Research articlesCan the source–sink hypothesis explain macrofaunal abundance patterns in the abyss? A modelling test
Abstract
Low food availability is a major structuring force in deep-sea benthic communities, sustaining only very low densities of organisms in parts of the abyss. These low population densities may result in an Allee effect, whereby local reproductive success is inhibited, and populations are maintained by larval dispersal from bathyal slopes. This slope–abyss source–sink (SASS) hypothesis suggests that the abyssal seafloor constitutes a vast sink habitat with macrofaunal populations sustained only by an influx of larval ‘refugees' from source areas on continental slopes, where higher productivity sustains greater population densities. Abyssal macrofaunal population densities would thus be directly related to larval inputs from bathyal source populations. We evaluate three predictions derived from the SASS hypothesis: (i) slope-derived larvae can be passively transported to central abyssal regions within a single larval period, (ii) projected larval export from slopes to the abyss reproduces global patterns of macrofaunal abundance and (iii) macrofaunal abundance decreases with distance from the continental slope. We find that abyssal macrofaunal populations are unlikely to be sustained solely through influx of larvae from slope sources. Rather, local reproduction probably sustains macrofaunal populations in relatively high-productivity abyssal areas, which must also be considered as potential larval source areas for more food-poor abyssal regions.
1. Introduction
The abyssal seafloor, encompassing ocean depths between 3000 and 6000 m, is a vast habitat covering approximately 54% of the Earth's surface (reviewed by Thistle [1] and Smith et al. [2]). This habitat consists largely of plains and hills covered by fine sediments, punctuated by manganese nodules, seamounts and trenches. Sunlight is absent in abyssal ecosystems, which are overwhelmingly dependent on the sinking flux of detrital organic material from primary production in overlying surface waters. However, only a small fraction of net primary production reaches the abyssal seafloor (approx. 0.5–2%), with the amount varying geographically according to levels of primary production and export efficiency in overlying waters (e.g. [2]). Abyssal ecosystems are thus characterized in large part as extremely ‘food limited’.
Food limitation is widely considered to be a major structuring force in deep-sea benthic communities (reviewed by Smith et al. [2], Levin et al. [3] and Snelgrove & Smith [4]). Variation in carbon flux at both local and regional scales influences patterns of abundance, biomass, body size and levels of alpha-, beta- and gamma-diversity, as well as a broad range of ecosystem functions [2]. Food limitation in parts of the abyss can be so great that only extremely low macrofaunal population densities can be sustained [5]. Such circumstances could generate an Allee effect, whereby low densities severely hamper reproductive success and local population growth rates could reach zero, or even negative values [6]. These low-quality ‘sink’ habitats, in which reproduction is insufficient to balance mortality, occur in a variety of ecosystems [7]. ‘Source–sink theory’ describes the export of offspring from high-quality habitats (sources) to populations in sink habitats through dispersal and/or emigration [8,9]. Sink populations are unsustainable without this supplementation from source areas, such that regional population structure may depend on connectivity between these sources and sinks. In extreme cases, sink populations in highly marginal habitats sustain no local reproduction, and rely solely on the immigration of each generation of individuals from source populations.
Deep-sea biodiversity patterns and their underlying causes are of increasing interest given links to ecosystem functioning, and efforts to establish conservation guidelines for deep-sea resource exploitation, including oil extraction, mining and deep-sea fisheries [10–12]. Identification of source and sink habitats has clear relevance to conservation efforts, in that protection of sink habitat is unlikely to preserve and/or revitalize endangered populations, while protection of source habitats may be disproportionately important [7]. Indeed, marine conservation research—particularly efforts to design effective marine protected areas (MPAs)—emphasizes dispersal-mediated ‘connectivity’ to source populations or high-quality habitat patches (e.g. [13]).
The role of source–sink dynamics in governing abyssal community structure has been suggested previously (e.g. [3,4]), and Rex et al. [5] explicitly applied source–sink theory to explain bathymetric species diversity gradients. The well-documented decline in species diversity from mid-bathyal (i.e. mid-continental slope) to abyssal depths in the North Atlantic [14] may primarily be driven by attenuation of particulate organic carbon flux with depth [15]; however, the mechanisms underlying productivity-driven diversity clines continue to be a topic of discussion in deep-sea ecology [3,16,17]. Rex et al. [5] proposed the slope–abyss source–sink (SASS) hypothesis to explain this trend, suggesting that extremely low input of organic matter to abyssal sediments has created a source–sink system between continental margins and the abyssal seafloor, whereby abyssal populations are maintained only by larval dispersal from more productive areas on the surrounding continental slopes. As distance to productive coastal ecosystems increases, source–sink dynamics are said to become increasingly important, because population densities on the food-limited abyssal seafloor are so low as to be reproductively non-viable (i.e. suffer from an Allee effect).
Based on analysis of depth ranges in North Atlantic gastropods and bivalve molluscs, Rex et al. [5, p. 164] note that the abyssal fauna largely represents ‘sparsely occupied range extensions for a subset of bathyal species with high dispersal ability’, arguing against the existence of a truly endemic abyssal fauna. In this view, abyssal populations consist of non-reproductive individuals having dispersed as larvae from slope ‘source’ populations. The SASS hypothesis essentially describes large areas of the abyssal seafloor as a biodiversity sink, rather than a high-diversity region and centre of evolutionary radiation such as has been suggested by others (e.g. [18–21]). As such, the SASS hypothesis implies that anthropogenic activities restricted to the abyss (e.g. manganese nodule mining) are much less likely to result in species extinctions than previously expected, because the reproductively active populations of abyssal species actually lie at shallower depths on the continental slope. Thus, the SASS hypothesis has clear implications for both conservation efforts and evolution in the deep sea.
Rex et al. [5] reported that abyssal molluscs showed depressed α-diversity relative to bathyal depths and were composed of a limited subset of slope taxa with extended bathymetric ranges. Subsequent references to the SASS hypothesis by other authors have focused on this line of evidence, discussing whether the trends observed in molluscs were typical of other taxa and/or regions. Moreno et al. [22] attempted to test predictions of the SASS hypothesis by applying nestedness analysis to polychaete species richness data from the Pacific margin of South America, asking whether the abyssal species were a subset of the shelf and bathyal fauna. A significant pattern was detected, ostensibly supporting the SASS hypothesis. However, the authors reported only 11 polychaete species below depths of 3000 m, and only one species below 4000 m. Other abyssal Pacific studies report approximately 200 polychaete species within relatively small areas at depths below 4000 m, with species continuing to accumulate [23,24], suggesting the dataset of Moreno et al. [22] severely under-sampled the polychaete fauna and cannot be taken as support for the SASS hypothesis. Evidence of high species turnover and low gene flow across depth strata further contradicts the notion that abyssal taxa represent range extensions from slope areas (e.g. [17,18,25–30]). Indeed, distinct benthic assemblages are found within relatively limited depth zones, and numerous species are restricted to abyssal depths.
Our objective is to move beyond discussion of bathymetric trends to formulate and test predictions of the SASS hypothesis concerning abyssal community structure and dispersal dynamics. Specifically, we focus on the prediction that areas of the abyssal seafloor below 4000 m should constitute sink habitats inhabited by populations of benthic organisms that are sustained by the influx of larvae dispersing, in a single generation, from more productive continental slope areas. This idea suggests the following testable predictions:
(1) physical transport is sufficient for propagules from the continental slope to reach the central abyssal regions of the major ocean basins within the time frame of a single pelagic larval period; | |||||
(2) projected larval export from continental slopes to the abyss reproduces global patterns of macrofaunal abundance on the abyssal seafloor; and | |||||
(3) benthic faunal abundance decreases with integrated distance from the continental slope. | |||||
2. Testing predictions of the slope–abyss source–sink hypothesis
Source–sink dynamics are inextricably linked to larval dispersal dynamics, particularly in the case of small, relatively sessile marine invertebrates which rely on planktonic larvae as their primary means of colonizing distant habitat patches. The SASS hypothesis asserts that the abyssal seafloor constitutes a ‘marginal’ habitat (cf. [31]), wherein food is so scarce that organisms are unable to reproduce, creating a demographic sink dependent on influx of larvae that were produced in more productive bathyal source regions. This prediction assumes the majority of abyssal taxa disperse via planktonic larval stages, and that larvae are capable of reaching distant abyssal sink regions by means of transport in deep-sea currents [5].
Evidence of reproductive activity in the abyss is extensive [32–34], suggesting many taxa are capable of allocating resources to gamete production. More importantly, diverse groups of macrofaunal invertebrates (approx. 300–1000 μm) and the vast majority of metazoan meiofauna (≥45 μm) are exclusively direct developers (i.e. they do not produce intermediate larval stages at all), yet these groups dominate the abundance, biomass and diversity of the abyssal benthos [35,36]. Other dominant macrofauna such as the polychaete worms exhibit brooding in many species, producing non-dispersive larval stages [34]. Most notably, the Peracarida (amphipods, isopods, tanaids and cumaceans)—which are among the most speciose groups in the abyss, with an estimated several thousand species [37]—are exclusively brooding taxa; yet some groups exhibit strong evidence of adaptive radiations in situ at abyssal depths [18,19]. Such adaptive radiation is impossible in non-reproductive sink populations [8,31], and, combined with high local diversity, suggests that at least some highly diverse, ecologically important clades are not subject to SASS dynamics. Nonetheless, we consider below the dispersal dynamics of those taxa that do produce planktonic larval stages, and thus could conceivably be subject to source–sink dynamics.
(a) Deep-sea larval dispersal distances estimated from drifter data
The potential spatial scales over which dispersal of deep-sea larvae might occur depend on the speed and trajectory of bottom currents, and the amount of time larvae spend in the plankton (commonly referred to as planktonic larval duration, PLD). Measurements of PLD in individual abyssal taxa are rare, but estimates range from a few days to a few months [38,39]; conservatively, PLDs are likely to be less than 1 year for most taxa. Thus, could dispersive larvae, in particular the demersal forms characteristic of most deep-sea species that produce planktonic larvae, be transported from the slope to the centres of abyssal ocean basins in a single generation?
Deep-sea drifter and tracer-release studies, in which neutrally buoyant floats or chemical tracers are released and tracked below 1500 m depth in the open ocean, provide useful insights into potential transport distances for deep-sea larvae. Numerous studies in the deep Pacific and Atlantic Oceans indicate that over approximately 1-year time scales, transport distances in most of the open ocean (i.e. away from bathyal topography such as continental slopes and mid-ocean ridges) are of the order of 10 to several hundred kilometres per year, except in the equatorial zones and the Southern Ocean [40–45]. Considering that the distance from the nearest slope to the centre of the Pacific Ocean basin is up to 5000 km, and up to 3000 km in the South Atlantic and Indian Ocean basins (e.g. see fig. 3b in [46], and values of DMIN in electronic supplementary material S1), abyssal benthic populations in these regions cannot be maintained by the transport of planktonic larvae in a single hop from slope populations. Shorter distances are required for slope-to-abyss larval transport in the North Atlantic relative to the Pacific, but even here, many central abyssal areas are more than 1000 km from the nearest slope. Thus, we conclude that physical transport is very likely to be insufficient for propagules from the continental slopes to reach the central ocean abyssal regions within a single pelagic larval period.
(b) Modelling larval source–sink dynamics in the abyss
Given the spatio-temporal variability of the velocity field in the deep ocean revealed by drifter trajectories (e.g. [41,42,44]), horizontal dispersal on sufficiently large scales is expected to be random-walk-like (e.g. [47]), implying that concentrations of pelagic larvae originating on the continental slope should decrease as roughly the square of the transport distance from the slope. Thus, even if slope larvae were somehow transported to the central ocean basins, larval densities in the basins would be exceedingly low. True sink populations are recruitment limited, so if abyssal communities are maintained predominantly by larval input from slope sources, abundance of abyssal organisms should reflect the integrated larval input from surrounding slopes. We modelled potential larval transport from the slope to the abyss in order to ask: does projected larval export from continental slopes to the abyss reproduce regional patterns of benthic invertebrate abundance at the abyssal seafloor? Alternatively, are some abyssal areas more likely than slopes to serve as source areas for larvae dispersing to other abyssal sites?
To address these questions, we constructed a simple dispersal model which produces a larval slope source index (LSSI). The LSSI is a relative estimate of the total larval input to a given location on the seafloor from all slope sites within a conceivable dispersal range. The model was initialized by setting the larval concentration of each ‘source pixel’ (topographic pixel in the 2000–3000 m depth range) to the product of the pixel surface area As and particulate organic carbon (POC) flux [48]. The pixel surface area As was calculated by dividing the horizontal area of the pixel (based on latitude/longitude extents) by the cosine of the topographic slope angle. We used 2000–3000 m as the depth range for the source of slope-derived larvae because, for the vast majority of species occurring below 4000 m for which bathymetric ranges are known (including molluscs, sea stars and holothurians), the upper limit of the range is deeper than 2000 m [5,17]. Thus, for most abyssal species, potential source populations on the slope lie below the 2000 m depth contour. For the first model run, the larval concentrations of all source pixels deeper than 4000 m and shallower than 2000 were set to zero. After initialization, the LSSI for each sink pixel was calculated by summing the contributions to that location from all slope source pixels. Each source pixel contribution was calculated by dividing the larval concentration by the squared dispersal distance. Normalizing larval input with the square of dispersal distance is consistent with the type of random-walk dispersal patterns that are suggested by drifter studies, which show convoluted, swirling trajectories.
Here we define dispersal distance as the length of the shortest connection between the source and destination pixels along great-circle segments that are not blocked by topography shallower than 1500 m. The dispersal distance of a source pixel to itself was set to ¼√As, which is an approximation of the mean distance of any location within the pixel to the pixel centre. Our simple dispersal model does not explicitly resolve time, and thus cannot be used to evaluate dispersal distances achievable within a time frame approximating PLD. We initially limited dispersal distances to less than or equal to 2000 km in order to keep computation tractable. This limitation creates an underlying assumption that larval PLDs are sufficient for dispersal across such long distances. However, as discussed above, drifter studies indicate PLDs of 4–5 years would be required for slope larvae to travel such distances (i.e. much longer than published estimates of PLD). The 2000 km dispersal distance used in our model is thus a very generous estimate such that we are probably overestimating the potential recruitment to abyssal populations from slope source areas. Nonetheless, as long as the dispersal distance is sufficiently large to allow flux to all locations on the abyssal seafloor, the model solution is not sensitively dependent on its value, because larval input is normalized to the square of the distance to the source. The closest source areas will thus dominate the spatial patterns of the LSSI. We tested this expectation with additional model runs at low spatial resolution using different dispersal distance limits and confirmed that patterns were similar across runs (results not shown).
Drifter and tracer data from the abyssal Atlantic [44,49] indicate significantly greater zonal than meridional displacement from deployment locations (i.e. ‘dispersal’) on time scales of 2–3 years, except near topography. Zonal tracer tongues occur in all major ocean basins [50], suggesting greater zonal than meridional dispersal may be typical. We thus modified our model to simulate zonal-to-meridional dispersal anisotropy (ζ) by multiplying zonal and meridional distances by 1/√ζ and √ζ, respectively, ensuring that the horizontal surface area of each pixel is conserved.
Initial model construction was based on a reduced topographic dataset derived by pre-averaging the Smith & Sandwell V12.1 topography [51] with a nominal resolution of 1′ into 9 × 9 pixels. To test the effect of this topographic filtering, a run was carried out with the domain limited to the Pacific but with approximately doubled resolution (5 × 5 pixel pre-averaging). Results from the standard- and high-resolution runs are very similar, except near the open boundaries (Drake Passage and a region south of Australia). The main effect of the increased horizontal resolution is that there are more potential larval source areas on seamounts and small islands. However, since the slope area associated with these small-scale topographic features is small, they do not significantly affect large-scale patterns.
(i) Effect of particulate organic carbon flux on larval supply
The SASS hypothesis predicts larval source areas on the continental slope based on the observation that biomass and standing stock decline with depth, presumably due to decreased input of POC [5]. However, the abyssal seafloor is not uniformly low in POC flux. Areas that underlie productive surface waters, such as the Pacific equatorial upwelling zone, receive elevated levels of flux, and thus support elevated biomass and abundance of benthic organisms [52]. We predicted that these high-flux regions not associated with continental margins may also serve as larval source areas for dispersal into the oligotrophic central basins. We thus modified the LSSI model to allow scaling of larval production with seafloor POC flux from Lutz et al. [48]. Model runs with a zonal-to-meridional dispersal distance anisotropy factor of 3 show interesting trends in the LSSI (figure 1). In particular, high-productivity regions associated with both coastal and equatorial upwelling zones impact the LSSI, indicating these regions may account for a disproportionately large part of larval export relative to other slope areas (figure 1b). More interestingly, additional model runs relaxed the depth restriction for larval source areas and considered the entire abyss as a potential larval source, scaled again with POC flux, which is an important predictor of benthic biomass [53]. In this scenario, both POC flux and distance from slope can influence larval supply to a given abyssal location, yet large-scale patterns in the LSSI closely resemble POC flux patterns (figure 1c). This result is consistent with our prediction that local productivity plays an important role in governing larval supply. Indeed, the importance of POC flux as a driver is enhanced by considering abyssal regions as potential larval sources.
Figure 1. (a–c) Dispersal distance normalized larval input calculated with zonal-to-meridional dispersal distance anisotropy factor of 3. Because larval input is a relative quantity, the data in each of the three panels were scaled with a constant factor into a common range (20–5000) and are reported without units. (a) Constant larval source strength, accounting only for distance to slope. (b) Source strength scaled to POC flux, with larval supply from continental slope only. (c) Source strength scaled to POC flux, allowing entire abyss to serve as a larval source area. (d) Estimated abundance of benthic macrofauna at the deep seafloor smoothed to 0.1 degree resolution and displayed in logarithm scale (base 10). Adapted from [53].
The results of our LSSI modelling are striking. The models based solely on slope larval sources (figure 1a,b) produce macrofaunal abundance patterns dramatically different from those documented in the deep sea (figure 1d). In particular, vast regions of the northeast and equatorial North Pacific abyss exhibit very weak larval input. If the supply of larvae from continental slopes is limiting abyssal macrobenthic abundance, these areas should have macrofaunal abundance approximately two orders of magnitude lower than abyssal areas near continental margins. In fact, these are regions of high abundance [53] (figure 1d). Thus, the prediction that projected larval export from slopes to the abyss reproduces regional patterns of macrofaunal abundance at the abyssal seafloor can be rejected. By contrast, when abyssal depths of more than 4000 m are included as potential larval sources (larval source strength varying with seafloor POC flux as in earlier runs), modelled larval inputs reproduce most regional patterns in abyssal macrofaunal abundance (figure 1c,d). Thus, if source–sink dynamics are important in controlling abyssal macrofaunal abundance, some abyssal regions are likely to be important source habitats. In this model, large regions of the abyssal seafloor emerge as potential larval sources (figure 1c), including areas underlying productive surface waters such as the equatorial Pacific upwelling zone. Source–sink processes controlling population densities are thus likely to be influenced by regional productivity patterns, rather than resulting solely from dispersal of larvae from distant slopes.
(c) Trends in benthic faunal abundance with distance from slope
Source–sink theory describes sink habitats as having a carrying capacity of zero, such that species abundance in these low-quality habitats will decline in the absence of immigration [7,8]. Moreover, abundance within a sink habitat is a function of productivity in the source habitat, and resulting rate of dispersal to the sink. Thus, the abundance of abyssal species would be expected to decline with distance from the productive source habitat, because habitat quality decreases due to low productivity, and populations become increasingly dispersal-limited with distance to the source population [5]. Here we ask: Does macrofaunal standing stock, or the standing stock of predominant macrofaunal taxa, decrease with the distance from the nearest continental slope?
We used existing datasets for macrofaunal abundance in the abyssal North Pacific [23,52,54–60] (electronic supplementary material, S1 and S2) to test the prediction that benthic faunal abundances are inversely related to distance from the continental slope (i.e. that standing stock declines with distance to slope). Two metrics were calculated for each abundance data point in order to quantify the range of dispersal distances required for colonization of abyssal sediments from the continental slope. These metrics were based on the approximate position of the 2000 m isobath along the continental slope (commonly used to delineate slope from abyssal seafloor), north of 15° S, and include the shortest linear distance to the isobath (DMIN) and the integrated distance (∑(1/r2)) from 868 equally spaced points along the isobath (DSUM) to account for dispersal from multiple locations along the slope (electronic supplementary material S1). If faunal abundance is inversely related to distance to the continental slope, then we would expect a negative relationship between faunal abundance and both DMIN and DSUM.
We conducted a stepwise multiple regression using latitude, longitude, depth, DMIN, DSUM and POC flux to the seafloor (derived from Lutz et al. [48]) as predictor variables, to evaluate potential predictors of macrofaunal abundance. Latitude and longitude were initially included to account for spatial correlations among sampling locations. However, evaluation of variance inflation factors for predictor variables indicated problems with colinearity between longitude and depth, and between DMIN and DSUM. Thus, DMIN and longitude, which had the highest variance inflation factors, were excluded from the regression analysis. Separate analyses were conducted using abundance data for total macrofauna and for dominant macrofaunal taxa including polychaetes, bivalves, gastropods, total molluscs, tanaids and isopods (table 1). Examination of diagnostics indicated that all datasets met the assumptions of normality and equal variances, so all analyses were conducted using untransformed data with the software SPSS Statistics v. 22.
| taxonomic group | beta coeff | s.e. | t-value | p-value | partial correlation coeff | adj R2 |
|---|---|---|---|---|---|---|
| total macrofauna | ||||||
| intercept | −651.243 | 211.292 | −3.082 | 0.003 | ||
| POC flux | 805.977 | 99.636 | 8.089 | <0.001 | 0.683 | 0.516*** |
| DSUM | −4.308 | 1.740 | −2.476 | 0.016 | −0.275 | 0.547*** |
| Polychaeta | ||||||
| intercept | −471.397 | 76.247 | −6.182 | <0.001 | ||
| POC flux | 409.018 | 46.805 | 8.739 | <0.001 | 0.713 | 0.501*** |
| Bivalvia | ||||||
| intercept | −44.304 | 11.504 | −3.851 | <0.001 | ||
| POC flux | 39.141 | 7.062 | 5.543 | <0.001 | 0.542 | 0.284*** |
| Gastropoda | ||||||
| intercept | −7.576 | 5.779 | −1.311 | 0.194 | ||
| POC flux | 8.897 | 3.548 | 2.508 | 0.014 | 0.280 | 0.066** |
| total Mollusca | ||||||
| intercept | −51.880 | 14.367 | −3.611 | 0.001 | ||
| POC flux | 48.038 | 8.819 | 5.447 | <0.001 | 0.535 | 0.277*** |
| Tanaidacea | ||||||
| intercept | −137.814 | 26.435 | −5.213 | <0.001 | ||
| POC flux | 151.806 | 12.479 | 12.165 | <0.001 | 0.818 | 0.691*** |
| DSUM | −0.864 | 0.217 | −3.982 | <0.001 | −0.422 | 0.742*** |
| Isopoda | ||||||
| intercept | −114.111 | 18.844 | −6.056 | <0.001 | ||
| POC flux | 93.758 | 11.567 | 8.105 | <0.001 | 0.686 | 0.463*** |
Regression analysis revealed significant relationships to DSUM only for total macrofauna and Tanaidacea. In both cases, abundance was negatively related to DSUM, indicating an increase in abundance with distance to the slope, in contrast to predictions of the SASS hypothesis. For all other taxa, only POC flux was retained in the regression model as a significant predictor of abundance. For all categories, abundance showed a positive trend with increasing POC flux. These results suggest that dispersal from shelf and slope populations alone is not the limiting factor controlling macrofaunal densities observed in Pacific abyssal habitats, and that productivity is more likely to be driving patterns of macrofaunal abundance.
3. Suggested revision of the slope–abyss source–sink hypothesis: the oligotrophic sink hypothesis
Source–sink population dynamics ultimately arise from differences in ‘habitat quality’ that can occur over a variety of spatial scales [9]. At the abyssal seafloor, some environmental parameters typically used to characterize habitat quality, such as temperature, flow regime and substrate type, can be relatively uniform over tens to thousands of kilometres. However, food availability in the form of surface-derived POC flux has long been known to exhibit spatial and temporal heterogeneity in the deep sea, from the landscape scale of abyssal hills and valleys to the regional scales of oligotrophic gyres and upwelling zones [2]. Direct measurements of POC flux at abyssal depths are relatively rare, but recent modelling efforts estimate fluxes for the global ocean that clearly demonstrate this variability in food availability (e.g. [48]). Heterogeneous POC flux could thus produce areas of higher or lower habitat quality over spatial scales relevant to larval dispersal.
Biomass and diversity are generally elevated in high-flux areas of the deep sea (e.g. [23,35,61]) distant from slope or shelf areas, suggesting these regions constitute high-quality habitat relative to areas underlying oligotrophic central gyres. These areas (e.g. the equatorial Pacific upwelling zone) are just as likely to function as ‘sources' of dispersive larvae as are productive areas along the continental margins. Considering higher-flux abyssal regions as potential larval source areas greatly increases both the effective population size and the geographical area from which to draw the pool of potential colonists for abyssal sediments. Moreover, potential distances larvae must travel to reach low-productivity regions are significantly reduced. Considering source–sink dynamics as a eutrophic–oligotrophic phenomenon encompassing the entire deep seafloor thus overcomes limitations of the strict slope–abyss interpretation. Nonetheless, for species lacking long-distance (approx. 1000 km) dispersal abilities (such as diverse groups of direct developers and brooders), even the oligotrophic central gyres are very unlikely to contain pure sink populations.
(a) Identifying abyssal source and sink habitats
According to classic definitions of source and sink habitats (cf. [9,31]), demographic sinks are areas in which local population growth rates are less than 0; thus, sub-optimal habitats are often demographic sinks, while source areas are found in areas of high-quality habitat. For the food-poor abyssal seafloor, productivity is likely to be the most effective predictor of habitat quality for many soft-sediment organisms. Analyses presented above add additional support to the body of evidence that benthic standing stock in the abyss is best predicted by POC flux, suggesting that the abyssal equatorial Pacific Ocean and Southern Ocean may serve as source areas of diversity for taxa with long-distance dispersal that can enhance the regional species pool. Productivity patterns may further distinguish areas where slope–abyss source–sink dynamics are particularly unlikely to occur. For example, the abyssal Southern Ocean is within reach of the Antarctic continental shelf and is also characterized by relatively high levels of POC flux [48]. Thus, any effects of dispersal from the slope on abyssal populations are likely to be masked by effects of high local productivity [62]. In fact, the abyssal Southern Ocean may have provided a source of propagules to re-colonize the Antarctic slope and shelf following species extinctions during glacial periods [62,63].
Particular taxa may be more or less susceptible to source–sink dynamics based on life-history patterns. In particular, populations of species lacking dispersive larval stages will be largely unaffected by source–sink dynamics [64,65], and as mentioned above, these groups represent the most abundant and diverse taxa found in the abyss. However, adaptive evolution in sink habitats can lead to increased survival and reproduction, to the degree that the habitat no longer functions as a sink [31]. In particular, sink populations that are unsustainable without dispersal or immigration and do receive a supply of immigrants may experience high rates of niche evolution [66]. These populations may, following lengthy periods of maladaptation to the sink habitat, shift rapidly into an adapted state in which stable population size no longer depends on immigration. In this scenario, extreme oligotrophy may have led to new adaptations and adaptive radiation. Currently stable abyssal populations may well have functioned as sinks in the past, especially if they evolved from slope species. In short, many species-rich, poorly dispersing taxa appear to be well adapted to abyssal oligotrophic conditions, whereas taxa with widely dispersing larvae (e.g. gastropods and bivalves) may be poorly adapted, but persist as sink populations in areas of the oligotrophic abyss.
(b) Testing the ‘oligotrophic sink hypothesis' in the deep sea: directions for future research
Some have suggested that the only true test of whether an area functions as a sink habitat is to isolate it from dispersal and then to monitor it to see if purported ‘sink’ populations disappear [31,67]; this is not a tractable scenario for abyssal seafloor studies. Nonetheless, two key pieces of information may indicate whether source–sink dynamics exist between two habitat patches: detailed demographic data to confirm that local population growth rates are less than 0 in the sink habitat, and information on directional movement or dispersal between sources and sinks [67]. Demographic data could be used to distinguish a true demographic sink from what is simply a lower-quality habitat with reduced carrying capacity. Alternatively, indicators of habitat quality (e.g. food availability) may be evaluated with respect to their effects on population dynamics, reproduction rates or individual fitness. For example, measurements of fitness such as fecundity or proportions of reproductive individuals could be compared between proposed source and sink habitats.
Directional dispersal between abyssal habitat patches may be more challenging to quantify. Modelling may be needed to estimate transport rates of small particles in abyssal currents. However, empirical observations are also needed to constrain such models, and these observations are still rare at abyssal depths. Genetic tests of population differentiation are frequently applied in marine dispersal studies and can yield some evidence of the distances over which dispersal is likely to be acting (e.g. [68,69]), and/or the presence of potential sink populations (genetic diversity is often depressed in marginal or sink populations due to small effective population size [31]). However, genetic studies provide limited information on directionality of dispersal, and are particularly poorly suited to examining processes occurring over ecological time scales, such as single dispersal events. For example, immigration rates required to maintain genetic connectivity (often taken to be one individual per entire population per generation) can be orders of magnitude lower than those required to maintain a non-reproducing sink population (equivalent to at least the average population density per unit area per generation) [70]. Ultimately, identifying the directions and rates of abyssal larval dispersal to maintain potential sink populations will require studies of abyssal hydrodynamics, as well as larval distribution and behaviour, and may require innovative larval tracking and modelling techniques such as those recently applied in shallow-water ecosystems (e.g. [71,72]).
Data accessibility
Source code for the LSSI model has been deposited to Dryad Digital Repository http://dx.doi.org/10.5061/dryad.1s445. The dataset used for regression analysis was compiled from multiple sources, as described in the text, and uploaded as electronic supplementary material.
Acknowledgements
We thank M. Rex and C. McClain for thoughtful and lively discussions that helped shape the ideas presented here. We also thank C. Morgan and D. Tittensor for assistance with data processing. We thank the ABYSSLINE Project (Seabed Resources Development Ltd) for their support of CRS, resulting in insights into the structure of abyssal communities that shaped this manuscript.
Funding statement
The International Seabed Authority supported the compilation of abundance data used in the modelling effort through an award to C.R.S.
Authors' contributions
S.M.H. compiled the macrofaunal abundance data from the abyssal Pacific and conducted the regression analysis. A.M.T. and C.R.S. designed and constructed the LSSI model, conducted model runs and produced figure 1. S.M.H. drafted the manuscript. All authors collaborated in design of the study, contributed towards manuscript writing and gave final approval for publication.
Conflict of interests
The authors have no competing interests.
