Global Patterns of Biodiversity
Diversity is a function of two factors: number of species (Species Richness) and number of specimens belonging to these species (Evenness). Several indices measuring diversity have been proposed, giving more or less weight to either of these two factors. To illustrate global patterns of biodiversity, Hurlbert’s index, for a sample size of 50 specimens, is calculated on the OBIS data holdings.
Download this reportbiodiversity illustration final.pdf — PDF document, 863Kb
One of the more intuitive criteria on which conservation efforts are based is ‘Diversity’ – but measuring diversity is not straightforward. What is usually perceived as diversity consists of two independent components(1)species richness, and (2) the relative abundance of specimens belonging to these species, often calculated as ‘evenness’. The more even the number of individuals from each species in a sample, the higher the diversity. The higher the number of species, the higher the diversity as well. Since species richness and evenness are independent factors, there is no unambiguous way to formulate the ultimate ‘diversity index’. Many indices have been proposed, often differing in the weight that is attached to either of the two components. Hill (1973) was able to show that many of the more commonly used indices are mathematically related. Several publications give a good overview of diversity indices, and their respective advantages and drawbacks (e.g. Magurran 1988; Grassle et al. 1979; Heip et al. 2001). ). Additionally, some measures also consider how different the species are from each other, such as how far they are separated on the “tree of life;” e.g. taxonomic distinctness (Warwick & Clarke, 1995; Clarke & Warwick, 1998).
Using the Ocean Biogeographic Information System (OBIS) data holdings, it is possible to investigate global patterns of biodiversity. Most OBIS records contain information on the presence/absence of a species, not its abundance. Unfortunately, abundance data is an essential component of most currently used diversity indices. One way around this problem is to use the number of records for a particular species as a proxy for its abundance.
A very simple approach to represent the global patterns in biodiversity is to count the number of species in a given area, and repeat this for a set of non-overlapping squares, or other polygons. This is illustrated in fig. 2b. The drawback of this simple method is that species richness is very much influenced by the sampling bias: the more observations are made, the more species will have been discovered. This is illustrated in the graphs in fig. 3. Sampling bias is a significant problem for open oceans and deep seas, where sampling effort has concentrated on discrete areas, while other areas or entire regions remain largely unexplored.
One of the most popular diversity indices is the Shannon index. Unfortunately, this index is also very sensitive to sampling effort. One assumption of the index is that all species of the community are present in the sample; obviously this will only be true if the number of sampled individuals is very large. Again, this is illustrated in the scatterplots in fig. 3, and can be seen from comparing the maps in figs 2a and 2c.
A measure of biodiversity that is both intuitive and relatively insensitive to observation bias is Hurlbert’s Index, which is calculated as the number of distinct species expected to be present in a random sample of, for example, 50 individuals from an area. es(50), the expected number of species in 50 individuals, is calculated for 1 degree cells in the main illustration (fig. 1), and repeated for comparison purposes using 5 degree squares, as fig. 2d.
Figure 2:(a) total records in OBIS, corrected for differences in surface area between squares on different latitude; (b) the total number of species, corrected for differences in surface area between squares on different latitude; (c) Shannon Index; (d) Hurlbert’s Index, es(50)
Figure 3: matrix of scatterplots of the different quantities discussed, showing interrelationship between them. log_n: log-transformed number of records in OBIS, per 5 degree square; log_s: log-transformed number of species per 5 degree square; shannon: Shannon diversity index; es: hurlbert’s index, in this case calculated for a sample of 50 observations from a 5 degree square. The value on the x-axis is read from the column (for example, the x axis for all scatterplots in the first column is log_n); the y-axis from the row (for example, all y axes on the third row are Shannon). The main point of this figure is to illustrate the strong correlation between number of observations, number of species and the Shannon Diversity Index, and the relatively low correlation of these factors with Hurlbert’s index.
Sources of data
Raw biogeographic data are available from many sources, such as large museums, national monitoring programmes, fisheries data, data from individual datasets, and so forth. The challenge is that these data are not always easily accessible, and that individual datasets are usually collected on a limited scope – geographic, taxonomic and temporal. The Ocean Biogeographic Information System (OBIS) was initiated to create a data warehouse to integrate this multitude of data in one comprehensive, quality-controlled system. OBIS is a work in progress, and increases steadily the quantity and quality of the data available through its portal. Its main contributors are
- the field projects of the international Census of Marine Life
- Regional OBIS Nodes, a network of often national organizations coordinating OBIS input for a region
- thematic contributors such as OBIS SEAMAP (mammals, birds, turtles), MicrOBIS (microbes), FishBase or Hexacorallia
- individual data holders such as the Smithsonian Museum, or the Continuous Plankton Recorder of the Sir Alister Hardy Foundation for Ocean Sciences
An assumption inherent in the approach of many diversity indices is that species are interchangeable – which obviously is not true. Diversity mainly consisting of invasive species is not a desirable situation; endangered and/or endemic species clearly should rank higher when making conservation decisions. So any final decision should take into consideration the actual species composition. Phylogenetic issues (i.e., issues pertaining to the evolutionary relatedness of species) have also been raised suggesting that it is better to have an area with species that are more distantly related than an area with only closely related species (e.g. Humphries et al. 1995). Indices that include relatedness exist (e.g. Warwick and Clarke, 1995, Clarke & Warwick, 1998), but require additional work before they can be calculated on large global data sets like OBIS.
There are significant differences in the intensity with which the oceans are studied: many more data are available for coastal areas than for the open ocean. In the open ocean, surface waters are more intensively sampled than the bottom, and even fewer data are available for the the mid-waters. Some large datasets (such as a long-line fisheries dataset from South Africa), with a massive number of records (more than 3 million) for very limited number of target species, results in low estimate of es(50) because longline fisheries data swamp very good biodiversity surveys of the same region. One example of such an approach is the Aquamaps model, which is discussed elsewhere in this document. Other models are, e.g., GARP and WhyWhere. Many of these tools are available through the OpenModeller web site.
The content of OBIS, as it is now, is suitable for the study of broad patterns of the distribution of biodiversity. The content is not sufficient to allow a detailed analysis on a regional scale, or to study the distribution patterns of individual taxa. But OBIS does provide a framework for capture and re-use of existing data, and will grow in time. It is a mechanism for data sharing, including data repatriation from the West to developing nations and small island states. The latter often do not have sufficient capacity to set up their own biodiversity data capture and management systems, and therefore could greatly benefit from global initiatives.