Preprint volume - SIBM

Pre-print Volume –Lecture
Topic 1: BIODIVERSITY AND CONSERVATION SCIENCE: CONTRIBUTING TO MANAGEMENT
Large scale patterns in biodiversity have been quantified using different multispecies
surrogates including: 1) priority species (on the assumption that protecting
rare/declining/threatened species provides an effective umbrella for overall species
richness in an area, but this does not always hold (Bonn et al., 2002)); 2) structural or
‘ecosystem engineer’ species (Jones et al., 1997); 3) specific groups e.g. molluscs,
polychaetes, mammals and sharks that are taxonomically stable and evenly recorded
(but are not necessarily indicators of total biodiversity (Smith, 2008)); 4) death
assemblages – remains of shell bearing molluscs (Warwick and Light, 2002). All these
surrogates have limitations, few have been correlated to overall biodiversity and some
approaches are not appropriate for marine assessments such as endemic species (not
applicable due to the higher connectivity and lower endemism in marine systems
compared with terrestrial).
There are an array of different measures to quantify species diversity (e.g. diversity
indices, number of species, number of priority species and taxonomic distinctness) and
many of these can also be applied to identify large-scale patterns at the scale of
habitats. Each yields a different representation of diversity, and there is often a lack of
congruence between measures (Orme et al., 2005). This has led to the combination of a
range of measures being used to capture patterns in biodiversity (e.g. Myers et al.,
2000, Hiscock & Breckels, 2007).
Each individual measure has its own limitations in terms of data requirements,
sensitivity to variability inherent in the data (such as uneven sampling effort), ways of
assigning confidence and the application of output in terms of the suitability to
represent overall biodiversity over large scales. Species and habitat richness are the
most commonly used measures of diversity and do not require abundance data, which
can add a significant bias when handling data from multiple sources. Similarly
taxonomic distinctness can be calculated without abundance data and provides
additional information on the phylogenetic diversity of a site which is arguably more
meaningful in terms of assessing ecosystem function.
An approach for large scale pattern detection - The greatest constraint on the
approach taken relates to the available data: large scale analyses are dependent on
existing data that were originally collected for a multitude of different purposes using
various sampling methods. Any approach that attempts to combine these data must take
into consideration the quality of the data by filtering out low quality data (that is either
old, inaccurate in terms of taxonomy, spatial reference etc.) but also standardize for
sampling method and coverage (some areas may be intensively sampled while others
rarely visited). We have used regression to standardise for sampling effort, carrying out
analyses on data collected by broadly similar collection methods and recombining
scores by grid cell (Langmead et al., 2008; Jackson et al., 2009). Other techniques
include resampling.
The data also determine the spatial scale of analysis, since the resolution of any grid
applied to the area of study needs to be of optimal size; small sized grids result in many
empty cells while large sized grids in the loss of fine scale resolution (Fig. 1). In our
work around the coast of Wales we used several grids matched to the data coverage: 1)
a small grid for the intertidal area (where sampling had better coverage); 2) a large grid
for the subtidal area (where data coverage was lower) and 3) an intermediate sized grid
to ensure comparison of the intertidal and subtidal areas (Jackson et al., 2009).
41 st S.I.B.M. CONGRESS Rapallo (GE), 7-11 June 2010
22