THE DEVELOPMENT AND APPLICATION OF ANALYTICAL METHODS IN BENTHIC MARINE INFAUNAL STUDIES

The major problems addressed here have not changed since the early 1900s. The methodology has changed dramatically. In sampling the diffuse literature on this topic, one gets the impression that progress in generalizing, simplifying and explaining the processes involved in benthic community organization, is painfully slow in comparison with more clearly defined areas of scientific study. The key to further progress in this field is to develop sampling and analytical techniques in concert. This presupposes that research aims and study hypotheses can be more clearly stated in benthic studies than has often been the case in the past. We have tried to provide a perspective on the development of analytical and sampling methods in benthic survey studies. Methods have evolved from the original intuitive approach based on the indicator species (or group) concept, to objective univariate indices which provide a useful initial characterization of a community or spatial pattern. These univariate methods are being progressively replaced or enhanced by more rigorous descriptive and inferential multivariate methods. Researchers of benthic systems can be simply overwhelmed by the variety and complexity of analytical methods available. In this chapter, we have not attempted to provide handbook (or cookbook) procedures. In all honesty, we cannot define right or wrong methodology, though this has been attempted to some degree (cf. Green and Vascotto, 1978; Field et al., 1982; Smith et al., 1988; Gray et al., 1988). Instead, we have attempted to provide the prospective researcher with a feel for the assumptions involved with certain methods, or the particular prejudices that have led to the emergence of currently “popular” methods. If anything can be concluded from the often contradictory opinions of different ecological authorities, it is that the analytical approach should be straightforward, avoiding the common trap of using a whole suite of complex (often uninterpretable) methods when one or two would be sufficient. If the data set is “robust”, the results will probably not be seriously affected by the use of questionable statistical methods. Unfortunately, the validity of results cannot be judged by other readers without an adequate description of the sampling design and the analytical methods. This is probably the most common shortcoming in many applied papers in benthic ecology. In addition, data management is a pressing problem in many studies. Much more attention should be paid to the biases and consistency of sampling methods and how they relate to the assumptions inherent in the statistical tests, to the validity of data transformations or reductions, and to the taxonomic adequacy. In most cases, the researcher may have reasonable confidence in the power of statistical inferential tests, due to the large sample sizes characteristic of most benthic studies. On the other hand, large sample sizes can lead to the problem of over-interpretation of “trivial” but statistically significant results. Therefore, the effect size of interest should be carefully considered. Because of the limitations inherent in species identifications, the use of biomass/size spectra to analyse trophic relationships and population structure in benthos is an alternative approach which seems to be receiving some attention. The potential usefulness of combining taxonomically based and biomass-based studies has not yet been fully explored, partly because of the additional costs and labour required for accurate descriptions. The appeal of non-parametric simulation or randomization methods for hypothesis testing is expected to increase, since these methods eliminate many of the problems encountered when attempting to fit aggregated, multispecies data to parametric models. Most researchers would agree that the introduction of reliable and flexible methods to simplify the often confusing and frustrating process of data analysis would be welcome. To this end, more effort should be made to invent clear graphical methods for depicting complex statistical results, for the benefit of managers, public groups and political agencies with policy decisions to make (see Figs 10 and 11). As methods and understanding of basic mechanisms affecting benthic communities improve, researchers seem to be attempting more ambitious studies. Long-term data are now available in many areas, as well as the type of widespread sampling coverage and data-handling methods which may eventually encourage researchers to readdress broad community issues of the type raised originally by Petersen (1913-191 5) and later by Thorson (1957, 1966). © 1990 Academic Press Inc.