230 P
. Kraufvelin J. Exp. Mar. Biol. Ecol. 240 1999 229 –258
Cerastoderma glaucum realism. Finally the factors and processes restricting mesocosm per- formance are outlined and their consequences are briefly discussed. It is concluded that the degrees
of replicability, repeatability and ecological realism are too low for straightforward use of these and probably most other mesocosms in predictive risk assessment or in extrapolation of results to
natural ecosystems.
1999 Elsevier Science B.V. All rights reserved.
Keywords : Baltic Sea; Community structure; Ecological realism; Mesocosm; Multivariate
statistics; Repeatability; Replicability; Rocky shore macrofauna
1. Introduction
It is a disquieting fact that so much of the research in marine mesocosms just seems to have been driven by the need to investigate the fates and effects of pollutants with the
consequence that basic underlying information on the principles that govern the functioning of these experimental systems, and of the natural systems of which they are
living models, have been given inadequate attention Pilson, 1990. Replicability, repeatability and realism are all important aspects of experimental ecosystems, but
nevertheless almost unknown when it comes to larger, more complex systems. Replicability is defined as the degree of similarity of spatially replicated experimental
units that are meant to represent the same conditions by definition and design, i.e. in this paper mesocosm controls run during the same year. The term repeatability is used to
describe the similarity of responses in independent systems that are observed at different points of time within the same research facility and by use of the same scientific
methods, i.e. in this study mesocosm controls run at the same place but during different years. The ecological realism or accuracy of the mesocosm is the degree of similarity
between the artificial system and the natural ecosystem mimicked. All definitions originate from Giesy and Allred 1985.
The economically and logistically optimal implementation of replicability and repeatability is more or less diametrically opposed to the essential concept of test system
realism and a central problem of ecological experimentation Kuiper et al., 1983; Hurlbert, 1984; de Lafontaine and Leggett, 1987. This is mainly due to the fact that all
these aspects of mesocosm similarity are related to scale Gamble, 1990; Cairns and McCormick, 1991; SETAC-Europe, 1991; Landis et al., 1997. Ecological realism tends
to increase gradually with increasing spatial and temporal scale of a study, while our abilities to replicate and repeat an experiment logically decrease more rapidly, because
of the restrictions all artificial designs impose on the included ecosystem parts Carpenter, 1996. Therefore a high degree of replicability and repeatability is generally
sacrificed on behalf of ecological realism, which mostly is stated to be the major reason for carrying out a mesocosm study in the first place. Although it has been stressed that
the value of a model ecosystem resides in its ability to mimic some real system, not itself Perez, 1995, all these three aspects of internal and external mesocosm similarity
are closely interlinked when it comes to interpretation and validation of results. Since all three cannot be met with the same intensity simultaneously and by use of the same
P . Kraufvelin J. Exp. Mar. Biol. Ecol. 240 1999 229 –258
231
methods, the scientist is often forced to emphasise one or the other. The typical trade-off from increased ecological realism gained with a larger mesocosm size is less experimen-
tal control, often expressed as less information on individual processes, less isolation of cause and effect as well as less ability to define accurately the densities of contained
biota, which ultimately makes it harder to interpret results Steele, 1979; Stephenson et al., 1984; Crossland and La Point, 1992.
Only a few case studies are available, where the replicability of aquatic mesocosms
3
i.e. artificial ecosystems bigger than 1 m have been thoroughly described Takahashi et al., 1975; Pilson et al., 1980; Brazner et al. 1989; Heimbach et al. 1992; Rosenzweig
and Buikema, 1994; Jenkins and Buikema, 1998; Kraufvelin, 1998. The ecological realism of aquatic mesocosms has on the other hand been discussed more frequently
e.g. Gearing, 1989; Lalli, 1990; Adey and Loveland, 1991; Clark and Cripe, 1993; Kennedy et al., 1995, although most comparisons between mesocosms and the field
have not been very ambitious or innovative Pilson, 1990; Perez, 1995. Precise data on mesocosm repeatability, finally, have never been presented at all. This last finding is
least to say surprising, especially taking into account the large number of aquatic mesocosm test systems currently in operation world-wide and the fact that also some
insight in repeatability is needed if mesocosms are to be used for prediction of real effects in natural systems Crane, 1997. In the recent paper by Kraufvelin 1998 some
problems with the replicability of BHB-mesocosm were pointed out. Coefficients of variation CVs were presented for a large number of variables, many of which
previously have been used as test endpoints in the mesocosm in question e.g. Landner et al., 1989; Lehtinen and Tana, 1992; Lehtinen et al., 1993, 1994, 1995, 1996, 1998; Tana
et al. 1994. These CVs were generally so high that they would probably have prevented any detection of significant differences between controls and treatments, if only real
replicates not ‘simple pseudoreplicates’, as defined by Hurlbert 1984, had been used. Partly in order to accentuate the problem with poor replicability and partly since many
large-scale community studies, that more effectively could be analysed by other means, still are stuck with univariate methods, the paper by Kraufvelin 1998 was only
concerned with the univariate one-way ANOVA. A more proper ecological approach to analyse this data would have been to carry out some kind of multivariate statistical
analysis parallel to the univariate ones. This would increase the understanding of the behaviour of the bladder-wrack macrofauna communities in several replicated and
repeated control mesocosms. Comparisons with simultaneous measurements in the field would further strengthen the overall picture of the performance of BHB-mesocosms and
other large-scale experimental ecosystems.
In this paper I therefore first present the replicability, repeatability and ecological realism of BHB-mesocosms visually by ordination. Then I discriminate the parallel
mesocosms replicability, different years repeatability and the mesocosms and mother system ecological realism analytically by various ANOSIM tests. This is first done by
using subsampled bladder-wrack plants replicability and then the mesocosms them- selves as replicates repeatability and realism. Note that the former tests do not imply
pseudoreplication as long as they are just considered along the line ‘is mesocosm A different from mesocosm B’ and no issues of causality e.g. confounding with
treatments are incorporated Hurlbert, 1984. Further objectives include to search for
232 P
. Kraufvelin J. Exp. Mar. Biol. Ecol. 240 1999 229 –258
the causes behind the observed patterns in macrofauna community structure and to demonstrate some often overlooked problems when working with living communities.
This will be accomplished by examination of bladder-wrack macrofauna communities both in the mesocosms and in the field mother system and by pin-pointing the species
and processes basically responsible for observed differences between studied groups. This is possible thanks to the low number of species present in the mesocosms and in the
Baltic bladder-wrack zone in general, at least compared to fully marine environments Haage, 1975; Wallentinus, 1991; Kautsky et al., 1992. Finally I discuss the conse-
quences of these findings for an effective use of large-scale mesocosms in ecological and ecotoxicological research.
2. Materials and methods