) and (b)). The change in the assemblages between Times 1 and 2 involved a very subtle shift of the assemblage found at Wharf Marsh West toward that at Bennelong Pond and away from that at Wharf Marsh East (Figure 2(a) and (b)). This spatial-temporal interaction prior to drainage makes it difficult to interpret significant differences among the locations after drainage (Table 1). No general patterns were evident. Wharf Marsh East remained significantly different from Bennelong Pond throughout the study. Wharf Marsh West was different from Bennelong Pond during five of the six sampling times (after drainage). In addition, the two drained locations in Wharf Marsh had significantly different assemblages during three of these six times, indicating that they did not show similar responses to drainage. At all times,

assemblages differed considerably among replicate cores and replicate sites (illustrated for Time 5 in Figure 2(c)). The third hypothesis predicted increasing dissimilarity between Bennelong Pond and each of the drained locations among the three periods after drainage. There was a significant interaction between Comparisons and Periods (F D 4Ð38, 4 and 9 df, P < 0Ð05). Periods were then compared for each pair-wise comparison using SNK tests. Average Bray-Curtis measures of dissimilarity did not change significantly for Wharf Marsh West vs. Bennelong Pond or Wharf Marsh East vs. Bennelong Pond. This was not as predicted by the hypothesis (there should be increasing dissimilarity between the three periods). In addition, there was greater dissimilarity between Wharf Marsh West and Wharf Marsh East during the second period (Times 3 and 4) than either the first (Times 1 and 2) and third (Times 5 and 6). This was also not as predicted because it was expected that the two drained locations would show similar differences at all times. The temporal trends in the assemblages are illustrated in Figure 3, using the centroids from each location at each time. Despite small-scale variability when each time is examined separately, the three locations had relatively similar assemblages prior to drainage (shown by the open symbols in Figure 3). After drainage, however, the assemblage in Bennelong Pond showed relatively small changes among times. The two drained locations, in contrast, showed large fluctuations, even between times only a few

381

382

M. Lindegarth and M. G. Chapman

weeks apart (shown by the closed symbols in Figure 3). These were not in the same direction, also illustrating that assemblages in the two drained locations responded differently to drainage.

(a)

Univariate changes

Stress = 0.09 (b)

Stress = 0.15 (c)

Because of the dominance of the fauna by oligochaetes and chironomid larvae, changes in abundance of these two taxa, along with the number of taxa, were tested using analyses of variance. The mean number of taxa and the abundance of chironomids were strongly and significantly affected by draining (indicated by the significant interaction between Periods and Drained vs. Control locations in Table 2). Planned contrasts showed significantly more taxa in Wharf Marsh than at Bennelong Pond before draining started. At all subsequent sampling times, there were more taxa at Bennelong Pond (Figure 4(a)), significantly so at Times 3, 4, 7 and 8. Similar differences were shown for the mean abundances of chironomids (Figure 4(b)). The mean abundance of oligochaetes was not significantly affected by draining (Table 2), although there was strong, potentially seasonal variability in their abundances. Nevertheless, there were greater mean abundances of oligochaetes at Bennelong Pond compared to Wharf Marsh at all times after drainage. There were no significant differences in the spatial variance among cores for the number of taxa or abundances of oligochaetes (P > 0Ð05 for all levels in the analysis). Variability of chironomids was significantly affected by draining (Periods ð Drained vs. Control; F D 7Ð42, 3 and 24 df, P < 0Ð01). Abundances were more variable in Wharf Marsh before draining but decreased after drainage, co-incident with the decrease in mean abundances.

Stress = 0.10

Discussion Figure 2. nMDS plots illustrating the relative similarity of assemblages in each core in Site 1; Site 2 Site 1, Site 2 at Wharf at Wharf Marsh East; Site 1; Site 2 at Bennelong Marsh West and Pond, respectively. Times of sampling illustrated are: (a) Time 1, August, 1997; (b) Time 2, September, 1997; (c) Time 5, September, 1998; n, five plots per site each time (fewer plots are shown when cores contained no animals, or similar assemblages so points overlapped).

The ultimate aim of these managerial actions was to enhance numbers of feeding individuals of the Japanese Snipe in the freshwater wetland. This was to be achieved by decreasing the level of water in the pond and thereby increasing the size of emersed feeding habitat and/or by increasing the densities of prey

Managing habitat for wading birds

Stress = 0.04

Figure 3. Temporal trends in the assemblages at , Wharf Marsh East; , Wharf Marsh West; , Bennelong Pond between Times 1 and 2 prior to drainage (open symbols) and Times 3 to 8 after drainage (closed symbols). Table 2. Asymmetrical analyses of variance and relevant contrasts for testing hypotheses about effects of draining on mean number of taxa and abundance of chironomids and oligochaetes Source

Taxa df

PeriodsDP Times(Periods)DT(P) Locations Drained vs. ControlsDDs vs. C Between drained locationsDDs Sites(Locations)DS(L) PðLo PðDs vs. C PðDs PðS(L) T(P)ðL T(P)ðDs vs. C T(P)ðDs T(P)ðS(L) Residual Cochran’s test Contrasts Ds vs. C Period 1 Period 2 Period 3 Period 4

3 4 2 1 1 3 6 3 3 9 8 4 4 12 192

Chironomids

Oligochaetes

Test

F

P

F

P

F

P

8Ð9 0Ð4 1Ð6

ŁŁ ns ns

38Ð3 No test 3Ð1

ŁŁ ŁŁ

No test 0Ð6 4Ð0

ns Ł

F E D

1Ð3 1Ð4 2Ð2

ns ns Ł

0Ð7 2Ð2 1Ð0

ns ns ns

2Ð2 0Ð2 1Ð1

ns ns ns

C B A

0Ð08

ns

0Ð08

ns

0Ð06

ns

C
C
F and P-values shown only for tests relevant to hypotheses about effects of draining. Results of contrasts are shown with means in rank order starting with the smallest mean. The right column refers to the sequence described in Figure 1. nsDP>0Ð05; Ł P<0Ð05; ŁŁ P<0Ð01; ŁŁŁ P<0Ð001

383

M. Lindegarth and M. G. Chapman 6 (a)

Mean (SE)

5 4 3 2 1 0

12

34

56

78

56

78

56

78

Times

50

(b)

Mean (SE)

40 30 20 10 0

12

34 Times

80 (c) 60 Mean (SE)

384

40

20

0

12

34 Times

Figure 4. Changes in mean (šSE) (a) number of taxa, (b) abundances of chironomids and (c) abundances of oligochaetes per core at: Wharf Marsh East; Wharf Marsh West; and Bennelong Pond.

in the sediment at the edges of the pond. Observations during sampling indicated that the area of emersed sediment increased along the edges of Wharf Marsh. More importantly, however, abundances and diversity of benthic animals decreased dramatically in the sediment in the locations which were drained. This decrease was evident during the first summer (4 months after the draining) and persisted throughout the experiment. Thus,

at the end of the sampling, the net effect of drainage was less potential food for birds feeding in Wharf Marsh. The results of this study clearly demonstrate that the managerial action taken was not appropriate for managing the abundance of animals in the sediment and should not be further attempted unless substantially modified and tested. Apart from the specific conclusions about the ‘success’ of the managerial actions, these results illustrate several general issues about problems in practical environmental management and the design of sampling programmes that can be used to evaluate logically the success of managerial actions. First, it is clear that managerial actions should be thought as experiments (e.g. Walters and Holling, 1990; Carpenter, 1990; Underwood, 1996). This includes management to restore habitats (Pastorak et al., 1997; McMahon, 1998; Chapman, 1999; Cairns, 2000; Thom, 2000). Inabilities of ecologists to do whole-system manipulations in the field due to logistic constrains are common themes in critiques of contemporary applied ecology (e.g. Hilborn and Walters, 1981; Peters, 1991; Underwood, 1998). Managerial actions represent unique opportunities for ecologists to test quantitative hypotheses about the success of management by considering them as manipulative field experiments. From the managers’ perspective, testing explicit hypotheses about effects of management is the only reliable way to evaluate the success of management. Without unambiguous information on its effects, further decisions about management cannot be founded on a scientific basis and management cannot evolve in an ‘adaptive’ manner (Walters and Holling, 1990). Second, this study demonstrates that appropriate replication is necessary for evaluation of the success of managerial actions. This is evident because the interpretation of the effects of draining on overall structure and composition of assemblages differed among comparisons when only one of the two drained locations were included in the comparison, i.e. there was only one drained and control location. In addition, there were differences between the two drained locations at many times of sampling, i.e. assemblages in the drained sites did not respond in the same way to the same managerial actions (Table 1, Figure 4).

Managing habitat for wading birds

Unpredictable, natural variability in abundances of animals at multiple spatial and temporal scales, similar to what was found here, are typical of assemblages in benthic habitats, including marine and freshwater soft sediments (e.g. Morrisey et al., 1992a,b; ´ and B´ır´o, 1998). This has profound Specziar consequences for the types of replication necessary to logically detect environmental impacts (Underwood, 1998; Lindegarth et al., 2000). This study illustrates how asymmetrical sampling designs (‘Beyond BACIdesigns’) may be used to successfully evaluate effects of managerial actions at multiple spatial scales, even under severe practical constraints which are typical of many situations where effects of environmental management need to be evaluated (Underwood, 1991b, 1992, 1993, 1994). For example, replication of ‘managed’ sites is often not possible to achieve because it is too expensive or unethical, while replicate control sites can generally be found (however, see Glasby and Underwood, 1998). As long as replication can be achieved in one of the experimental treatments (either control or managed sites), asymmetrical designs and analyses of variance can be used to contrast the magnitude of changes in managed areas compared with natural variability observed in unmanaged areas. Third, large short-term variability in abundance of benthos shown here illustrates the importance of replication at hierarchical temporal scales. For example, the number of taxa varied significantly at individual sites and differences among locations in abundance and composition differed among sampling times two weeks apart. Large changes were also observed among times in abundances of chironomids and oligochaetes at individual locations (Figure 4). Therefore, the accuracy of estimates of means at different periods will be substantially improved if samples are taken at replicate times within periods. Furthermore, one striking effect of draining was the increase in temporal variability in structure and composition of benthic assemblages (Figure 3). This is an effect that has been suggested as a sign of stress (Warwick and Clarke, 1994), but its generality has yet to be widely demonstrated Chapman et al., 1995). Sampling designs to detect such effects must, however, necessarily include sampling at multiple, hierarchical temporal scales.

Although recent developments of analytical techniques for testing hypotheses about changes in whole assemblages of animals have had a large influence on methods for environmental impact assessment, few studies have used multivariate techniques to evaluate the efficiency of restoration and creation of habitats (Chapman, 1999). Such techniques offer possibilities to make a sensitive, coherent assessment of effects on the level of whole assemblages. On the other hand, they are not yet developed to the degree that they can fully deal with the complex sampling designs which are needed to evaluate effects of managerial actions against a background of natural spatial and temporal variability at multiple scales. Nevertheless, the multivariate analyses revealed major changes in temporal variability as a consequence of draining. Therefore, the multivariate analyses provide an important complement to tests of hypotheses about individual taxa. Actions such as those described here are done because managers (or influential groups of the community) believe that they will achieve certain desirable endpoints. In all cases of environmental management (including attempts to create and restore habitats), there are no guarantees that managerial actions will have the desired effects. To ensure that valuable resources for management of natural resources are wisely used, it is essential that the managerial actions are evaluated properly (Hobbs and Norton, 1996; Chapman and Underwood, 1997). Like in any ecological experiment, this can only be achieved if the desired (and predicted) outcome is clearly defined beforehand (McDonald and Ericson, 1994; Grayson et al., 1999; Chapman, 1999). It is unfortunate that this attempt by managers to enhance the number of birds by increasing the available amount of food was initially guided by very vague aims and predictions. Procedures to evaluate whether some of the aims were met were not planned until after the management started, i.e. draining was done. For other aims, most notably how the actions worked to enhance the welfare of the bird populations, explicit plans of evaluation were, to our knowledge, never presented. Therefore, there are no credible evidence of how the bird populations changed, let alone what the draining did to the birds.

385

386

M. Lindegarth and M. G. Chapman

Despite problems of finding appropriate reference locations and some confusion about predicted endpoints, this study nevertheless emphatically demonstrated the abundances of benthic invertebrates were not enhanced by creating emersed patches of sediment along the edges of Wharf Marsh. Therefore, although the proposed model for management of the abundance of animals in the sediment was unsuccessful, the evaluation of its effects was successful and unambiguous. These results should be very useful if any further attempts are made to manage assemblages of benthic animals and ultimately populations of birds in this or other areas.

Acknowledgements This project was funded by the Olympic Co-ordination Authority and funds from the Centre for Research on Ecological Impacts of Coastal Cities, University of Sydney. We thank A. J. Underwood for discussions on the role of ecological science in environmental management. Thanks also to numerous members of the research staff for assistance with fieldwork, sorting data entry and checking and production of graphics and to two anonymous referees for helpful comments on an earlier draft of this paper.

References Adam, P. (1995). Reversing the trend. Wetlands 14, 1–5. Anderson, P. (1995). Ecological restoration and creation: a review. Biological Journal of the Linnean Society 56, 187–211. Bradshaw, A. D. (1993). Restoration ecology as a science. Restoration Ecology 1, 71–73. Bradshaw, A. D. (1996). Underlying principles of restoration. Canadian Journal of Fisheries Aquatic Sciences 53, 3–9. Bray, J. R. and Curtis, J. T. (1957). An ordination of the upland forest communities of Southern Wisconsin. Ecological Monographs 27, 325–349. Burchett, M. D., Pulkownik, A., Grant, C. and Macfarlane, G. (1999). Rehabilitation of saline wetlands, Olympics 2000 site, Sydney (Australia)I: management strategies based on ecological needs assessment. Marine Pollution Bulletin 37, 515–525. Cairns, J. Jr. (2000). Setting ecological restoration goals for technical feasibility and scientific validity. Ecological Engineering 15, 171–180. Carpenter, S. R. (1990). Large-scale perturbations: opportunities for innovation. Ecology 71, 2038–2043.

Chapman, M. G. (1999). Improving sampling designs for measuring restoration in aquatic habitats. Journal of Aquatic Ecosystem Stress and Recovery 6, 235–251. Chapman, M. G. and Underwood, A. J. (1997). Concepts and issues in restoration of mangrove forests in urban environments. In Frontiers in Ecology: Building the Links (N. Klomp and I. Lunt, eds), pp. 103–114. Oxford: Elsevier Science. Chapman, M. G. and Underwood, A. J. (2001). The need for a practical scientific protocol to measure successful restoration. Wetlands Australia 19: 28–49. Chapman, M. G., Underwood, A. J. and Skilleter, G. A. (1995). Variability at different spatial scales between a subtidal assemblage exposed to the discharge of sewage and two control assemblages. Journal of Experimental Marine Biology and Ecology 189, 1–2. Clarke, K. R. (1993). Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18, 117–143. Clarke, K. R. and Warwick, R. M. (1994). Change in Marine Communities: An Approach to Statistical Analysis and Interpretation. Plymouth, UK: Natural Environmental Research Council, Plymouth Marine Laboratory, UK. Downes, B. J., Lake, P. S. and Schreiber., E. S. (1993). Spatial variation in the distribution of stream invertebrates: implications of patchiness for models of community organization. Freshwater Biology 30, 119–132. Falk, K., Nøhr, H. and Rasmussen, L. M. (1994). Margrethe-Kog and the artificial saltwater lagoon: evaluation of a habitat restoration project in the Danish Wadden Sea. Environmental Conservation 21, 133–144. Frith, H. J., Crome, F. H. and Brown, B. K. (1977). Aspects of the biology of the Japanese snipe Gallinago hardwickii. Australian Journal of Ecology 2, 341–368. Gilbert, O. L. and Anderson, P. (1998). Habitat Creation and Repair. Oxford: Oxford University Press. Glasby, T. M. (1997). Analysing data from postimpact studies using asymmetrical analyses of variance: a case study of epibiota on marinas. Australian Journal of Ecology 22, 448–459. Glasby, T. M. and Underwood, A. J. (1996). Sampling to differentiate between pulse and press perturbations. Environmental Monitoring and Assessment 42, 241–252. Glasby, T. M. and Underwood, A. J. (1998). Determining positions for control locations in environmental studies of estuarine marinas. Marine Ecology Progress Series 171, 1–14. Grayson, J. E., Chapman, M. G. and Underwood, A. J. (1999). The assessment of restoration of habitat in urban wetlands. Landscape and Urban Planning 43, 227–236. Green, R. H. (1979). Sampling design and statistical methods for environmental biologists. New York: John Wiley & Sons.

Managing habitat for wading birds Green, A. J. and Hilton, G. M. (1998). Management procedures required to increase chironomid availability to waders feeding on artificial lagoons remain unclear. Journal of Applied Ecology 35, 9–12. Havens, K. E. and Aumen, N. G. (2000). Hypothesis-driven experimental research is necessary for natural resource management. Environmental Management 25, 1–7. Higgins, P. J. and Davies, S. J. (1996). Handbook of Australian, New Zealand and Antarctic Birds. Melbourne: Oxford University Press. Hilborn, R. and Walters, C. J. (1981). Pitfalls of environmental baseline and process studies. EIA Review 2, 265–278. Hill, D. A. (1989). Manipulating water habitats to optimise wader and wildfowl populations. In Biological Habitat Reconstruction (G. P. Buckley, ed.), pp. 328–343. London: Belhaven Press. Hobbs, R. J. and Norton, D. A. (1996). Towards a conceptual framework for restoration ecology. Restoration Ecology 4, 93–110. H¨otker, H. (1994). Wadden Sea birds and embankments - can artificial wetlands compensate for losses due to land claims. Ophelia 6, 279–295. Hurlbert, S. H. (1984). Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54, 187–211. Kentula, M. E. (2000). Perspectives on setting success criteria for wetland restoration. Ecological Engineering 15, 199–209. Kusler, J. A. and Kentula, M. E. (eds.) (1990). Wetland Creation and Restoration: The Status of the Science. Washington DC: Island Press. Lindegarth, M., Valentinsson, D., Hansson, M. and Ulmestrand, M. (2000). Interpreting large-scale experiments on effects of trawling on benthic fauna: an empirical test of the potential effects of spatial confounding in experiments without replicated control and trawled areas. Journal of Experimental Marine Biology and Ecology 245, 155–169. Lubchenco, J., Olson, A. M., Brubaker, L. B., Carpenter, S. R., Holland, M. M., Hubbell, S. P., et al. (1991). The sustainable biosphere initiative: and ecological research agenda. Ecology 72, 371–412. McDonald, L. L. and Erickson, W. P. (1994). Testing for bioequivalence in field studies: has a disturbed site been adequately reclaimed? In Statistics and Environmental Monitoring (D. J. Fletcher and B. F. Manly, eds), pp 183–187. Dunedin: University of Otago Press. McMahon, J. A. (1998). Empirical and theoretical ecology as a basis for restoration: an ecological success story. In Successes, Limitations and Frontiers in Ecosystem Science (M. L. Pace and P. M. Groffman, eds), pp. 220–246. New York: Springer-Verlag. Morrisey, D. J., Howitt, L., Underwood, A. J. and Stark, J. S. (1992a). Spatial variation in soft-sediment benthos. Marine Ecology Progress Series 81, 197–204. Morrisey, D. J., Underwood, A. J., Howitt, L. and Stark, J. S. (1992b). Temporal variation in

soft-sediment benthos. Journal of Experimental Marine Biology and Ecology 164, 233–245. Moy, L. D. and Levin, L. A. (1991). Are Spartina marshes a replaceable resource? A functional approach to evaluation of marsh creation efforts. Estuaries 14, 1–16. Murkin, H. R. and Kadlec, J. A. (1986a). Relationships between waterfowl and macroinvertebrate densities in a northern prairie marsh. Journal of Wildlife Management 50, 212–217. Murkin, H. R. and Kadlec, J. A. (1986b). Responses by benthic macroinvertebrates to prolonged flooding of marsh habitat. Canadian Journal of Zoology 64, 65–72. Murkin, H. R., Kaminski, R. M. and Titman, R. D. (1981). Responses by dabbling ducks and aquatic invertebrates to an experimentally manipulated cattail marsh. Canadian Journal of Zoology 60, 2324–2332. Murkin, H. R., Murkin, E. J. and Ball, J. P. (1997). Avian habitat selection and prairie wetland dynamics—a 10-year experiment. Ecological Applications 7, 1144–1159. Naarding, J. A. (1985). Latham’s snipe Gallinago hardwickii in Australia and Japan. Tasmania, Australia: National Parks and Wildlife Service. Oliver, I. and Beattie, A. J. (1996). Designing a cost-effective invertebrate survey: a test of methods for rapid assessment of biodiversity. Ecological Applications 6, 594–607. Pastorok, R. A., MacDonald, A., Sampson, J. R., Wilber, P., Yozzo, D. J. and Titre, J. P. (1997). An ecological decision framework for environmental restoration projects. Ecological Engineering 9, 89–107. Peters, R. H. (1991). A Critique for Ecology. Cambridge: Cambridge University Press. Rehfisch, M. M. (1994). Man-made lagoons and how their attractiveness to waders might be increased by manipulating the biomass of an insect benthos. Journal of Applied Ecology 31, 383–401. Reitan, O. and Sandvik, J. (1996). An assessment of retaining dams in hydropower reservoir for enhancing bird habitat. Regulated Rivers: Research and Management 12, 523–534. Short, F. T., Burdick, D. M., Short, C. A., Davis, R. C. and Morgan, P. A. (2000). Developing success criteria for restored eelgrass, salt marsh and mud flat habitats. Ecological Engineering 15, 239–252. ´ A. and B´ır´o, P. (1998). Spatial disSpecziar, tribution and short-term changes of benthic macrofauna in Lake Balaton (Hungary). Hydrobiologia 389, 203–216. Thom, R. M. (2000). Adaptive management of coastal ecosystem restoration projects. Ecological Engineering 15, 365–372. Thompson, S. P., Paerl, H. W. and Malia, C. G. (1995). Seasonal patterns of nitrification and denitrification in a natural and a restored salt marsh. Estuaries 18, 399–408. Thrush, S. F., Pridmore, R. D. and Hewitt, J. E. (1994). Impacts on soft-sediment macrofauna: the effects of spatial variation on temporal trends. Ecological Applications 4, 31–41.

387

388

M. Lindegarth and M. G. Chapman Todd, M. (1998). Feeding Ecology of the Latham’s Snipe Gallinago hardwickii in the Lower Hunter Valley, N.S.W. Sydney, Australia: Australasian Wader Studies Group, Birds Australia. Underwood, A. J. (1990). Experiments in ecology and management: their logics, functions and interpretations. Australian Journal of Ecology 15, 365–389. Underwood, A. J. (1991a). Biological monitoring for human impact: how little it can achieve. In Proceedings of the 29th Congress of the Australian Society of Limnology. (R. V. Hyne, ed.), pp. 105–123. Jabiru, Northern Territory, Australia. Underwood, A. J. (1991b). Beyond BACI: experimental designs for detecting human environmental impacts on temporal variations in natural populations. Australian Journal of Freshwater Research 42, 569–587. Underwood, A. J. (1992). Beyond BACI: the detection of environmental impacts on populations in the real, but variable, world. Journal of Experimental Marine Biology and Ecology 161, 145–178. Underwood, A. J. (1993). The mechanics of spatially replicated sampling programmes to detect environmental impacts in a variable world. Australian Journal of Ecology 18, 99–116. Underwood, A. J. (1994). On beyond BACI: sampling designs that might reliably detect environmental disturbances. Ecological Applications 4, 3–15. Underwood, A. J. (1996). Detection, interpretation, prediction and management of environmental disturbances: some roles for experimental marine ecology. Journal of Experimental Marine Biology and Ecology 200, 1–27.

Underwood, A. J. (1997). Experiments in Ecology: Their Logical Design and Interpretation using Analysis of Variance. Cambridge: Cambridge University Press. Underwood, A. J. (1998). Design, implementation, and analysis of ecological and environmental experiments: Pitfalls in the maintenance of logical structures. In Experimental Ecology: Issues and Perspectives (W. J. Resetarits and J. Bernardo, eds) pp. 325–249. Oxford: Oxford University Press. Walters, C. J. and Holling, C. S. (1990). Largescale management experiments and learning by doing. Ecology 71, 2060–2068. Warwick, R. M. (1986). A new method for detecting pollution effects on marine macrobenthic communities. Marine Biology 92, 557–562. Warwick, R. M. and Clarke, K. R. (1993). Increased variability as a symptom of stress in marine communities. Journal of Experimental Marine Biology and Ecology 172, 215–226. Warwick, R. M. and Clarke, K. R. (1995). New ‘biodiversity’ measures reveal a decrease in taxonomic distinctness with increasing stress. Marine Ecology Progress Series 129, 301–305. Winer, B. J., Brown, D. R. and Michels, K. M. (1991). Statistical Principles in Experimental Design. New York: McGraw-Hill. Inc. Zedler, J. B. (1996). Ecological issues in wetland mitigation: an introduction to the forum. Ecological Applications 6, 33–37. Zedler, J. B., Fellows, M. Q. and Trnka, S. (1998). Wastelands to wetlands: links between habitat protection and ecosystem science. In Successes, Limitations and Frontiers in Ecosystem Science (M. L. Pace and P. M. Groffman, eds), pp. 69–112. New York: Springer-Verlag.