Quantifying Assemblage Turnover and Species Contributions at Ecologic Boundaries

Not all boundaries, whether stratigraphical or geographical, are marked by species-level changes in community composition. For example, paleodata for some sites do not show readily discernible glacial-interglacial contrasts. Rather, the proportional abundances of species can vary subtly between glacials and interglacials. This paper presents a simple quantitative measure of assemblage turnover (assemblage turnover index, ATI) that uses changes in species' proportional abundances to identify intervals of community change. A second, functionally-related index (conditioned-on-boundary index, CoBI) identifies species contributions to the total assemblage turnover. With these measures we examine benthonic foraminiferal assemblages to assess glacial/interglacial contrasts at abyssal depths. Our results indicate that these measures, ATI and CoBI, have potential as sequence stratigraphic tools in abyssal depth deposits. Many peaks in the set of values of ATI coincide with terminations at the end of glaciations and delineate peak-bounded ATI intervals (PATIs) separated by boundaries that approximate to glacial terminations and to transgressions at neritic depths. These measures, however, can be used to evaluate the assemblage turnover and composition at any defined ecological or paleoecological boundary. The section used is from Ocean Drilling Program (OPD) Hole 994C, drilled on the Blake Ridge, offshore SE USA.


Introduction
Biostratigraphers historically have sought means to subdivide the sedimentary stratigraphic record as finely as possible. There is, however, evidently a limit to the zonal resolution that can be attained using a single fossil group [1]. For example, the majority of the Pleistocene, the base of which is placed at 2.588 Ma [2], is currently ascribed to the Globorotalia truncatulinoides truncatulinoides (dOrbigny) planktonic foraminiferal Zone [3] or to Zones PT1a and PTIb based on planktonic foraminifera [4] (author names are given at the first mention of any species).
It has long been appreciated, however, that glacial and interglacial fauna and flora within the Pleistocene differ [5], both on land [6] and in the oceans [7]. In Chile, for example, plant leaf architecture changes from a mixture of species belonging to the cool temperate North Patagonian Forest and more thermophilous rain forest vegetation between glacials and interglacials [8]. Glacial-interglacial contrasts in the insect community have been recorded in Greenland [9]. Such changes have been recorded among some foraminifera, but not at all sites. Schott [10] found that Globorotalia menardii (d'Orbigny) in the Indian Ocean was more abundant in interglacial than in glacial sediments. Phleger et al. [11], in a study of North Atlantic foraminifera, noted the presence in piston cores of ''layers of faunas normal for their latitude alternating with faunas typical of a higher latitude,'' while Bandy [12] was able to distinguish glacials from interglacials off southern California using the ratio between populations of sinistrally and dextrally coiled Neogloboquadrina pachyderma (Ehrenberg). Streeter [13] found that Atlantic benthonic foraminifera at depths .2500 m varied greatly over the last 150 ka and suggested that this arose because of depression and elevation of faunas through a depth range of several hundred meters between glacials and interglacials [14]. Streeter and Lavery [15] recorded that uppermost Pleistocene faunas in cores from the western North Atlantic slope and rise north of 35uN were dominated by Uvigerina, but that Hoeglundina dominated in the Holocene. This faunal transition was apparently diachronous, occurring at ,12 ka at 3,000 m, but at ,8 ka at 4,000 m. Thomas et al. [16] examined benthonic foraminifera in two lower bathyal (,1700 m) and abyssal (,3500 m) piston cores spanning the last 45 ka in the northeastern Atlantic Ocean. They found that Epistominella exigua (Brady) and Alabaminella weddellensis (Earland), which bloom opportunistically where a spring plankton bloom produces a pulse of phytodetritus, were rare during the last glacial maximum but abundant in the Holocene. In contrast, Cassidulina, Pullenia, bolivinids, buliminids and uvigerinids were common during glacial MIS (marine isotope stages) 2, 3 and 4, although the interglacial MIS 3 was not as warm as other Late Quaternary interglacials [17]. Thomas et al. [16] suggested that this reflects an enhanced organic carbon flux during glacials, rather than sluggish glacial bottom circulation leading to poorly oxygenated bottom water. This may be related to the plankton multiplier effect proposed by Woods and Barkmann [18], in which a diminished greenhouse effect during glacials reduces radiative forcing of the ocean, increasing the depth of winter convection. This in turn increases the annual resupply of nutrients to the euphotic zone, which leads to increased annual primary production. Gaby and Sen Gupta [19] found glacial and postglacial assemblages of the abyssal Venezuela Basin to differ, the Holocene fauna containing abundant Cibicides wuellerstorfi (Schwager), Melonis pompilioides (Fichtel and Moll), Nuttallides umbonifera (Cushman), and Pullenia sp., while the fauna in the last glacial was dominated by Massilina sp., Globocassidulina subglobosa (Brady) and Nummoloculina irregularis (d'Orbigny).
Glacial-interglacial contrasts in the benthonic foraminiferal fauna are not everywhere marked, however. Streeter and Lavery [15] wrote that on the western North Atlantic continental rise below 4,000 m, ''the glacial to modern faunal shift is subtle, but it clearly occurs later than on the upper rise.'' Sen Gupta et al. [20] examined benthonic foraminifera over the past 127 ka in three bathyal cores (depths near 2000 m) from the western Grenada Basin, eastern Caribbean Sea. They found only subtle changes, rather than drastic turnovers, at glacial-interglacial boundaries based on the abundance of Globorotalia menardii. They stated that neither species richness S nor the information function H~{ P p i : ln (p i ), where p i is the proportional abundance of the ith species) showed any distinct stratigraphic trend (although H is not expected to show such a trend [21]). However, they suggested Nuttallides umbonifera, Bulimina buchiana dOrbigny and Chilostomella oolina Schwager to be rarer in the last glacial than in the two bounding glacials. Wilson [22,23] examined the benthonic foraminifera in two bathyal piston cores near the northern Leeward Islands, eastern Caribbean Sea. He did not find any marked faunal changes at the Pleistocene-Holocene boundary, but showed that the organic flux in one core decreased gradually through the entire core. Wilson [24] found only weak evidence of Milankovich cycles in the Upper Quaternary of ODP Hole 1006A (Santaren Channel, offshore western Bahamas), where Globocassidulina subglobosa and Cibicidoides aff. C. io (Cushman) were smaller assemblage components during most glacial MISs. However, the percentages of these species varied between odd-numbered MISs and they were insignificantly correlated with one another, G. subglobosa being rare in MIS 9 while C. aff. C. io was common.
The inability to detect glacial-interglacial contrasts at all sites appears to arise because not all sites show marked changes in community composition at the species level at glacial-interglacial boundaries. Rather, the proportional abundances of species vary between glacials and interglacials to differing degrees. This paper presents a simple quantitative measure, the assemblage turnover index (ATI), which uses changes in species' proportional abundances to identify intervals of marked community change. Whittaker [25] distinguished two categories of diversity: inventory diversity, which calculates the diversity of associations within samples (point diversity) or habitats (a diversity); and differentiation diversity, which examines the change in diversity between samples (pattern diversity) or habitats (b diversity). The assemblage turnover index presented here is a form of differentiation diversity. A conditioned-on-boundary index (CoBI), developed as a function of the ATI, identifies species that contribute most to maxima and minima in sets of values of the ATI. The section used in this demonstration is from Ocean Drilling Program (OPD) Hole 994C, drilled on the Blake Ridge, offshore SE USA ( Figure 1). Although Bhaumik and Gupta [26,27] and Mohan et al. [28] have examined Neogene benthonic foraminifera from this and nearby ODP Holes, glacial-interglacial contrasts have not been recorded at the Blake Ridge before this study.

Site Description
ODP Site 994 (31u47.1399N, 75u32.7539W; water depth 2799 m) is situated on the sediment drift deposit that forms Blake Ridge [29,30], a drift deposit consisting of current-lain sediment east of the Blake Plateau. The Blake Ridge was deposited by the Western Boundary Undercurrent, this being a thermohaline-induced contour current (a current that flows parallel to bathymetric contours), which flows southward along the U.S. continental margin [31]. The Western Boundary Undercurrent transported clays eroded from eastern North America north of 40uN, at least as far south as Puerto Rico [32]. ODP Hole 994C cored a 700 m-thick succession of clays with calcareous nannofossils within which there are no obvious depositional hiatuses [30]. In this paper we examine the topmost ,14 m of sediment in ODP Hole 994C. Mass sediment transport complexes are absent. The studied section, which is a distinct lithostratigraphic unit termed Unit 1 [30], consists of light gray to gray and greenish gray nannofossil-rich clays in beds up to 1.20 m thick. The presence of the trace fossil Zoophycos indicates that some bioturbation has taken place, but mostly at bedding planes. Biostratigraphic correlation within the Quaternary of ODP Hole 994C is limited. Okada [33] found the first occurrence of Emiliania huxleyi (Lohmann) at 8.05-9.05 meters below the seafloor (mbsf), between OPD Hole 994C Core 2H-3, 65 cm and 2H-4, 15 cm, for which he suggested an age of 0.26 mya. This indicates a depositional rate of 3.5 cm/ka in the uppermost part of the Hole. The first appearance of E. huxleyi has subsequently been placed between 0.262-0.264 Myr [34].
Bhaumik and Gupta [27] studied the benthonic foraminifera in nearby ODP Hole 997A, and found Brizalina paula (Cushman and Cahill), Cibicidoides kullenbergi (Parker), Uvigerina hispidocostata Cushman and Todd and Uvigerina peregrina Cushman to be abundant in that part of the section within the gas hydrate zone. Bhaumik and Gupta [26]

Materials and Methods
Sixty seven samples of 20 cm 3 were taken at 20 cm intervals from ODP Hole 994C, Cores 1 and 2, between 0.08-13.25 mbsf. They were provided by the Ocean Drilling Program (ODP) that is sponsored by the U.S. National Science Foundation (NSF) and participating countries. The cored site being in international waters, no specific permissions were required for these locations/ activities. The material comprising fossils, sampling did not involve endangered or protected species. Each sample was ,2 cm thick and represents ,600 years. Samples were soaked in water until disaggregated, washed over a 63 mm mesh to remove silt and clay, and dried over a gentle heat. An attempt was made to pick N = 250-300 specimens of benthonic foraminifera from the .63 mm fraction from each sample. However, only 42 samples yielded .250 specimens (mean, 251 specimens per sample, minimum 104). The methods on assemblages used here have been reported by Wilson [35]. The foraminifera were sorted into species and identified using Cushman [36,37,38,39,40,41,42], Cushman and Henbest [43], Phleger and Parker [44], Phleger et al. [11], Parker [45], Pflum and Frericks [46] and Mohan et al. [28]. The number of specimens (n i ) was recorded for each species or species group (i.e., rare species in the same genus that were left in open nomenclature and grouped together).
Elphidium excavatum (Terquem), Epistominella takayanagii Iwasa, Quinqueloculina poeyana d'Orbigny and Quinqueloculina ex gr. lamarckiana d'Orbigny, which are typical of neritic water, were regarded as allochthonous and excluded from this analysis. Elphidium is typically regarded as a shallow-water genus [47] that has to be removed from data sets of studies of bathyal foraminifera [48]. Sen Gupta and others [49] recorded abundant E. excavatum on the Louisiana continental shelf. In the southern North Sea, it dominates the foraminiferal fauna at depths between 25-30 m [50]. Epistominella takayanagii has been recorded from Chaleur Bay eastern Canada, mostly in waters ,100 m deep [51], and may have been transported southwards to ODP Site 994C. The proportional abundance of both E. excavatum and E. takayanagii peaked in MIS 10, glacial cycle E, as did the percentage of overall allochthonous, shallow-water species. A single specimen of Stilostomella lepidula (Schwager) recovered from 5.45 mbsf was presumed to be reworked, this species having gone extinct during middle Pleistocene times [52], and was excluded from further analysis. This left a presumed predominantly in situ abyssal assemblage, within which there may have been some slight downslope transport of Angulogerina occidentalis (Cushman), Bulimina aculeata d'Orbigny, B. alazanensis Cushman, Cibicides sp., Fursenkoina fusiformis (Williamson), Globocassidulina obtusa (Williamson) and Nuttallides rugosa (Phleger and Parker) [35]. This presumed in situ assemblage forms the subject of the remainder of this paper. To examine turnover of an entire assemblage quantitatively across a delineated boundary we developed the ATI index. For a set of samples from a given site, the Assemblage Turnover Index for each pair of adjacent samples is defined as in which p i1 and p i2 are the proportional abundances of the ith species, i = 1,…, s, in the lower and upper samples (see Appendix S1 for a glossary of terms). This assemblage turnover index between samples will be denoted as ATI s . Note that although for each sample P p i~1 , the measure ATI s can be .1. Thus, ATI s gives the proportion or percent of turnover or change specifically across a defined or particular boundary. We calculated the mean x x, and standard deviation, s, of values of ATI s over all samples within the core. To develop our control chart we determined all points with ATI s w x xzs ð Þ, which were then deemed to be positions of major turnover. Oba et al. [53] presented a d18O curve for ODP Hole 994C (see Figure 2). Their samples were taken at irregular intervals (sample spacing 7-49 cm; mean 22.6 cm, sd 10.9 cm). The values of d18O for the samples used here were interpolated from Oba et al. 's [53] curve and correlation between ATI s and interpolated d18O was calculated. Because Oba et al. 's [53] uppermost sample was taken at 0.14 mbsf, it was not possible to estimate the d18O value for the uppermost sample picked for this study. Point (sample) values of species richness S and the information function H were calculated. Dominance was determined using max(p i ), the proportional abundance of the most abundant species in each sample [54]. We chose to calculate correlations between ATI s , S, H and max(p i ) using the upper (younger) sample in each sample pair.
Peaks in ATI s , those values larger than our designated control value of x xzs ð Þ, were used to divide the succession into peakbounded ATI s (PATI-) intervals. These intervals were numbered, commencing from PATI-1 for the most recent. PATI-1 and the oldest PATI are incomplete, their upper and lower boundaries respectively not being bounded by ATI s peaks.
To assess which species contributed most to the ATI at the PATI boundaries, a conditioned -on-boundary index CoBI was derived. CoBI provides the proportion that each species within an assemblage contributed to the change or turnover specifically across the PATI boundary. For each species at any PATI boundary where p ij , j = 1,2 are the ith species proportions on either side of the selected boundary of interest and at which the ATI is calculated.
There are two forms of CoBI: 1. Partial conditioned-on-boundary index, CoBI p , in which the assemblage turnover index ATI s was calculated between the entire set of samples within the PATI below the ATI peak and the first sample immediately above the peak. In this case, the ATI is designated as ATI p . The value of ATI ( = ATI p ) was substituted into equation (2), as were p i1 , the proportional abundance of the ith species in the entire PATI below the peak in ATI p , and p i2 , the proportional abundance of that ith species in the first sample above the peak in ATI s. The proportional contribution of each species to ATI p was assessed from the vector of CoBI p values at each ATI s peak. 2. Thorough conditioned-on-boundary index CoBI t , in which the ATI is denoted as ATI t , was calculated between the values in two complete PATIs separated by the peak in ATI s (see Figure 2). The value of ATI t was substituted into equation (2), as were p i1 and p i2 , the proportional abundance of the ith species in the two PATIs separated by the peak in ATI t . The proportional contribution of each species to the value of ATI t was assessed from the vector of partial CoBI for each ATI s peak.
Thus, to detect change between the total set of samples from the assemblages within two distinct PATIs we evaluate the ATI = ATI t at the boundary between these two. The partial indices are used to detect assemblage change exactly at the boundary between a single PATI and the next contiguous sample.

Assemblage Turnover Index (ATI = ATI s ) Between Samples
Our presumed in situ, abyssal fauna comprised 16,184 specimens in 157 species (see File S1), and was dominated by  Figure 3.
The assemblage turnover index between adjacent samples ranged from ATI s = 0.263-1.421(x = 0.710, s = 0.233) (Figure 2), indicating total assemblage change from 26% to 142%. The value of ATI s exceeded x+ s = 0.943 across nine pairs of samples. We chose to include the borderline value of ATI s at 9.85 mbsf, where it nevertheless formed a pronounced peak. We computed correlations of ATI with the indices H and max (p i ) and with d18O. Although the formulae for these measures utilize the relative abundances, there is no linear functional relationship among them that necessitates a significant correlation. The ATI s was positively correlated with the information function H for the younger sample in the pair (r = 0.62, p,0.0001). ATI s was negatively correlated with max(p i ) (r = -0.65, p,0.0001), which indicates a change in dominance across peaks in ATI s , and negatively correlated with d18O (r = -0.32, p,0.01). H and d18O were in turn significantly negatively correlated (r = -0.53, p,0.0001).

Thorough Conditioned-on-Boundary Index (CoBI t )
The ATI t used for CoBI t ranged between 0.36-1.33 (PATI-10/ 9 and PATI-4/3 boundaries respectively; Table 2), or an observed Assemblage Turnover at Boundaries PLOS ONE | www.plosone.org total assemblage change ranging from 36% to 133% between the PATI pairs from PATI-1/2 through PATI-10/11. Thirty three species resulted in a value of CoBI t .0.02 (i.e., accounted for .2% of ATI t ) at any one PATI boundary. Twenty eight species contributed to .0.02 of both ATI p and ATI t . Seventeen species had a CoBI t of 0.02-0.05 at any one boundary. Thus, only sixteen species (,10% of all species recorded) had a CoBI t .0.05 across any one PATI boundary. The highest CoBI t was 0.31 for Brizalina lowmani, indicating pronounced change in dominance across the PATI-3/2 boundary, while Bulimina aculeata presented a comparable CoBI t of 0.29 across the PATI-11/10 boundary. All other species contributed a maximum CoBI t ,0.20, although the maxima for Globocassidulina obtusa and Melonis baarleeanus were 0.19 and 0.17 respectively. The CoBI t for B. lowmani was .0.02 across eight PATI boundaries, while those for G. obtusa were .0.02 across six boundaries and for M. baarleeanus and Cibicidoides robertsonianus across five.

Discussion
The assemblage turnover index is a form of differentiation diversity sensu Whittaker [25]. It is here presented at two scales, of which the end members are (a) ATI s , which corresponds to Whittaker's (1972) pattern or between-sample diversity, and (b) ATI t , which is here taken as corresponding to his between-habitat or b-diversity. ATI s was strongly positively correlated with the information function H and negatively correlated with max(p i ) for the younger of the samples in the sample pairs used in its calculation. This indicates that assemblage turnover -the sum of the changes in proportional abundances of species -increases with increasing diversity and with decreasing dominance (i.e. increasing equitability).
The correlation between ATI s, and interpolated values of d18O was significant and negative ( Figure 2). Oxygen isotopes have been used to erect a paleotemperature record of marine isotope stages (MISs) that is reliable back to MIS 16, 650 ka [17], within which odd numbered MISs are interglacials [55] and against which faunal changes can be compared. Much of the time between MIS 1-12 (the interval examined in this study) consists of 100 ka MIS couplets [56,57]. Broecker and Van Donk [58] grouped the MISs into glacial cycles (segments of two to four MISs) that were separated by terminations (i.e., pronounced boundaries between isotopic maxima and minima). Because warming during deglaciation occurs more rapidly than does cooling during the development of glaciation [59], each termination separates a preceding glacial from a succeeding interglacial. MIS 3 being a subdued glaciation, there is no termination between MIS 4 and 3, but Termination T-I occurs between MIS 2 and 1, and T-II between MIS 6 and 5. The interval examined during this study encompasses terminations T-V to T-I, which separated Glacial Cycles F to A (Figure 3). Cheng et al. [60] positioned terminations at the mid-point between the peaks and troughs in the graph of d18O. However, because the d18O curve for ODP Hole 994C presented by Oba et al. [53] was based on irregularly spaced samples their technique cannot be used here. It does appear, however, that the boundaries between PATI-11/10, 10/9, 7/6 and 6/5 and 2/1 approximate to terminations T-V through T-I, respectively. Thus, the slower onsets of glacials are marked by low levels of turnover, ATI s , while the more rapid transitions to interglacials are marked by peaks in ATI s . Because the peaks in ATI s occurred within terminations, the correlation between ATI s and d18O is low.
Not all peaks in ATI s detected across our samples coincide with terminations. The boundary between PATI-9/8 and 8/7 occurred within MIS 8 and indicates an increase in the flux of organic carbon through that glacial MIS. This may reflect increasing efficiency of the plankton multiplier of Woods and Barkmann (1993).
The close grouping of the boundaries between PATI-5 through PATI-1, all of which occurred during the transition from MIS 2 to MIS 1, show this to have been an interval of protracted environmental change at ODP Site 994. Sea level rose by ,120 m during termination T-1 [61], but did so in several decimeter steps [62]. It is possible that the closely spaced Assemblage Turnover at Boundaries PLOS ONE | www.plosone.org boundaries from PATI-5 through PATI-1 reflect these steps. Brizalina lowmani did not decrease in proportional abundance across all boundaries between PATI-5 through PATI-1, but increased across the PATI-3/2 boundary. Some data suggest that the changes in the fauna across the peaks in values of ATI s reflect changes in either (a) dissolved oxygen levels, (b) the organic carbon flux and (c) bottom current strength, although the first two of these factors are frequently correlated [63]. For a paleoenvironmental summary, see Table 3.
den Dulk et al. [64] studied benthonic foraminifera under an upwelling system in the northern Arabian Sea. They recognised two groups of foraminifera:  [65] found that Bulimina aculeata, which dominated in PATI-11 ( Figure 3B), lives primarily where the flux of organic carbon exceeds 3 g m 22 yr 21 . Melonis baarleeanus (Figure 3E), although placed in group 1 by den Dulk et al. [64], was shown by Qvale and Van Weering [66] to prefer a fine-grained substrate with a relatively high organic carbon content [67]. Mackensen et al. [68] suggested that in the South Atlantic Ocean it prefers seasonally varying productivity. Taldenkova et al. [69] found this species to be more abundant in  the upper bathyal Holocene of the Arctic Ocean than the latest Pleistocene, and ascribed it to a distal-river group of relatively deep-water species that thrive on slightly altered organic matter and is therefore restricted to areas with periodic delivery of organic matter. Murray [70] noted that M. baarleeanus has been recorded live in all oceans except the Indian Ocean. In ODP Hole 994C this species accounted for .0.02 of the CoBI t across six of the ten PATI boundaries, and was abundant in the early part of PATI-10 and in PATI-5 and PATI-4. It was rare to absent in PATI-3 through PATI-1. This suggests that seasonality varied through the Late Quaternary at Blake Ridge. Unlike in the Arctic Ocean [69], at Blake Ridge seasonality was much reduced in the Holocene, after termination T-1.
Globocassidulina subglobosa, which is found throughout the Atlantic, Pacific and Southern Oceans (Murray, 2013), has been suggested to be an oxic indicator [71] that prefers an elevated mean organic carbon flux of 15 g m 22 yr 21 [72]. This species was abundant in PATI-4 (which equates to a brief episode in the glacial MIS 2). Smart and Gooday [73] examined trends in benthonic foraminiferal abundances along an organic enrichment gradient on the continental slope off North Carolina, eastern Atlantic Ocean. They found Bulimina aculeata and Globocassidulina subglobosa to be equally abundant at all sites, suggesting that these cannot be used as proxies for the organic flux. It is unclear, however, if Globocassidulina obtusa and G. murrhina have the same tolerances.    Kaiho [74] suggested that many of the species recovered from ODP Hole 994C are indicative of suboxic bottom waters, although he also suggested that Cibicides wuellerstorfi (figure 3C) and Cibicidoides robertsonianus are indicative of oxic water [75]. The abundance of C. wuellerstorfi was relatively high during PATI-6. Altenbach et al. [72] recorded the annual organic carbon flux levels best tolerated by some live benthonic foraminifera. Most species recorded by Altenbach et al. [72] that were also recovered from ODP Hole 994C lived under a flux rate of ,2-6 g m 22 yr 21 (Bulimina striata mexicana, C. robertsonianus, Pullenia bulloides, P. quinqueloba). Schönfeld [71] recorded these four species as living infaunally within the sediment at a variety of depths down to 4.5 cm. However, Altenbach et al. [72] recorded live C. wuellerstorfi primarily at low organic carbon flux rates of 1.5-3 g m 22 yr 21 and Oridorsalis umbonatus at 2-3.5 g m 22 yr 21 . Cibicides wuellerstorfi is an epiphytal species living on raised substrate particles that prefers active bottom currents [76,77]. CoBI t showed in ODP Hole 994C that C. wuellerstorfi was recovered primarily from PATI-9, 6, 3 and 1, during which the organic carbon flux may have been low, the strength of the Western Boundary Undercurrent enhanced, or both. Altenbach et al. [72] recovered Hoeglundina elegans mainly from areas with a flux rate of 4.5-15 g m 22 yr 21 , although Schönfeld [71] suggested it to be an oxic indicator. The costate species Uvigerina mediterranea and U. peregina, which have morphologies comparable to U. hispidocostata, primarily under a carbon flux of 3-9 g m 22 yr 22 . Schönfeld [71] recorded U. peregrina as living mostly at depths of 0-1 cm below the sediment water interface, but U. mediterranea as occurring down to 6 cm below the interface. Uvigerina hispidocostata in ODP Hole 994C was recovered mainly from PATI-8 and PATI-7, which coincide with MIS 8, for which it could indicate an interval of enhanced organic carbon flux but might also highlight an interlude in which uvigerinids penetrated deeper into the sediment. Schoenfeld and Altenbach [78] found that Uvigerina spp. in the north-eastern Atlantic Ocean were more abundant during glacial MIS 2 than interglacial MIS 1, and ascribed this to a widespread change from glacial to modern productivity characteristics across termination T-I. MIS 8 is similarly a glacial stage. Seiglie [79] noted B. lowmani to be indicative of high organic carbon levels and Sen Gupta and Strickert [80]  Gooday et al. [81] found Epistominella exigua ( Figure 3D to be abundant in well-oxygenated, abyssal water below the oxygen minimum zone of the Arabian Sea. Smart et al. [82] showed that E. exigua colonises aggregates of phytodetritus and they speculated that this opportunistic, epifaunal species may represent a proxy for seasonal phytodetritus pulses originating from surface primary productivity in open ocean eutrophic areas. They suggested that inputs added over a geologically prolonged period of time would be reflected in peaks of E. exigua. This species was at its most abundant in PATI-6, having a ATI p .0.05 across the PATI-7/6 boundary and a ATI t .0.05 across both the PATI-7/ 6 and PATI-6/5 boundaries. This implies a brief interlude of enhanced seasonality in MIS 7 and 6, and may be related to a change in surface circulation and the position of the Gulf Stream at that time. Sequence stratigraphy is the correlation of sedimentary rock successions using key events produced by worldwide changes in sea level [1]. These events are used to divide the succession into packages (systems tracts) that are bounded by characteristic surfaces [83,84]. Benthonic foraminifera have long been used in sequence stratigraphy at neritic paleodepths [85]. However, the use of benthonic foraminiferal assemblage characteristics for sequence stratigraphic purposes at abyssal depths has thus far been problematic. For example, at neritic depths the planktonic/ benthonic foraminiferal ratio has been used to determine changes in sea level [86,87]. However, this index cannot be readily applied at depths of more than ,500 m, at which planktonic foraminifera typically form .99% of the assemblage. At neritic depths, maximum flooding surfaces are reflected by peaks in uvigerinid abundance that have been ascribed to sluggish circulation at times of maximum transgression [88]. This is contrary, however, to the enhanced abundance of bathyal and abyssal Uvigerina during glacial lowstands. Nagy et al. [89] suggested that at neritic depths the information function H is low on interglacial maximum flooding surfaces. In ODP Hole 994C, however, H is negatively correlated with d18O and the index is lower during glacial, even numbered MIS than it is during interglacial MIS. Therefore, we propose that ATI s peaks show strong potential as a sequence stratigraphic tool for abyssal deposits, some peaks at the PATI boundaries coinciding with terminations that are marked by transgressive systems tracts. However, the apparent coincidence between peaks in ATI s and terminations must be applied with caution, since not all peaks coincide with terminations; two peaks occurred within glacial MIS 8. However, this can be avoided by judicious use of the control limits.

Conclusions
Assemblages are not constant entities, but change over time as the proportional abundance of each species within a community changes. As one species acquires a higher proportional abundance, one or more others must decrease in abundance. Peaks in ATI s , the ATI between successive samples, delimit peak-bounded intervals (PATIs-) within which the community is relatively stable. The current inability to detect glacial-interglacial contrasts in general appears to arise because not all sites show marked changes in community composition at the species level at glacialinterglacial boundaries. Both ATI and CoBI can be applied to successions for which there is no immediately obvious differentiation of glacial and interglacial assemblages. Peaks in ATI s in the Upper Quaternary of ODP Hole 994C, Blake Ridge, define eleven PATIs. Eight of the PATI boundaries approximate to terminations, although, as shown by termination T-I, a termination can be marked by more than one PATI boundary if, like termination T-I, it consists of a series of events marked by decimeter changes in sea level. While it appears that for our data set all terminations were marked by at least one PATI boundary, not all PATI boundaries coincided with terminations; two PATI boundaries were recorded within MIS 8. Nevertheless, this suggests that PATIs and peaks in assemblage turnover as measured by our index, ATI s , have potential as a sequence stratigraphic tool. Our quantitative approach allows some sequence stratigraphic concepts to be extended into the abyssal environment.
Both CoBI p and CoBI t suggest that species that changed markedly across PATI boundaries were responding to changes in paleo-oxygenation, the organic matter flux, or bottom current strength. A transitory peak in Epistominella exigua within PATI-6 implies a brief interlude of enhanced seasonality in MIS 7 and 6, and may be related to a change in surface circulation and the position of the Gulf Stream at that time.
The assemblage turnover (ATI) and conditioned-on-boundary (CoBI) indices have here been applied to the ecostratigraphy of abyssal benthonic foraminifera. However, these measures can also be used to detect and characterise boundaries for any taxon and applied in both paleoecological and ecological studies.

Supporting Information
Appendix S1 Terms introduced in this paper and their definitions.