You are here: Home Our Work Overlap between hotspots of marine mammal biodiversity and global seamount distributions

Overlap between hotspots of marine mammal biodiversity and global seamount distributions

submitted by Kaschner, K., J. Ready, E. Agbayani, P. Eastwood, T. Rees, K. Reyes, J. Rius & R. Froese

Figure 1(a) Global map of predicted marine mammal species richness and seamount density and 1(b) the highly significant relationship between the two, spearman’s rho = 0.76, p < 0.0001) (both graphs modified from Kaschner, 2007)

AquaMaps is a species distribution model available as an online web service that generates standardized range maps and the relative probability of occurrence within that range for currently more than 9000 marine species from available point occurrences and other types of habitat usage information (Kaschner et al, 2006, Ready et al, accepted). By overlaying AquaMaps predictions for a subset of individual species (namely 115 marine mammals), we produced a global map of biodiversity patterns that shows the co-occurrence of predicted hotspots of marine mammal species richness and off-shore seamounts.

Download this report

marine mammal biodiversity and global seamounts.pdf — PDF document, 412Kb

Scientific Background

Species are not uniformly distributed on Earth.  Heterogeneous physical features and community evolution drive the mix of species found in a given location.  Places with high species diversity tend to contain greater genetic diversity and variety of physical habitats, as well as exhibiting greater resilience to environmental changes.  Biological diversity, most commonly measured as the number of species present or species richness (versus number of genes or ecological habitats), provides an important composite measure of biological and ecological importance.

Areas of high biological diversity, depending on their size and accessibility, may be identified either directly from intensive sampling or survey efforts. Alternatively, such hotspots of species richness may be inferred using habitat prediction models by overlaying a large number of predicted species occurrence data layers in a Geographic Information System (GIS) which will highlight areas of high co-occurrence of species.

The vastness of the offshore marine environment and the resulting paucity of data for the open ocean or deep sea will likely preclude the identification of most areas of high biological diversity in the high seas through direct survey efforts alone in the foreseeable future. Thus large-scale species distribution models represent a useful alternative to evaluate the biological or ecological significance of an area based on this criterion by making inferences from better studied areas to those comparatively less well known (Wood et al, in review).

In general, just as the physical environment often determines the limits of a single species’ distribution, high diversity in habitat can be a good predictor for species richness. In this example, high marine mammal species richness is predicted to correlate with greater seamount density.  Seamounts aggregate prey and provide a wide range of environmental gradients for multiple species to thrive, often with enough physical separation to support diverging evolutionary histories.  Here we see this diversity effect cascading all the way to top predators, i.e. marine mammals.

How the areas of high biological diversity were identified

We used AquaMaps, an environmental envelope model which is a modified version of the relative environmental suitability model (RES) developed by (Kaschner et al. 2006) to visualize and investigate patterns of species richness. AquaMaps is available online and has been used to produce standardized range maps of > 9000 marine species to date. Predictions are provided in the form of relative probabilities of a given species to occur in each grid cell of a global grid of 0.5 degree latitude by 0.5 degree longitude cell dimensions. Using the tools available on the project web site, we selected and superimposed predictions for a subset of species, namely 115 marine mammals to visualize global patterns of biological diversity for these taxa. For each species, we assumed a relative probability threshold of 0.4 to define species presence in a given area. We then compared predicted marine mammal species richness in relation to a global seamount data set (Kitchingman et al, 2007) and found a significant relationship between areas of high marine mammal diversity and areas of high seamount density (Kaschner 2007). This preliminary analysis provided support that offshore seamounts represent biologically significant areas even for highly mobile and transient species such as most marine mammals.

There are a wide range of available habitat prediction models requiring different types of input data and all coming with their own set of assumptions (see (Elith et al. 2006; Guisan et al. 2006; Guisan & Zimmermann 2000; Redfern et al. 2006). All of these models can, in theory, be used to generate predictions about species occurrence which can then be applied to make inferences about areas of high biological diversity. Most of these models tend to be more sophisticated and flexible in their assumptions than the AquaMaps approach. However, performance of all models is limited by the quality of their input data sets (Lozier et al, 2009) and currently available point occurrence data sets for large scale range predictions are affected by a large number of biases (see below). AquaMaps was specifically developed to deal with these biases to the extent possible and has been shown to perform as well as or better than other habitat prediction models when faced with currently available suboptimal, patchy, large-scale data sets (Ready et al., accepted).

Sources of Data

There are two types of input data used to generate AquaMaps species predictions. Firstly, available point occurrence records for the respective species, used to calculate all environmental envelopes (except for depth preferences) are harvested (and continually updated) from online data repositories such as the Ocean Biogeographic Information System (OBIS), OBIS-SeaMap (a sub-node of OBIS dealing specifically with marine mammals, seabirds and sea turtles), or the Global Biodiversity Information Facility (GBIF). Such point data sets are compiled from a variety of different sources and are generally affected by a number of sampling biases including, but not limited to, non-representative coverage of habitats and species misidentifications. To correct for misidentifications or geographic misallocations of occurrence records, information about the general occurrence of species in different ocean basins is used as a filter to select “good” points. This information is harvested from existing online species databases such as FishBase and SeaLifeBase where it is provided in the form of FAO statistical area checklists and/or bounding boxes delineating the known maximum range extent boundaries for species as described in the scientific literature. The concentration of sampling efforts in continental shelf areas often results in a mis-representation of the true depth usage of species. To counteract this bias, AquaMaps relies on depth usage information taken from the literature, as encoded in online species databases. To enable further correction of known biases or to predict occurrence for species for which point records are currently lacking, AquaMaps explicitly allows for the incorporation of further or alternative input data in the form of expert knowledge about habitat usage. This type of information can be used to adjust both envelope settings as well as bounding box/FAO area definitions during an expert review process. 

Important considerations

Large-scale species distribution models currently probably represent the best, if not only choice to identify potential areas of high biological diversity in most of the often data poor off-shore regions of the world’s oceans. However, the concentration of sampling effort in more accessible habitats, such as the continental shelf regions of the northern hemisphere also represents a great challenge for the application of any species distribution modeling technique and results of all models therefore need to viewed with some caution. Most commonly, species distribution models predict broad range extents that often do not consider seasonal movements of animals or subspecies level population structure, and may thus potentially overlook critical habitat needed during certain life stages or for maintaining subspecies level diversity. When simply adding up the number of species, other useful information can be missed.  For a given species, we are often interested in attributes such as:  abundance, genetic uniqueness, endemism, and endangered status.  However, most models and subsequently diversity indices derived from such predictions do not consider relative or absolute abundances of individual species and are indifferent to species substitutions. Hence, mapping of biodiversity hotspots may not reliably pick up on areas important to species of special concern, such as endangered and/or extremely rare species, although it is possible species weights can be assigned to add up a relative diversity measure inclusive of some of these aspects.

Despite these caveats, which affect most currently existing models, an exercise such as the one conducted here may provide a starting point for  evaluating the significance of an area. As indicated above, the tools and features available on the AquaMaps website allow for the selection of different subsets of species based on a range of different conservation and management criteria. Currently, taxa such as ray-finned fish and elasmobranchs as well as marine mammals are either complete or comprehensively covered by Aquamaps (see Figure 2), but coverage is currently being expanded to invertebrate, algae and hexacoral taxa. The incorporated expert review process represents a Wiki approach that can greatly facilitate the review of existing data and resulting predictions through expert panels such as IUCN species working groups.

However, to most reliably identify areas of high biological diversity, a range of different modeling techniques should ideally be applied to determine which regions are consistently – across all model outputs - predicted to represent hotspots. Model selection and spatial and temporal scales of the analysis should be based on data availability and the ecology and life history of the taxa in question and outputs should be validated with independent, effort-corrected survey data to the extent possible. Forward projections of changes in species distributions and related areas of high biodiversity under different climate change scenarios can help to identify those significant areas  most likely to ensure long-term protection of high biological diversity. 


Figure 2: Example of AquaMaps current coverage: Map of ray-finned fish species richness and hotspots of biodiversity for more than 6000 species

Document Actions