1. Introduction
The benthic foraminiferal and glauconite contents of Quaternary sediment on the continental rise off NE South America, NW of the mouth of the River Amazon and at latitude 0–10°N, remain little known, even though numerous piston cores (McGreary & Damuth, Reference McGreary and Damuth1973; Damuth, Reference Damuth1975; Damuth & Kumar, Reference Damuth and Kumar1975; Bé et al. Reference Bé, Damuth, Lott, Free, Clune and Hays1976) and several holes from the Deep Sea Drilling Program (DSDP) and Ocean Drilling Program (ODP) have been taken from this region. Schnitker (Reference Schnitker1980, fig. 3) speculated from samples taken northwards of c. 12°N that the continental rise off NE South America should currently be occupied by a fauna dominated by Uvigerina peregrina Cushman, with lesser Hoeglundina elegans (d’Orbigny) and Gyroidina spp. Kaiho (Reference Kaiho1994) suggested that such a fauna lives in suboxic bottom waters. Schnitker's (Reference Schnitker1980) speculation has yet to be tested. He did not indicate clearly the dominant species on the continental rise during the last glacial, but suggested that the adjacent abyssal plane was occupied by a mixed Epistominella exigua (Brady)/Oridorsalis umbonatus (Reuss) fauna. Hofker (Reference Hofker1983) included a station at 940 m water depth in his study of benthic foraminifera off Surinam, Guyana and French Guiana, but noted that, due to the coarse mesh size of the Agassiz trawl used, this surface sample contained larger specimens only. (Hofker (Reference Hofker1983) did not, however, state the precise mesh size of the trawl.) He suggested that the abundant Bulimina aculeata d’Orbigny, B. alazanensis Cushman, B. striata mexicana Cushman, Cibicidoides pseudoungerianus Cushman, Osangularia cultur (Parker and Jones), Sphaeroidina bulloides d’Orbigny and U. peregrina in the sample are indicative of upwelling of cool water in the area.
The foraminifera on the northern South American continental shelf are better known than those on the continental rise (Cushman and Parker, Reference Cushman and Parker1931; Drooger and Kaasschieter, Reference Drooger and Kaasschieter1958; Hofker, Reference Hofker1983; Mikhalevich, Reference Mikhalevich1983; Vilela, Reference Vilela2003; Wilson, Reference Wilson2006, Reference Wilson2010; Pascual et al. Reference Pascual, García, Lázaro and Pujos2009). Pascual et al. (Reference Pascual, García, Lázaro and Pujos2009) examined the foraminiferal fauna in 16 piston cores taken from the shelf (15–200 m water depth) between French Guiana and the Atlantic coast of Venezuela, which they found to include Ammonia tepida (Cushman), Hanzawaia concentrica (Cushman) and Nonionella atlantica Cushman. Such an assemblage is indicative of river influence (Adegoke, Omatsola & Salami, Reference Adegoke, Omatsola, Salami, Schafer and Pelletier1976) and presumably reflects the north-westwards transport of Amazon water as a low-salinity plume (van der Zwaan & Jorissen, Reference van der Zwaan, Jorissen, Tyson and Pearson1991; Hellweger & Gordon, Reference Hellweger and Gordon2002; Hu et al. Reference Hu, Montgomery, Schmitt and Muller-Karger2004). This plume seasonally extends the entire 1600 km between the mouth of the Amazon and the Orinoco delta (Eisma, Reference Eisma1998; Archer, Reference Archer, Blum, Marriott and Leclair2005). Pascual et al. (Reference Pascual, García, Lázaro and Pujos2009) found U. peregrina across much of the shelf. This species is typically recorded from abyssal water depths (Balsam et al. Reference Balsam, Gary, Healy-Williams and Williams1987) and is indicative of a high nutrient loading (Schnitker, Reference Schnitker1980; Schmiedl & Mackensen, Reference Schmiedl and Mackensen1997; Altenbach et al. Reference Altenbach, Pflaumann, Schiebel, Thies, Timm and Trauth1999; Murray, Reference Murray2006), Finally, Pascual et al. (Reference Pascual, García, Lázaro and Pujos2009) recorded H. elegans and Planulina ariminensis d’Orbigny, which are more usually associated with North Atlantic deep water (NADW), from the outer continental shelf off Surinam and concluded that they reflect upwelling of nutrient-rich water. Ganssen & Sarnthein (Reference Ganssen, Sarnthein, Suess and Thiede1983) had previously found Uvigerina sp. and H. elegans in association with upwelling off NW Africa. Hofker (Reference Hofker1983, p. 7) wrote that the foraminiferal shelf fauna off Surinam and Guyana ‘does not appear to be very tropical’ and is instead indicative of cooler water.
Upwelling off northern South America has been little studied. Predictive sedimentological and micropalaeontological indicators of upwelling that might be detected off South America have been well documented elsewhere however, and are outlined here. The five currently most pronounced upwelling systems globally (California, Peru, Namibia, Morocco and Somalia) are all reflected in both the micropalaeontological and sedimentological records, especially due to the formation of a high-stress oxygen minimum zone (OMZ) at depth (Levin & Sibuet, Reference Levin and Sibuet2012). An OMZ is a layer of water within which dissolved oxygen concentrations can be as low as 0.1–1.0 mL L−1 (Smart, Reference Smart and Haslett2003). OMZs develop where seasonal upwelling of oceanic water enhances the nutrient availability in the photic zone and stimulates primary productivity. In turn, this boosts the rain of organic matter into deeper water, where oxidation of the descending organic matter depletes dissolved oxygen levels.
The sedimentological record differs between the five most pronounced, modern upwelling systems (Vercoutere et al. Reference Vercoutere, Mullins, McDougal and Thompson1987). Burnett, Roe & Piper (Reference Burnett, Roe, Piper, Suess and Thiede1983) and Parrish (Reference Parrish1998) suggest that sediment below upwelling centres is characterized by chert, organic-rich shale and phosphorites (see also Levin et al. Reference Levin, Gutierrez, Rathburn, Neira, Sellanes, Munoz, Gallardo and Salamanca2002). In contrast, Fischer & Arthur (Reference Fischer, Arthur, Cook and Enos1977) suggested that sediment bathed in water with a low dissolved oxygen content but moderately high dissolved silica and iron levels contains copious glauconite. Mullins et al. (Reference Mullins, Thompson, McDougall and Vercoutere1985) and Vercoutere et al. (Reference Vercoutere, Mullins, McDougal and Thompson1987) found abundant glauconite along the upper edge of the OMZ off California. Burnett (Reference Burnett1980) suggested that the formation of glauconite, phosphorite and apatite off Peru was enhanced during sea-level highstands during late Quaternary time. Off northern South America, Hofker (Reference Hofker1983) recorded glauconite in scattered samples below 33.5 m water depth off Surinam and western Guyana, while Odin, Mackinnon & Pujos (Reference Odin, Mackinnon, Pujos and Odin1988) recorded it from the continental shelf off French Guiana in water of depth >150 m.
The OMZ foraminiferal fauna has been most exhaustively studied off California. There the glauconite is associated with a benthic foraminiferal fauna dominated by Cassidulina (up to 90% of total benthic foraminifera; Mullins et al. Reference Mullins, Thompson, McDougall and Vercoutere1985). Vercoutere et al. (Reference Vercoutere, Mullins, McDougal and Thompson1987) recorded that Uvigerina spp. form up to c. 30% of the benthic foraminiferal community near the boundaries of the California OMZ, and a pronounced minimum near the OMZ core. Examination of figure 3 in Mullins et al. (Reference Mullins, Thompson, McDougall and Vercoutere1985) suggests that the percentage abundances of Cassidulina spp. and Uvigerina spp. along the upper edge of the OMZ are inversely correlated, the peak abundance of Cassidulina spp. occurring shallower that Uvigerina spp. Within the OMZ, grain size and the abundance of glauconite, carbonate and foraminiferal fragments all decrease towards the OMZ core such that Uvigerina spp. off California are not closely associated with glauconite. Vercoutere et al. (Reference Vercoutere, Mullins, McDougal and Thompson1987) suggested that the number of benthic foraminiferal genera off California is higher at the OMZ core, where the fauna is dominated by Bolivina and Brizalina, than along the margins. Off California the fauna at the OMZ lower boundary is dominated by Globobulimina and Praeglobobulimina and glauconite disappears.
This paper examines whether upwelling, noted on the upper continental slope and outer continental shelf of NE South America, also impacted the Demerara Rise. Using benthic foraminifera in the upper Quaternary lower bathyal ODP Hole 1261A, located on the Demerara Rise (water depth 1899 m, Fig. 1), we report that the foraminiferal fauna and glauconite content show that the upper Quaternary sediment in ODP Hole 1261A was deposited along the upper margin of an OMZ induced by upwelling, but that the position of ODP Site 1261 relative to the OMZ margin varied over time. Possible reasons for that variation are presented.
Location map, ODP Site 1261.
2. Site description
ODP Hole 1261A was drilled c. 350 km NE of Surinam on the gently dipping (c. 1°) NW slope of Demerara Rise. The Demerara Rise is a submarine plateau located at c. 5°N off the coasts of Surinam and French Guiana. It extends c. 380 km along the coast and is c. 220 km wide from the shelf edge to an escarpment along its NE edge, at which water depths increase quickly from 1000 to >4500 m. The Demerara Plateau comprises continental crust and is a bathymetric extension of the continental shelf offshore Surinam and French Guiana (Basile et al. Reference Basile, Maillard, Patriat, Gaullier, Loncke, Roest, Mercier de Lépinay and Pattier2013) that is conjugate with the Guinea Plateau offshore West Africa (Aslanian et al. Reference Aslanian, Moulin, Olivet, Unternehr, Matias, Bache, Rabineau, Nouzé, Klingelheofer, Contrucci and Labails2009) and has a complex rifting and post-rift history (Heine & Brune, Reference Heine and Brune2014). This history includes middle Cainozoic seawards tilting that Basile et al. (Reference Basile, Maillard, Patriat, Gaullier, Loncke, Roest, Mercier de Lépinay and Pattier2013) suggested may have been driven by either:
-
1. Mantle dynamics, although they could not find a clear reason as to why this mechanism would be restricted to the Demerara Plateau and not the adjacent continental margins and why it would have occurred over such a short time interval.
-
2. Subsidence coeval with the late Miocene uplift of the Eastern Cordillera of Colombia and the Venezuelan Andes, which changed the course of the Orinoco from northwards flowing into the Caribbean Sea to eastwards flowing into the Central Atlantic Ocean (Díaz de Gamero, Reference Díaz de Gamero1996) and connected the Amazon to the Atlantic (Hoorn et al. Reference Hoorn, Guerrero, Sarmiento and Lorente1995). These diversions led to the development of the Amazon deep-sea fan to the SE of the Demerara Plateau (Manley & Flood, Reference Manley and Flood1988) and the Orinoco deep-sea fan to the NW (Callec et al. Reference Callec, Deville, Desaubliaux, Griboulard, Huyghe, Mascle, Mascle, Noble, Padron de Carillo and Schmitz2010). These large sediment accumulations loaded the oceanic lithosphere, inducing ocean-wards flexing of the Demerara Plateau and adjacent continental margin.
Similar ocean-wards tilt has not been recorded on the conjugate Guinea Plateau, where flanking deep-sea fans are absent. The tilting hypothesis presented by Basile et al. (Reference Basile, Maillard, Patriat, Gaullier, Loncke, Roest, Mercier de Lépinay and Pattier2013) is problematic in that they suggest that the majority of the tilting and subsidence occurred during late Oligocene and early Miocene time, whereas the uplift of the Eastern Cordillera and Venezuelan Andes occurred during late Miocene time. The Amazon deep-sea fan began developing during middle–late Miocene time, and reached its present shape and size during late Pliocene time (Figueiredo et al. Reference Figueiredo, Hoorn, van der Ven and Soares2009). The first indication of the Orinoco River in Trinidad is in the upper Miocene – lower Pliocene San José Calcareous Silt Member of the Manzanilla Formation (Wilson, Reference Wilson2013). Manley & Flood (Reference Manley and Flood1988) however suggested that the Amazon fan, growth of which helped induce the tilting, was deposited in two or three sedimentation cycles that span later Neogene time. Callec et al. (Reference Callec, Deville, Desaubliaux, Griboulard, Huyghe, Mascle, Mascle, Noble, Padron de Carillo and Schmitz2010) found that turbidite deposition on the Orinoco fan persisted throughout the last glacial interval but slowed during the post-glacial sea-level rise. This suggests that Pleistocene and ongoing subsidence of the Demerara Rise due to crustal loading is possible.
Although ODP Hole 1261A lies c. 800 km NW of the mouth of the River Amazon, the world's largest river with a discharge of (1.93 ± 0.13)×105 m3 s−1 (Perry, Duffy & Miller, Reference Perry, Duffy and Miller1996), the Shipboard Scientific Party (Reference Erbacher, Mosher and Malone2004) found little evidence of riverine influence in the sediment on the Demerara Rise, where the upper Quaternary sediment of ODP Hole 1261A is a pelagic nannofossil ooze composed of foraminifera, calcareous nannofossils, carbonate debris and diagenetic calcite. This reflects the observation by Damuth & Kumar (Reference Damuth and Kumar1975) that, although much Amazonian sediment is transferred into deep water as turbidity currents (Pirmez & Imran, Reference Pirmez and Imran2003) and regional mass-transport complexes (Maslin et al. Reference Maslin, Mikkelsen, Vilela and Haq1998), there are no contour currents to transport sediment northwest from the fan to the Demerara Rise. Nor is there much clay within the upper Quaternary sediment of ODP Hole 1261A. This is because the large quantity of suspended sediment disgorged by the Amazon (Milliman & Meade, Reference Milliman and Meade1983; Meade, Reference Meade1994) is generally not washed far offshore but is instead transported northwest along the continental shelf as mudcapes driven by the Guyana Current (Lentz, Reference Lentz1995; Eisma, Reference Eisma1998), such that about half the mud forming the Orinoco Delta is of Amazon origin (Aslan et al. Reference Aslan, White, Warne and Guevara2003). In addition, the Shipboard Scientific Party (Reference Erbacher, Mosher and Malone2004) recorded neither turbidites nor extensive bioturbation within the upper Quaternary sediment of ODP Hole 1261A. They interpreted a sharp colour change from pale yellow/pale brown to olive grey/greenish grey at c. 0.25 m below the seafloor (mbsf) to be a diagenetic redox boundary that has propagated down from the seafloor. Below this colour change, the upper Quaternary sediment comprises light greenish-grey sediment with only subtle variations. This suggests that the upper Quaternary layer in ODP Hole 1261A contains mostly in situ sediment and benthic foraminifera that will provide an excellent record of Late Pleistocene palaeoenvironments off NE South America.
The Shipboard Scientific Party (Reference Erbacher, Mosher and Malone2004) examined the calcareous nannoplankton in two core catcher samples from the upper Quaternary sediment of ODP Hole 1261A. They found that a sample from 12–17 cm within the core catcher for Core 1 (3.83–3.90 mbsf) comprised upper Quaternary sediment with abundant Emiliania huxleyi (Lohmann), which is indicative of Zone NN21, the first appearance of which occurred during 0.262–0.264 Ma. The next core catcher at the base of Core 2 (4.61 mbsf) contained Pseudoemiliania lacunosa (Kamptner), which is indicative of Zone NN19 and which became extinct at c. 485 ka (Thierstein et al. Reference Thierstein, Geitzenauer, Molfino and Shackleton1977). Note that, these being core catcher samples, they probably do not reflect the precise positions of the species’ evolutionary appearance (E. huxleyi) or extinction (P. lacunosa).
3. Materials and methods
Forty-six samples of 20 cm3 each were taken from the uppermost two cores from ODP Hole 1261A, 0–4.7 m below the seafloor (mbsf) every 10 cm, although sample spacing was disrupted through the lack of recovery of a sample at 2.9 mbsf.
In total, 13690 benthic foraminifera were recovered from the samples (mean 298 specimens per sample). Specimens of Siphonodosaria lepidula (Schwager), the extinction of which is a marker for the basal Middle Pleistocene layer (Gupta, Reference Gupta1993; Lutze, Reference Lutze, von and Ryan1979) at 570 ka (Hayward et al. Reference Hayward, Kawagata, Sabaa, Grenfell, Kerckhoven, Johnson and Thomas2012), were used to determine the probable base of the Middle Pleistocene layer. A specimen of S. lepidula recovered from 2.3 mbsf, above the lowest recorded occurrence of E. huxleyi, is concluded to be reworked. A further eight specimens of S. lepidula were recovered from 4.5 mbsf, just above a recorded occurrence of P. lacunosa at 4.61 mbsf. Since the last appearance datum of P. lacunosa (which became extinct at c. 485 ka) is probably not recorded, it is possible that the S. lepidula at 4.5 mbsf are in situ and indicate an age of c. 570 ka at 4.5 mbsf. If so, then they indicate a mean sedimentation rate of c. 0.8 cm ka−1 for the upper Quaternary section. This is far lower than the rate of deposition on the Orinoco fan, where rates of 13–42 cm ka−1 have been recorded (Hoffmann et al. Reference Hoffmann, Bahr, Voigt, Schönfeld, Nürnberg and Rethemeyer2014).
This study comprises two modules. In the first, 200 grains of dimension >105 μm were picked from each sample, a foraminiferal fragment being picked and counted if it retained the proloculus. These 200 grains were used to compute: (1) the percentage of grains as glauconite (%G); and (2) the percentage of all foraminifera as benthic specimens (%B = [B/(B+P)]×100, where B and P are the numbers of benthic and planktonic specimens respectively). The two lowest samples contained %G > 70%, which would have made the calculation of %B unreliable using only 200 grains. For these two samples, a total of 200 foraminifera were picked and used to determine %B. All percentage data were transformed as log plus unity when subjected to statistical testing. The runs test was used to determine if %G and %B fluctuated in a random manner.
For the second part of the study, c. 300 specimens of benthic foraminifera were picked from the >105 μm fraction of each sample. These were sorted into species and identified using illustrations in Phleger & Parker (Reference Phleger and Parker1951), Phleger, Parker & Peirson (Reference Phleger, Parker and Peirson1953), van Morkhoven, Berggren & Edwards (Reference van Morkhoven, Berggren and Edwards1986), Bornmalm (Reference Bornmalm1997), Mohan, Gupta & Bhaumik (Reference Mohan, Gupta and Bhaumik2011) and Holbourn, Henderson & MacLeod (Reference Holbourn, Henderson and MacLeod2013). The percentage abundance of each species in each sample was calculated and fluctuations in these were compared with %G and %B using the transformation ln(x+1) for all measures and Pearson's correlation coefficients. Trends in the percentage abundances of U. peregrina and C. laevigata within the examined section were similarly analysed using the log plus unity transformed percentage abundances to avoid attenuation problems. All such data are hereafter referred to as transformed percentage abundances.
The information function H was used to assess diversity, calculated as
\begin{equation}
H = - \sum {p_i \,\ln \left( {p_i } \right)} ,\end{equation}
where pi is the proportional abundance of the ith species, i = 1, 2, 3,. . . S, where S is species richness. Dominance was determined using the max(pi ) index of Berger & Parker (Reference Berger and Parker1970). For details of both, see Hayek & Buzas (Reference Hayek and Buzas2013).
The benthic foraminiferal index (BFOI) of Kaiho (Reference Kaiho1994) was calculated using those species or genera included in Kaiho's list. This measure uses only the relative abundance of oxic and dysoxic foraminifera in calculating BFOI, so the proportional abundances of suboxic and species not ascribed to either the oxic, suboxic or dysoxic groups were also calculated.
Hayek & Wilson (Reference Hayek and Wilson2013) developed an assemblage turnover index (ATI), which they applied to the upper Quaternary section of ODP Hole 994C (Blake Ridge) (see also Wilson & Hayek Reference Wilson and Hayek2014). For a set of samples from a stratigraphic section, the ATI between each pair of adjacent samples is calculated from
\begin{equation}
{\rm ATI}_{\rm s} = \Sigma \left| {p_{i2} - p_{i1} } \right|\end{equation}
in which ATIs is the between-sample assemblage turnover and pi
1 and pi
2 are the proportional abundances of the ith species. ATIs therefore yields the proportion or percent of faunal turnover or change between samples. The mean (
$\bar x$
) and standard deviation (σ) of values of ATIs were calculated over all 46 samples from ODP Hole 1261A. A normal quantile plot test for non-normality and a critical level of p = 0.05 were used to determine that the values of ATIs were normally distributed, as expected of a statistic that is a sum. A runs test was performed to determine whether values of ATIs were randomly distributed around
$\bar x$
. Using control charts, all points with a control limit of ATIs > (
$\bar x$
+ σ) were deemed to be positions of elevated taxonomic turnover against background rates of variability.
Values of ATIs > (
$\bar x$
+ σ) were used to delimit peak-bounded ATI intervals (PATIs), which were numbered from the top of the section down (see Hayek & Wilson, Reference Hayek and Wilson2013 for details). The conditioned-on-boundary index (CoBI) of Hayek & Wilson (Reference Hayek and Wilson2013) was used to assess which species contributed most to the total turnover, ATI, at the PATI boundaries. CoBI therefore furnishes the proportion that each species within an assemblage contributes to the total assemblage turnover across the PATI boundary. For each species at any PATI boundary,
\begin{equation}
{\rm CoBI} = \frac{{\left| {p_{i2} - p_{i1} } \right|}}{{{\rm ATI}}},\end{equation}
where pij, j = 1,2, are the ith species proportions on either side of the selected boundary of interest and at which the ATI is calculated. This study uses both the partial conditioned-on-boundary index (CoBIp) and the thorough conditioned-on-boundary index (CoBIt) of Hayek & Wilson (Reference Hayek and Wilson2013). For the CoBIp, the assemblage turnover index ATI is calculated between the entire set of samples within the PATI, below the PATI boundary and the first sample immediately above it. In this case, the ATI is designated as ATIp. The value of ATI ( = ATIp) is substituted into Equation (3), as are pi 1, the proportional abundance of the ith species in the entire PATI below the peak in ATIp, and pi 2, the proportional abundance of that ith species in the first sample above the peak in ATIs. The proportional contribution of each species to ATIp is assessed from the vector of CoBIp values at each ATIs peak.
For CoBIt, the calculation of ATI ( = ATIt) uses all samples in the two PATIs separated by the peak in ATIs and the value of ATI ( = ATIt) is substituted into Equation (3), as are pi 1 and pi 2.
4. Results
The mean value of the percentage of 200 grains as glauconite (%G) was 9.1% (median 3.25%, range 0–89%, s.d. 17.1%). The runs test (z = 2.98, p = 0.003) shows that the data for %G were not generated by a random process at the 0.05 significance level. Values of %G below the median appear to be concentrated in the three intervals 0.8–1.2, 2.0–3.1 and 3.9–4.2 mbsf (Fig. 2). Meanwhile, the mean percentage of the foraminiferal fauna as benthic specimens (%B) was 9.9% (median 8.7%, range 1.5–31.8%, s.d. = 6.7%). The runs test showed that values of %B were not randomly distributed around the mean (z = 2.17, p = 0.03). Values of %B lower than the median are concentrated between 0.0–0.4 and 2.5–4.2 mbsf. The %G and %B are significantly correlated (r = 0.35, p = 0.019).
The 200 grain study and assemblage turnover index for the upper Quaternary section of ODP Hole 1261A: (a) percentage of grains as glauconite (%G); (b) percentage of grains as benthic foraminifera (%B); and (c) between-sample assemblage turnover index (ATIs ). Vertical dashed line – mean ATIs plus one standard deviation; horizontal bashed lines – peaks in ATIs (PATI boundaries). PATIs numbered from top downwards.
The 13690 specimens recovered were placed in 134 species or species groups, of which 13 each formed >1% of the total recovery (Bulimina aculeata, B. striata mexicana, Cassidulina laevigata, Epistominella exigua, Fissurina marginata (Montagu), Globocassidulina subglobosa (Brady), Gyroidinoides lamarckiana (d’Orbigny), Hoeglundina elegans, Oridorsalis umbonatus (Reuss), Planulina wuellerstorfi (Schwager), Sigmoilopsis schlumbergeri (Silvestri), Uvigerina auberiana d’Orbigny, U. peregrine; see online Supplementary Material 1 at http://journals.cambridge.org/geo). The distributions of selected species are shown in Figure 3. Total recovery comprised mostly B. aculeata (38%), with lesser U. peregrina (23%), S. schlumbergeri (4.6%) and C. laevigata (4.4%). These four species collectively formed 44.6–91.7% of each sample (mean 71.6%; s.d. 14.2%). The combined Uvigerina spp. had a mean of 25% of each sample (range 4.7–46.5%, s.d. 11.1%). As shown by max(pi ) however, individual samples were mostly dominated by either B. aculeata or U. peregrina. One sample at 1.8 mbsf was dominated by E. exigua. Brizalina spp., which dominate at the cores of OMZs, formed <0.2% of the total recovery.
Percentage abundances of selected species in the upper Quaternary ODP section of Hole 1261A: (a) Bulimina aculeata; (b) Cassidulina laevigata; (c) Epistominella exigua; (d) Gyroidinoides lamarckiana; (e) Oridorsalis umbonatus; (f) Planulina wuellerstorfi; (g) Sigmoilopsis schlumbergeri; and (h) Uvigerina peregrina. Horizontal dashed lines – PATI boundaries; further dashed lines in B and H – linear regressions between percentage abundances and depth below seafloor.
Relationships between the transformed percentage abundances of the 13 most-abundant species and measures from the 200 grain module were investigated (see online Supplementary Material 2 at http://journals.cambridge.org/geo). Only the transformed percentage abundance per sample of E. exigua was positively correlated with transformed %G at a significance level of <0.05 (r = 0.457). Transformed percentage abundances of B. aculeata and U. peregrina were positively and significantly correlated with transformed %B at r = 0.49 and 0.44 respectively, while B. striata mexicana, C. laevigata, G. subglobosa, G. lamarckiana, H. elegans, O. umbonatus, P. wuellerstorfi, S. schlumbergeri and U. auberiana were negatively correlated with transformed %B. The transformed percentage abundance of total Uvigerina spp. was positively correlated with transformed %B (r = 0.37, p = 0.01), as was the transformed percentage abundance of the benthic foraminiferal fauna per sample as U. auberiana (r = 0.36, p = 0.01).
Online supplementary Material 2 (available at http://journals.cambridge.org/geo) shows that C. laevigata and U. peregrina were negatively correlated (r = –0.49, p = 0.001; Fig. 3). Linear regression shows the transformed percentage abundance of U. peregrina to be higher towards the top of the section (ln(%Up) = 3.458 – 0.171D, r = 0.47, p = 0.001, where %Up is the percentage abundance of U. peregrina and D is the depth of the sample in metres below the seafloor). Cassidulina laevigata has lower percentage abundances towards the top of the succession (ln(%Cl+1) = 0.402+0.337D, r = 0.46, p = 0.001, where %Cl is the percentage abundance of C. laevigata). However, the section at 0.0–0.7 mbsf has relatively high proportional abundances of C. laevigata.
Of the total recovery, 0.4% of specimens were in the dysoxic group, 5.5% oxic and 81.6% suboxic. The remaining 12.2% of specimens were not assigned to any one particular group. The proportion of suboxic species (which principally comprised Bulimina spp., Uvigerina spp. and Cassidulina spp.) was generally higher at 2.4–0.3 mbsf than in the remainder of the section. The BFOI ranged from 50 to 100 (mean = 90.42). The graph of BFOI (Fig. 4) falls into three subsections: (a) 4.5–2.3 mbsf within which, as indicated by the coefficient of variation (CV), the mean BFOI is generally high and the BFOI varies little (mean = 95.3, sd = 5.5, CV = 5.8); (b) 2.2–0.5 mbsf, within which the mean is lower and the BFOI more variable (mean = 82.8, sd = 15.3, CV = 18.5), indicating an interval with depressed dissolved oxygen levels; and (c) 0.4–0.0 mbsf, within which the mean is high and the BFOI varies little (mean = 95.5, sd = 2.6, CV = 2.7).
Measures in the upper Quaternary section of ODP Hole 1261A. (a) Proportional abundance of the most abundant species in any sample, max(pi ). White – Bulimina aculeata dominant; light grey – Uvigerina peregrina dominant; dark grey – Epistominella exigua dominant. (b) Information function, H. (c) Benthic foraminifera oxygenation index, BFOI. (d) Proportion of assemblages as suboxic specimens.
The information function H ranged between 1.16 and 2.91 (Fig. 4). It was generally higher (mean H = 2.34) at 4.5–2.5 mbsf than at 2.4–0.0 mbsf (mean H = 1.84; observed value of student's t-test t obs = 3.68, critical value t crit = 2.01, p = 0.001, degrees of freedom df = 44).
The between-sample assemblage turnover index (ATIs) had a mean of
$\bar x$
= 0.58 (range 0.25–1.10, sd = 0.21; Fig. 2). The runs test showed that values of ATIs are not randomly distributed around the mean (z = 2.08, p = 0.04). Eight values of ATIs > (
$\bar x$
+ σ) delimited the PATI boundaries. We also chose to place a PATI boundary between 1.7 m and 1.6 mbsf since the value of ATIs, while less, nevertheless rounded to the control limit of (
$\bar x$
+ σ). This boundary is within a group of four relatively high values of ATIs. Nine peaks in ATIs were therefore recognized, dividing the section into ten peak-bounded ATI intervals (PATIs). The number of samples per PATI ranged from one (PATIs-2, 3, 10) to nine (PATI-4). Clearly, the uppermost PATI-1 and the lowermost PATI-10 were of necessity incomplete, their top and bottom boundaries respectively not being bounded by a peak in ATIs. That the one-sample PATIs 2 and 3 (0.7–0.5 mbsf) are adjacent indicates rapid assemblage turnover towards the top of the section. Some PATI boundaries coincide with marked changes in max(pi
), such as those between PATI-5 through PATI-8. However, others do not such as the PATI-8–9 and PATI-2–3 boundaries. Likewise, while some PATI boundaries coincide with marked changes in H, others do not.
The partial assemblage turnover index (ATIp) ranged between 0.47 and 1.0 (
$\bar x$
= 0.80, sd = 0.16), while the thorough assemblage turnover index (ATIt) ranged between 0.34 and 1.0 (
$\bar x$
= 0.65, sd = 0.25; Table 1). Values of ATIt decreased gradually from the PATI-9–10 boundary to the PATI-4–5 boundary. ATIt exceeded (
$\bar x$
+ σ) only across the PATI-3–4 and PATI-2–3 boundaries, indicating these to be times of especially marked faunal turnover.
Partial assemblage turnover indices (ATIp) and partial conditioned-on-boundary indices (CoBIp) in the Upper Quaternary in ODP Hole 1261A. Species included each show a CoBIp > 0.02 across any one PATI boundary. Values of CoBIp > 0.1 in bold. Italics indicate a decrease in proportional abundance across a PATI boundary.
Twenty-four species had a partial conditioned-on-boundary index (CoBIp) > 0.02 across any one PATI boundary (i.e. they contributed >2% to ATIp across that boundary; Table 1). However, 19 of these species had a CoBIp > 0.02 across ≤4 PATI boundaries. Of the five remaining species with CoBIp > 0.02 across ≥5 PATI boundaries, Planulina wuellerstorfi and Sigmoilopsis schlumbergeri had CoBIp > 0.02 across PATI boundaries low in the section only, while C. laevigata, B. aculeata and U. peregrina had CoBIp > 0.02 almost throughout the entire section. Values of CoBIp for U. peregrina and C. laevigata are significantly positively correlated (r = 0.73, p = 0.03), reflecting the fact that the percentage abundances of these species are negatively correlated. The major change in abundance of both species occurred near the PATI-5–6 boundary. Likewise, the CoBIp for P. wuellerstorfi and B. aculeata are correlated (r = 0.72, p = 0.03) and their percentage abundances are correlated (r = –0.59, p = 0.001). The percentage abundances of P. wuellerstorfi and C. laevigata are not correlated (r = 0.27, p = 0.07), while P. wuellerstorfi and U. peregrina are negatively correlated –0.46, p = 0.002). Only B. aculeata, C. laevigata, Epistominella exigua and U. peregrina had an especially large CoBIp > 0.1 across any one PATI boundary.
Thirty species had a thorough conditioned-on-boundary index (CoBIt) > 0.02 across any single PATI boundary (i.e. they contributed >2% to ATIt across that boundary; Table 2). Of the 24 species that had CoBIp > 0.02 across any boundary, all but Quinqueloculina ex gr. lamarckiana had a CoBIt > 0.02. Only B. aculeata, B. striata mexicana, C. laevigata, Epistominella exigua, Hoeglundina elegans, Oridorsalis umbonatus and U. peregrina had a CoBIt > 0.02 across ≥5 PATI boundaries, while only B. aculeata, C. laevigata, E. exigua, S. schlumbergeri and U. peregrina had a CoBIt > 0.1 over any one PATI boundary. E. exigua was most abundant in PATI-3 and in the uppermost parts of PATI-5 and PATI-10.
Thorough assemblage turnover indices (ATIt) and thorough conditioned-on-boundary indices (CoBIt) in the upper Quaternary section of ODP Hole 1261A. Species included each show a CoBIt > 0.02 across any one PATI boundary. Values of CoBIt >0.1 in bold. Italics indicate a decrease in proportional abundance across a PATI boundary.
5. Discussion
Evidence from both the 200 grain module and the benthic foraminiferal fauna module shows that the upper Quaternary section at ODP Site 1261 was deposited under the influence of an oxygen minimum zone (OMZ). Furthermore, the proximity of the site to the upper margin of the OMZ varied over time, the OMZ upper margin shallowing relative to the seafloor in the upper part of the section.
The percentage of the total (planktonic+benthic) foraminiferal fauna as planktonic specimens (%P) has been widely used as a proxy for palaeodepth since its introduction by Grimsdale & van Morkhoven (Reference Grimsdale and van Morkhoven1955) (e.g. Bandy & Arnal, Reference Bandy and Arnal1960; van der Zwaan, Jorissen & de Stigter, Reference van der Zwaan, Jorissen and de Stigter1990; de Rijk, Troelstra & Rohling, Reference de Rijk, Troelstra and Rohling1999; van Hinsbergen et al. Reference van Hinsbergen, Kouwenhoven and van der Zwaan2005). The percentage of the fauna as benthic specimens (%B), being the converse of %P, should similarly have potential as a palaeodepth proxy, being higher in shallower water. However, in showing that planktonic foraminifera are more susceptible to dissolution and fragmentation than benthic foraminifera, Nguyen & Speijer (Reference Nguyen and Speijer2014) cast doubt on the value of %P and %B as proxies for palaeodepth at sites prone to dissolution at bathyal depths and deeper. Mullins et al. (Reference Mullins, Thompson, McDougall and Vercoutere1985) found that foraminiferal fragments, which would indicate enhanced dissolution, are more abundant along the upper edge of the California OMZ than at the underlying OMZ core. %B should therefore have potential as a dissolution indicator in deeper-water sections around an OMZ (although it is possible that oxygen availability and the organic carbon flux to the seafloor played a role; see van der Zwaan et al. Reference van der Zwaan, Duijnstee, den Dulk, Ernst, Jannink and Kouwenhoven1999). The concentration in ODP Hole 1261A of values of %B exceeding the median of 0.5–2.4 mbsf perhaps indicates an interval of enhanced dissolution associated with relative shallowing of the upper margin of an OMZ. The information function H was generally lower at 2.4–0.0 mbsf than at 4.5–2.5 mbsf. This also reflects the relative upwards translation of the upper boundary of the OMZ, and Levin & Sibuet (Reference Levin and Sibuet2012) found that diversity is reduced within an OMZ core compared to the overlying water.
The composition of the benthonic fauna in our study reflects this low-frequency fluctuation in the position of the upper margin of an OMZ, but not necessarily that benthic foraminifera were concentrated by dissolution alone. Nguyen & Speijer (Reference Nguyen and Speijer2014) suggested that thick-walled benthic foraminifera (Lenticulina, Uvigerina, Sigmoilina, Cibicidoides, Osangularia and Eponides) are proportionally more abundant in areas of high dissolution, while Jones (Reference Jones2014) noted that members of the Buliminida (which includes Uvigerina) are more abundant within an OMZ. In the upper Quaternary section of ODP Hole 1261A, there are significant correlations between Bulimina aculeata, U. peregrina, total Uvigerina spp. and %B. This suggests that Jones’ (2014) hypothesis that %B is enhanced in OMZs because of the exclusion of macro-predators, perhaps coupled with dissolution, is a more likely explanation for the higher values of %B between 0.5 and 2.4 mbsf than dissolution alone. This hypothesis is supported by the positive correlation between %B and %G, glauconite being associated with OMZ margins (Parrish, Reference Parrish1998).
The Bulimina–Uvigerina–Cassidulina fauna in the upper Quaternary section of ODP Hole 1261A differs markedly from the Uvigerina–Hoeglundina–Gyroidina assemblage postulated by Schnitker (Reference Schnitker1980). Hoeglundina elegans formed only 1.6% and Gyroidinoides spp. only 2.1% of the total benthic foraminiferal assemblage recovered. Although G. lamarckiana contributed >0.02 to both CoBIp and CoBIt across several PATI boundaries (Tables 1 and 2), the contributions were either equal to or less than the mean of all species across all PATI boundaries. Hoeglundina elegans contributed more than the mean CoBIp across the PATI-8–9 and PATI-4–5 boundaries (Table 2), but contributed more than the mean CoBIt across only PATI-7–8 boundary. Thus, the proportional abundances of H. elegans and Gyroidinoides spp. fluctuated little between PATIs. The Bulimina–Uvigerina–Cassidulina fauna presumably reflects the influence of the upwelling-induced OMZ off northern South America.
This OMZ is suggested to be of the Californian type, where C. laevigata and U. peregrina are abundant near the upper margin of the OMZ, but where U. peregrina occurs in deeper water than C. laevigata. The negative correlation in ODP Hole 1261A between the percentage abundances of C. laevigata and U. peregrina, coupled with linear regressions showing that the proportional abundance of U. peregrina is higher towards the top of the section while that for C. laevigata was higher towards the section base, supports the suggestion that either ODP Site 1261A subsided across the upper boundary of an OMZ during late Quaternary time, or that the upper boundary of the OMZ migrated upwards into shallower water at that time. The major changes in the percentage abundances of C. laevigata and U. peregrina species occurred near the PATI-5–6 boundary (Fig. 3), suggesting that the relative upwards movement of the OMZ upper margin, whether due to subsidence or shallowing of the OMZ, occurred abruptly. Crustal subsidence, related to the growth of the Amazon and Orinoco deep-sea fans (Basile et al. Reference Basile, Maillard, Patriat, Gaullier, Loncke, Roest, Mercier de Lépinay and Pattier2013), is thought to have been a more likely cause than shallowing of the OMZ upper margin; we cannot think of a mechanism, such as enhanced upwelling and a consequent increase in the rain of organic matter, that could have caused shallowing of the OMZ upper boundary and persisted throughout much of the late Quaternary period and probably across several glacial–interglacial cycles.
Although Basile et al. (Reference Basile, Maillard, Patriat, Gaullier, Loncke, Roest, Mercier de Lépinay and Pattier2013) stated that most subsidence occurred during late Oligocene – early Miocene time, sedimentation on the fans persists and may be inducing subsidence at a small scale. However, this subsidence hypothesis is itself not without problems; it would have required that the upper margin of the OMZ remain at a constant position relative to the seafloor across several glacial–interglacial cycles. It is difficult to envisage that this was so, since factors such as sea level, ocean current strength, wind strength and productivity varied between glacials and interglacials. The precise amount of crustal subsidence or shallowing of the upper margin of the OMZ is unknown, but comparison with Mullins et al. (Reference Mullins, Thompson, McDougall and Vercoutere1985) suggests that it was of the order at least several tens of metres. The increase in the percentage abundance of C. laevigata towards the top of the section (0.0–0.7 mbsf) suggests that the OMZ upper edge moved back down into deeper water during the latest Pleistocene and Holocene. The significant correlation between CoBIp for P. wuellerstorfi and B. aculeata provides supporting palaeoenvironmental information for the hypothesis.
Kaiho (Reference Kaiho1994) suggested that B. aculeata is indicative of suboxic bottom waters, while the epifaunal P. wuellerstorfi is a suspension feeder indicative of oxic waters and enhanced current action (see also Schoenfeld, Reference Schoenfeld2001). This suggests that intervals with higher levels of P. wuellerstorfi are somewhat less dysoxic than those with few specimens. This species is more abundant towards the base of the section (4.5–2.6 mbsf) and the top (0.2–0.0 mbsf) than in the intervening section. The latest Pleistocene and Holocene migration of the OMZ upper edge back down into deeper water may have arisen through either the rate of ongoing sedimentation now exceeding the rate of crustal subsidence, or some other factor that depressed the OMZ relative to a stable seafloor. Epistominella exigua, which was most abundant in PATI-3 and the uppermost parts of PATI-5 and PATI-10, indicates periodic enhancement of spring plankton blooms producing pulses of phytodetritus (Thomas et al. Reference Thomas, Booth, Maslin and Shackleton1995). Hayek & Wilson (Reference Hayek and Wilson2013) found a transitory peak in E. exigua in marine isotope stage (MIS) 7 and 6 at ODP Hole 994C (Blake Ridge), from which they inferred a brief interlude of enhanced seasonality that they thought may be related to a change in surface circulation and the position of the Gulf Stream at that time.
The sedimentation rate at ODP Site 1261A, based on the occurrence of supposedly in situ Siphonodosaria lepidula at 4.50 mbsl, is c. 0.8 cm ka−1. Despite the proximity of the Amazon estuary, this is comparable with that of 0.14–3.2 cm ka−1 recorded elsewhere in the deep Atlantic Ocean (Carvalho, Oliveira & Soares, Reference Carvalho, Oliveira and Soares2011). Extrapolating from this inferred depositional rate suggests that the interval of greatest assemblage turnover (PATI-2 and -3, 0.7–0.5 mbsf) may have occurred between 6.25 and 8.75 ka. However, as was noted for upper Quaternary sediment in the nearby eastern Caribbean Sea (Reid, Carey & Ross, Reference Reid, Carey and Ross1996), it is unlikely that sedimentation rates were constant throughout late Quaternary time. This age estimate is therefore only approximate.
The pattern of between-sample, non-randomly distributed ATI ( = ATIs), indicates that higher-frequency signals are superimposed on the lower-frequency variation in the relative position of the upper margin of the OMZ, there being eight peaks at which ATIs >(
$\bar x$
+ σ) and one at which ATIs rounds to (
$\bar x$
+ σ). These peaks separate intervals (PATIs) within which the background rate assemblage turnover is relatively low. The mean value of ATIs in the Demerara Rise ODP Hole 1261A (
$\bar x$
= 0.58) was significantly lower than that in the upper Quaternary section in the Blake Ridge ODP Hole 994C (
$\bar x$
= 0.71, t
obs = 2.94, t
crit = 1.98, df = 109, p = 0.004), even though both were taken from continental rises. This indicates that mean ATIs for the upper Quaternary section of the North Atlantic Ocean varies geographically. ATIs seems to have no underlying predictive range given similar glacial–interglacial perturbations to the assemblages. The determination of areas with higher and lower mean ATIs may therefore prove to be of environmental significance.
Hayek & Wilson (Reference Hayek and Wilson2013) hypothesized that most PATI boundaries in the upper Quaternary section of ODP Hole 994C equate with the glacial Terminations recognized globally by Broecker and van Donk (Reference Broecker and van Donk1970). These Terminations occurred at times of particular instability in high-latitude ice masses and were triggered by Milankovich factors during late Quaternary time, at a periodicity of c. 100 ka (Berger, Reference Berger2013). It is not known however if the PATI boundaries in Hole 1261A equate with those found by Hayek & Wilson (Reference Hayek and Wilson2013) in Hole 994C although, should their hypothesis of PATI boundaries coinciding with Terminations prove correct, then the PATI boundaries should be useful for correlation between holes. The data from ODP Hole 1261A do not contradict this suggestion. Hayek & Wilson (Reference Hayek and Wilson2013) recorded 11 PATIs within the upper Quaternary section (MIS 1–12) of ODP Hole 994C, based on nine peaks at which ATIs >(
$\bar x$
+ σ) and one peak, at the PATI-9–10 boundary, at which ATIs ≈ (
$\bar x$
+ σ). In ODP Hole 1261A a peak at the PATI-4–5 boundary also rounded to (
$\bar x$
+ σ). The boundaries between PATIs 1–4 in Hole 1261A are closely spaced. Hayek & Wilson (Reference Hayek and Wilson2013) found Termination I, separating the Pleistocene and Holocene, is in ODP Hole 994C likewise marked by several closely spaced PATI boundaries. They suggested that this may be because the 120 m sea-level rise during this Termination (Poag & Valentine, Reference Poag and Valentine1976), occurred in several decimetre steps (Blanchon & Shaw, Reference Blanchon and Shaw1995) that were associated with changes in ocean chemistry. The first appearance of the calcareous nannofossil Emiliania huxleyi in Hole 994C occurred at 8.05–9.05 mbsf (Okada, Reference Okada, Paull, Matsumoto, Wallace and Dillon2000), which lies within MIS 9 (Oba, Shikama & Okada, Reference Oba, Shikama, Okada, Paull, Matsumoto, Wallace and Dillon2000) and PATI-9 of Hayek & Wilson (Reference Hayek and Wilson2013) at this hole. In ODP Hole 1261A E. huxleyi was recorded at 3.83–3.90 mbsf, although this was most likely not the lowest occurrence. This depth lies within PATI-7 at ODP Site 1261A. Pseudoemiliania lacunosa was in ODP Hole 1261A recorded from 4.61 mbsf which equates to PATI-10, although this is unlikely to be the highest occurrence. In ODP Hole 994C the highest occurrence of P. lacunosa was recorded at 19.04–20.04 mbsf, which Oba, Shikama & Okada (Reference Oba, Shikama, Okada, Paull, Matsumoto, Wallace and Dillon2000) suggested lies within MIS 12. This agrees with the finding by Thierstein et al. (Reference Thierstein, Geitzenauer, Molfino and Shackleton1977) that the extinction of P. lacunosa occurred synchronously worldwide during MIS 12. The closely spaced occurrences of E. huxleyi and P. lacunosa in ODP Hole 1261A at 3.90 and 4.61 mbsf, respectively, indicates that the MIS 9–12 section is here condensed. It is therefore probable that the 10 cm sample spacing in this study did not resolve all the potential PATI boundaries in the upper Quaternary section at Hole 1261A, and that long range correlation using PATI boundaries remains a distinct possibility.
6. Conclusions
The upper Quaternary sediment of the Demerara Rise was deposited under the influence of an oxygen minimum zone (OMZ), with ODP Hole 1261A situated near the upper boundary of the OMZ. The suboxic Bulimina–Uvigerina–Cassidulina fauna differs from the similarly suboxic Uvigerina–Hoeglundina–Gyroidina fauna anticipated by Schnitker (Reference Schnitker1980) and may reflect the impact of local upwelling. The lower and the uppermost few samples from the section contained abundant Cassidulina laevigata and glauconite, while the intervening samples contained lesser glauconite and abundant Uvigerina peregrina. This is concluded to indicate that the relative position of the upper boundary of the OMZ fluctuated during at least part of late Quaternary time. It may have shoaled, or some subsidence of the Demerara Rise may have occurred after the 485 ka evolutionary appearance of the calcareous nannofossil Pseudoemiliania lacunosa. Neither hypothesis can be accepted without question. While we cannot think of a process that would have caused shallowing of the OMZ upper margin across a series of glacial–interglacial cycles, neither can we envisage a means by which the palaeodepth of the OMZ remained constant relative to a subsiding seafloor over time. Following this crustal subsidence or shoaling of the OMZ, either the progradational accumulation of sediment on the sea floor or migration of the OMZ boundary back into deep water led to a latest Pleistocene and Holocene re-establishment of a shallower-water C. laevigata fauna. The percentage of grains as benthonic foraminifera (%B) is higher in the samples with abundant U. peregrina, suggesting that the higher values of %B are a product not of dissolution but of the exclusion of predatory macrobenthos towards the OMZ core. The distribution of the oxic indicator Planulina wuellerstorfi supports the model of crustal subsidence or migration of the OMZ boundary, this species being associated with C. laevigata but not associated with either B. aculeata or U. peregrina. Superimposed on this longer-term change in the position of ODP Site 1261 relative to the OMZ margin due to subsidence are shorter-term turnovers in the benthic foraminiferal assemblage. A conditioned-on-boundary index (CoBI) demonstrates that Bulimina aculeata, C. laevigata and U. peregrina participated most consistently in these short-term assemblage turnovers. An interval with abundant Epistominella exigua might indicate a brief interlude with a regular spring bloom of phytoplankton.
Previous workers examining the Quaternary micropalaeontological record have tended to search for glacial–interglacial contrasts between assemblages. The bathyal foraminiferal record and the stratigraphic distribution of glauconite on the Demerara Rise, off northern South America, suggests that locally at least there are longer-term fluctuations that spanned several marine isotope stages, in this case related to either shoaling of an OMZ or crustal subsidence. Micropalaeontological workers should therefore be wary of trying to force all interpretations of changes in deep-water benthic foraminiferal assemblages to fit a simple glacial–interglacial model.
Acknowledgements
This research received no specific grant from any funding agency, commercial or not-for-profit sectors. This research used samples provided by the Ocean Drilling Program (ODP) that is sponsored by the US National Science Foundation (NSF) and participating countries under management of Joint Oceanographic Institutions (JOI), Inc.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0016756814000673
