Planktic Foraminiferal Test Size and Weight Response to the Late Pliocene Environment

Atmospheric carbon dioxide (pCO2atm) is impacting the ocean and marine organisms directly via changes in carbonate chemistry and indirectly via a range of changes in physical parameters most dominantly temperature. To assess potential impacts of climate change on carbonate production in the open ocean, we measured size and weight of planktic foraminifers during the late Pliocene at pCO2atm concentrations comparable to today and global temperatures 2 to 3 °C warmer. Size of all foraminifers was measured at Atlantic Ocean Deep Sea Drilling Project (DSDP) Site 610, Ocean Drilling Program (ODP) Site 999, and Integrated Ocean Drilling Program (IODP) Site U1313. Test size was smaller during the Pliocene than in modern assemblages under the same environmental conditions. During the cold marine isotope stage (MIS) M2, size increased at Site 999, potentially linked to intensified stratification of the surface ocean in response to the closure of the Central American Seaway. At Site U1313, test size tracks the warming throughout the late Pliocene. Size‐normalized weight (SNW) of Globigerina bulloides at Site U1313 decreased during warmer temperature intervals. SNW of Globigerinoides ruber (white) at Site 999 displays high‐frequency variability not correlated to temperature. Yet during the glacial period within MIS M2, test weight was higher during higher temperatures. Our results support studies in the modern ocean, which challenge the view that carbonate chemistry is the primary driver for calcification. To better understand processes driving changes in SNW, computer tomography was used to quantify calcite to volume ratios. During interglacial periods, lower calcite volume but higher test volume suggests less suitable conditions for calcification. As this signal is not evident in SNW, subtle changes in calcification might not be observed by the weight‐based method.


Introduction
Atmospheric carbon dioxide (pCO 2 atm ) has increased from 280 parts per million (ppm) at the start of the industrial revolution (1750; Siegenthaler et al., 2005) to over 410 ppm today (2019; National Aeronautics and Space Administration, 2019). This increase in pCO 2 atm results in changes in the carbonate chemistry of the ocean, warming, increased stratification, and oxygen loss, all of which are projected to impact marine organisms (Pörtner et al., 2014).
Marine calcifying organisms play a fundamental role in the inorganic carbon cycle (Boyce et al., 2010;Henehan et al., 2017). Carbonate production in the ocean is roughly equally divided between the shelf and the open ocean (Cartapanis et al., 2018). In the open ocean ecosystem, planktic foraminifers are a significant sink for CaCO 3 , producing 32 to 80 % of CaCO 3 flux to marine sediments (Schiebel, 2002). Therefore, any changes in their calcification in response to climate change could have impacts on the inorganic carbon cycle.
The amount of carbonate produced by an individual foraminifer is a function of the specimen's size and the thickness of the test. Growth of foraminiferal tests is regulated by several environmental factors including pH, temperature, salinity, light, oxygen, and nutrient levels, together with food availability (Kucera, 2007). Today, larger test sizes are generally found in the warm subtropical to tropical oceans with decreasing size toward the poles (de Villiers, 2004;Schmidt, Renaud, et al., 2004); trends across the group are the result of species-specific size. This pattern reflects the larger species diversity, and inclusion of relatively large species, in warmer subtropical and tropical waters. Within a single species, size is tied to optimum growth conditions, outside of which growth rates decrease and reproduction ceases (Schmidt et al., 2006;Schmidt, Renaud, et al., 2004). Increased test size in warmer surface waters has been related to a combination of carbonate saturation, faster metabolic rates, higher light intensity, and greater niche diversity due to stronger stratification (de Villiers, 2004;Lombard et al., 2009;Schmidt, Renaud, et al., 2004). As several of these factors are related to temperature, this suggests that temperature is the dominant control on growth in lower latitudes (Bijma et al., 1990) though the modern dataset does not allow to distinguish between temperature per se or stratification . Modulating this overarching pattern, from oligotrophic to mesotrophic environments, increasing food facilitates growth to larger sizes until high productivity limits light to support symbiont bearing species, which are then replaced by smaller nonsymbiont bearing ones (Schmidt, Renaud, et al., 2004). Furthermore, sizes in highly changeable environments such as frontal systems and upwelling areas tend to be smaller than the general size temperature trend would suggest (Schmidt, Renaud, et al., 2004).
Other than size, foraminiferal weight is a fundamental indicator of foraminiferal carbonate production. Planktic foraminifera in general respond to changes in lower pH, carbonate ion concentration, and saturation with thinner tests and reduction in test size (Barker & Elderfield, 2002;Bijma et al., 1999;Henehan et al., 2017;Naik et al., 2010;Spero et al., 1997). This calcification response is more prominent in nonsymbiotic foraminifera than in symbiotic species, which are able to elevate the pH in surrounding waters due to photosynthetic CO 2 fixation (de Nooijer et al., 2009;Rink et al., 1998;Wolf-Gladrow et al., 1999). The drivers of weight change in foraminifer are still heavily debated as some species and regions show no dependency on the carbonate system suggesting that temperature and productivity drive a change in calcification (Beer et al., 2010;Weinkauf et al., 2013;Weinkauf et al., 2016). For example, reduced salinity may inhibit calcification in some species of foraminifera (Beer et al., 2010;Bijma et al., 1990;Weinkauf et al., 2013).
We tested our understanding of the impact of climate change on foraminiferal carbonate production by analyzing size and weight at two sites during the late Pliocene (3.4 to 2.9 Ma) with reconstructed pCO 2 peaking at~420 to 450 ppm (Dowsett & Gill, 2010;Martínez-Botí et al., 2015;Pagani et al., 2010;Seki et al., 2010). The Pliocene ocean had similar circulation and faunal distribution to modern (Robinson et al., 2008) and therefore is an ideal test interval for this comparison (Dowsett et al., 2012;Haywood et al., 2013). Global sea surface temperature (SST) is estimated to have been 2 to 3°C higher, surface pH~0.06 to 0.11 units lower, and sea level~25 m above current levels (Dowsett et al., 2012;Dwyer & Chandler, 2009;Haywood et al., 2016;Hönisch et al., 2012;Miller et al., 2012;Robinson et al., 2008 Karas et al., 2017); changes in pCO 2 also likely played an important role (Tan et al., 2017). Unfortunately, no high resolution pCO 2 records across MIS M2 currently exist to offer further insight into this question.
At Sites 610, 999, and U1313, we measure the size of all specimens within the late Pliocene assemblages and determine size-normalized weight (SNW) of Globigerinoides ruber (white) in the Caribbean Sea (Site 999) and Globigerina bulloides in the North Atlantic Ocean (Site U1313). These species were chosen due to their high abundance in the assemblage at each location to asses if smaller size is associated with thinner tests in response to the higher atmospheric pCO 2 in the late Pliocene. To interpret any changes in SNW, we used microcomputer tomography to measure changes in thickness and calcite on each specimen.
Our previous work suggests competing impacts on assemblage size of evolutionary tendency for larger size in the most modern icehouse assemblages ( Paleoceanography and Paleoclimatology ecological drivers result in largest sizes in highly stratified warm waters today (Schmidt, Renaud, et al., 2004). Consequently, low-resolution studies suggest that the size in foraminiferal assemblages is lower during the Pliocene than today , and therefore, we are expecting smaller sizes compared to modern. Therefore, supported by research on glacial-interglacial variability (Barker & Elderfield, 2002;Schmidt et al., 2003), we would expect the Pliocene assemblages to be overall smaller than modern and smallest during M2. Based on laboratory studies and sediment samples, the expectation is that in a warmer world with higher pCO 2 atm , the weight of individuals should be lower, compared to modern (Barker & Elderfield, 2002;Bijma et al., 2002). The growth and calcification response should be largest in regions of the ocean where the environment varies most between glacial and interglacial intervals (e.g., higher latitudes) and smaller in regions with lower amplitudes of change (e.g., lower latitudes; Schmidt et al., 2003).

Materials
We analyzed planktic foraminifers from Ocean Drilling Program ( Figure 1) is located in the Caribbean Sea. Hole 999A is predominantly composed of nannofossil clayey mixed sediments with foraminifers and foraminiferal clayey mixed sediment with nannofossils (Sigurdsson et al., 1996). We adopt the age model of Groeneveld (2005), which was initially established by Haug and Tiedemann (1998b). Sedimentation rate during the late Pliocene was 3.0 to 3.3 cm/kyr (Sigurdsson et al., 1996). One hundred thirty-four samples over 419 kyr were analyzed, resulting in an average temporal resolution of 3.1 kyr.
Site U1313 is situated at the base of the upper western flank of the mid-Atlantic Ridge (41°0′N, 32°57′W) in a water depth of 3,413 m (Channell et al., 2006; Figure 1). The core material consists of a carbonate nannofossil ooze, with less than 5 % clay (Channell et al., 2006). The age model is based on De Schepper et al. (2013) and Naafs et al. (2010). The sedimentation rate was 4.1 to 4.5 cm/kyr throughout the late Pliocene (Channell et al., 2006). One hundred twenty-four samples, from both Holes U1313B and U1313C, over 212 kyr were used with an average temporal resolution of 1.7 kyr.
Hole 610A was cored in the northern North Atlantic Ocean (53°13′N, 18°53′W; Figure 1) at a water depth of 2,417 m (Ruddiman et al., 1983) and is composed of cyclic alternations of pelagic sediments with glacial mud and interglacial nannofossil ooze (Ruddiman et al., 1983). We use the age model from the initial site report (Ruddiman et al., 1983) suggesting a sedimentation rate of 5.1 cm/kyr (Ruddiman et al., 1983). Sixteen samples over 87 kyr, with a temporal resolution of 5.4 cm/kyr, were used in the determination of latitudinal size gradients in the late Pliocene.
We incorporate Mg/Ca SST estimates, measured on Trilobatus sacculifer, from Groeneveld (2005) at Hole 999A, with a resolution of 6 kyr and compare to the alkenone temperature record with a resolution of 13 kyr (Bartoli et al., 2011;Martínez-Botí et al., 2015). The alkenone temperature record has been to complement the Mg/Ca SST as this record might be biased toward lower temperatures at Site 999 due to dissolution as indicated by fragmentation. We use the Site U1313 Mg/Ca temperature record, measured on G. bulloides, from De Schepper et al. (2013), with a resolution of 2 kyr and compare to the alkenone temperature record with a resolution of 4 kyr (De Schepper et al., 2013). The G. bulloides Mg/Ca SST record for Hole 610A is also taken from De Schepper et al. (2013). The LR04 benthic foraminifera oxygen isotope stack (Lisiecki & Raymo, 2005) has been used to correlate the samples. As no high-resolution atmospheric CO 2 (pCO 2 atm [ppm]) data exist for these sites, we use the boron isotope data from Martínez-Botí et al. (2015) and Bartoli et al. (2011) to provide information on pH changes.

Paleoceanography and Paleoclimatology
representatively using a microsplitter so that~800 to~3,000 specimens of each sample were photographed at a magnification of 160x following . Morphological parameters were analyzed using Olympus Stream Motion. Parameters were set to exclude particles with a maximum diameter below 150 μm, below sphericity of 0.5, and below the mean grey value of 40; the particles outside these thresholds represent broken foraminifers, other microfossils, and rock fragments such as ice rafted debris or volcanic ash. Benthic foraminifera, ostracods, diatoms, and other material were removed through manual assessment of the images with the largest particle size to avoid bias in the data. The 95 th percentile of the maximum diameter was calculated on the remaining data, following the method of Schmidt, Renaud, et al. (2004) as the differences between the size spectra of the assemblages are predominantly changing the skewness of the distribution. The average measurement error of the 95 th percentile of the assemblage size (size 95/5 ) across both sites is 0.84 %, equal to 3.07 μm, and is therefore within the calculated accuracy of diameter measurements of 3.31 μm . The error was determined by the resplitting and repeat measurement of 51 samples.

Preservation
Currently, the lysocline is found at a depth between 4 and 5 km in the Atlantic Ocean (Thunnel, 1982) and in the Caribbean Sea (Archer, 1996), and hence, all sites are significantly above the carbonate compensation depth (Sigurdsson et al., 1996). In the Caribbean, during the Pliocene, the lysocline was deeper than today (Sigurdsson et al., 1996;Tiedemann & Franz, 1997) with potential impacts on size spectra in the assemblages; fragmentation is apparent in many samples.
Fragmentation can bias the assemblage size toward both higher and lower values by selectively removing more dissolution susceptible species, such as Gs. ruber and Orbulina universa (Berger, 1970) or by increasing the abundance of small fragments. To determine changes in carbonate preservation, fragmentation was counted in at least 200 particles of all samples, following the method of previous work on Pliocene samples from Sites 999 and U1313 (Davis et al., 2013) and the method by Le and Shackleton (1992). The ratio of broken versus complete foraminifers was calculated.

Size-Normalized Weight
To determine the SNW of foraminifera at ODP Site 999 through the Pliocene, 20 to 30 individuals of Globigerinoides ruber sensu stricto ( Figure 2) were picked from the 300-to 355-μm size fraction of 86 samples following the method of Barker and Elderfield (2002). Other morphotypes are known to have different habitats (Aurahs et al., 2011), with Gs. ruber sensu lato ( Figure 2) precipitating tests in colder surface waters than Gs. ruber sensu stricto (Steinke et al., 2005). For SNW of Site U1313 samples, 15 to 30 specimens of G. bulloides ( Figure 2) were picked from the same size fraction for 48 samples, following the same method.
There are several morphospecies of G. bulloides within which are genotypes that inhibit a range of ecologies (Darling & Wade, 2008). Based on the location of Site U1313, G. bulloides types IIa-IIe, with IIa and IIb likely to be dominant, are the genotypes present (Darling & Wade, 2008). There are no studies to show a morphological difference between these types. Samples were weighed to six-decimal places using a Mettler Toledo analytical balance to determine the average weight of the test (sieve-based weight). For every sample, the length of each specimen in the same orientation was taken from digital images using ImageJ. SNW for all samples was calculated by normalizing sieve-based weight to the mean diameter for the corresponding size following Barker and Elderfield (2002). Our data were combined with data from Davis et al. (2013), which was determined using the same method on the same instruments.

Microcomputer Tomography
To better interpret any changes in SNW, 15 of the Gs. ruber specimens that were used for SNW measurements (see section 2.2.3) were subsampled from four samples of Site 999, guided by the oxygen isotope (δ 18 O) record, to cover samples before, during and after MIS M2. Microcomputer tomographic images were taken using a Nikon XT H 225 ST CT scanner (120 kV, 58 Va), configured at 2.1-to 2.3 μm resolution, with an exposure time of 0.5 s and a total of 3,141 projections. These images were processed in Avizo 9.0 (Thermo Scientific™). The calcite test of the foraminifera is isolated from any infill to give the volume of the calcite (Figure 3). The internal cavity of each chamber (incorporating any infilling, to represent the true internal space) was manually isolated following Caromel et al. (2015) and Schmidt et al. (2016). External volume and volume of the calcite test were added to determine the volume of the foraminifera. The calcite to volume (CV) ratio was calculated ( Figure 3). Due to the resolution of the scan, any change below 4 μm cannot be resolved with confidence. The method would also highlight any infilling with secondary calcite as a potential cause for weight changes.

Results
The It is important to note that the pCO 2 atm record is at a much lower resolution, in particular across MIS M2. There is very little variability of CO 2 within the precision of the data during and prior to MIS M2, questioning if CO 2 was the driver for the glaciation (Bartoli et al., 2011). At Site U1313, size 95/5 ranges from 290 to 444 μm ( Figure 4a) and has a weak positive correlation with Mg/Ca SST (r = 0.40, P < 0.001), and no correlation to alkenone SST (r = 0.24, P = 0.062). In general, warming is associated with larger size throughout the Pliocene.
To determine latitudinal size gradients in comparison to the Holocene, size data were added at lower resolution from Site 610 ( Figure 5).

Paleoceanography and Paleoclimatology
At Site U1313, SNW ranges from 14.26 to 21.70 μg for G. bulloides (Figure 4c), with no correlation with fragmentation (r = −0.08, P = 0.54). SNW and both Mg/Ca SST and alkenone SST are not correlated across the Pliocene (r = −0.23, P = 0.12; r = 0.004, P = 0.98), although test weight is generally lower at cooler temperatures (Figure 4b). Fragmentation increases during this period but is not correlated with SNW (r = −0.11, P = 0.34; Figure 4b). Both temperature and SNW fluctuate before and after the glacial period by 4.5°C and 4.46 μg and 6.5°C and 6.20 μg, respectively.

Paleoceanography and Paleoclimatology
TODD ET AL.
To better understand drivers of SNW, we analyzed CV ratios of Gs. ruber, the dominate species at Site 999. CV ratios during the Pliocene (Figure 6) show more calcite per volume during MIS M2 with averages of 46 % than compared to 40.3 % before MIS M2 and 41.0 % after.

Discussion
Planktic foraminifers strongly respond to their environment as they cannot regulate their temperature and are passively transported by ocean currents. As such they are an ideal group to study past records of climate change. Reconstructing biotic response, though, will always be hampered by our inability to describe the complex marine system with multiple drivers of change in its entirety. Carbonate chemistry is rather stable throughout the late Pliocene based on low-resolution measurements, though a higher frequency variability would be expected in analogy to Pleistocene glacial-interglacial cycles (Jansen et al., 2007;Siegenthaler et al., 2005). Therefore, our discussion focuses on the physical environment, specifically temperature and stratification, and its impact on size in the entire community of foraminifers and weight in the species Gs. ruber and G. bulloides. Based on our understanding of modern drivers of changes in size and weight in foraminifers, we would expect to see large changes in assemblage size in more variable environments and lower shell weight in lower saturation regions and time intervals where the physiological stress is strongest. Due to the evolutionary offset in size over the Neogene, assemblage size in the Pliocene should be recognizably smaller than today and increase toward the tropics.
Size 95/5 at Site 999 shows high-frequency variability throughout the Pliocene that is not related to temperature, which remains relatively stable through the study period (Groeneveld, 2005). Overall, there is little temperature change over a significant size change with cooler MIS M2 temperatures associated with the largest sizes ( Figure 7). As mentioned in the methods, the Mg/Ca temperature record at this site is biased toward cooler temperatures and thus might not faithfully surface temperatures. Generally, higher fragmentation is associated with warmer temperatures as expected if warmer temperatures are related to higher pCO 2 atm . While preservation increases during MIS M2 at Site 999 (Figure 4), there is no correlation between fragmentation and size. This suggests that there is no bias via selectivity removing fragile larger species such as T. sacculifer and Gs. ruber nor many small fragments biasing the size data. Additionally, although there is increased fragmentation, it does not overprint our weight signal, as no signs of dissolution are observed in the CT data obtained for SNW ( Figure 6; see section 3). Consequently, we interpret larger sizes and weight to reflect a biotic response, rather than a reduction of dissolution biasing the signal (see section 3).
Counterintuitively, the coldest and warmest temperatures at this location have been reconstructed during MIS M2. It is important to note that this is not a global signal but reflects the local conditions, which possibly highly sensitive to intermittent exchange between the Atlantic and Pacific Oceans in the final stages of the closure of the CAS (Bartoli et al., 2005;Haug & Tiedemann, 1998a). At this location, cooling starts before MIS M2 at 3.32 Ma and warming starts during the peak glaciation. De Schepper et al. (2013) interpreted the warming as the result of a sea level drop at the peak glaciation of MIS M2, which stopped the influx of cold Pacific water into the Caribbean.
The amplitude of size change at Site 999 is larger than at Site U1313 with a maximum change of 165 μm versus 154 μm, respectively. In contrast to Site 999, Site U1313 size is positively correlated with temperature with an increase in size 95/5 tracking the warming throughout the record (Figure 7). Alkenone SST suggests a cooling during MIS M2 at Site U1313, which is not apparent in the G. bulloides Mg/Ca SST, which reflect mixed layer temperatures. The surface warming corresponds with the increase in size 95/5 Figure 5. Holocene and Pliocene mean of the 95th percentile of assemblage size (μm) per biogeographic area plotted against sea surface temperature (SST;°C). The error bars represent the 95 % confidence intervals. The black solid line corresponds to the regression line (r = 0.938, P = 0.006) for the Holocene, using mean annual SST (Schmidt, Renaud, et al., 2004). The black dashed line indicates the trends though the new late Pliocene. Late Pliocene SST is determined as the average values: the blue symbols represent sample within MIS M2, and red symbols represent sample outside MIS M2 from Figure 4, see references for the temperature data in Figure 3 and section 2. All sites are plotted using Mg/Ca, with the exception of "999 alk," which uses alkenone SST. The black scale bar on the right of the graph represents the offset of the late Pliocene average assemblage size to the lowresolution data from Schmidt, Renaud, et al. (2004).
implying that temperature is the dominant control on assemblage size at this site. It is unclear what has caused the divergence between the two SST proxies, but it is possible that a different genotype of G. bulloides has been used to produce the Mg/Ca SST record (Darling et al., 2003;De Schepper et al., 2013). It is thought that G. bulloides type IId has a different environmental preference from the other genotypes (Darling et al., 2003), but it is not thought to be found at Site U1313 (Kucera & Darling, 2002), where there is potential for mixing of the warm and cool water genotypes (Darling et al., 2003;Darling & Wade, 2008). However, there are no high-resolution alkenone records available for Site 999 to compare to Site U1313. Temperature records though for Site U1313 show a warming at the end of the glacial MIS M2, which is attributed to the weakening of the Atlantic Meridional Ocean Circulation (AMOC), causing a deepening of the thermocline and surface warming (Karas et al., 2017;Steph et al., 2006;Steph et al., 2010).
Changes in foraminiferal assemblage size can be driven by changes in species diversity and by changes in size of the dominant species. A multitude of ecological conditions have been linked to size changes in planktic foraminifera, including pCO 2 , SST, stratification, salinity, and nutrient levels .
Changes to species abundance and extinctions of larger species are also important for size variation. Throughout the late Pliocene, the foraminiferal assemblage at Site 999 is affected by changes in species, either by extinction or fluctuations in occurrence (Berggren, 1969;Chaisson & Hondt, 2000). Dentoglobigerina altispira (3.1 Ma) and many of the Neogloboquadrina species go extinct (Chaisson & Hondt, 2000) in the late Pliocene. Fluctuations in the abundance of menardellids, Neoglobquadrina dutertrei, Globigerinita glutinata, Gs. ruber, T. sacculifer, and the Globoturborotalia group are observed (Chaisson & Hondt, 2000), all of which will impact the sizes ranges within the assemblages. Throughout the record, Site 999 is dominated by species of the Globigerinoides, Trilobatus, and Menardella genera, all are of the largest extant planktic foraminifers. The extant species of these groups have optimal growth conditions at temperatures between 25 and 28°C. The

10.1029/2019PA003738
Paleoceanography and Paleoclimatology minor temperature changes between 21 and 25°C documented for this site in the late Pliocene might therefore not lead to large changes in the size of individual species (Schmidt et al., 2006;Schmidt, Renaud, et al., 2004). However, the higher alkenone SST at~28°C (Bartoli et al., 2011;Martínez-Botí et al., 2015) brings the temperature closer to the optimal growth conditions, which could lead to individual size changes. Additionally, warming may lead to increases in size due to increases in abundance of Trilobatus over Globigerinoides when temperatures and salinities become closer to their species optima.
Furthermore, light availability might impact size of symbiont bearing species and their ability to thrive in these settings. The impact of light on foraminiferal assemblage composition and size has been first assessed by Ortiz et al. (1995) for the California upwelling system. They showed that high turbidity reduces light availability and thereby reduces size in symbiont-bearing species. The high productivity and resulting low light levels increase the abundance of nonsymbiont bearing species such as G. bulloides, which has typically smaller sizes than Gs. ruber and T. sacculifer, thereby decreasing overall sizes in the assemblage (Ortiz et al., 1995). Light availability could therefore contribute to size changes at Site 999, considering the likely changes in the position of the thermocline due to stratification changes throughout the Pliocene.
Site U1313 contains a mixture of warm water species such as Gs. ruber and intermediate and cold-water species such as G. bulloides, Globigerina falconensis, and Neogloboquadrina incompta (Channell et al., 2006). The dominant species in the analyzed time interval are G. bulloides, N. incompta, Gt. glutinata, and Globrotalia hirsta. G. bulloides sizes, for example, react to environmental change but show no overall net increase during this time period (Malmgren & Kennet, 1978). On contrast, a size increase of~75 μm in N. pachyderma has been described for the past 1.3 Ma in the subarctic and temperate North Atlantic (Huber et al., 2000). Large-temperature fluctuations due to migration of the frontal systems led to a high variability in assemblage size in the Pleistocene . The variability of the North Atlantic Current is expressed in a dinoflagellate cyst assemblage shift and supported by both Mg/Ca and alkenone SST reconstructions for this site of 16 to 23°C (De Schepper et al., 2013;Naafs et al., 2010), and modeling outputs (Haywood et al., 2016). Both assemblage composition and size of individuals would be predicted to change in response; warming leads to an increase in abundance of Gs. ruber and T. sacculifer, which are generally larger than G. bulloides, G. falconensis, and N. incompta (Schmidt, Renaud, et al., 2004). Additionally, the size of the colder water species would decrease further away from their optimum, while the size of the warm water species would increase. In the Holocene, samples recovered from near frontal systems are additionally smaller than expected from the global temperature size trend (Schmidt, Renaud, et al., 2004).
Higher amplitudes of warming result in large size change in both the Pleistocene  and the Pliocene. The amplitude of warming scales with size across time, suggesting that the drivers of assemblage size in the Pleistocene were the same in the Pliocene and as such have predictive power across the entire time interval. In the Pleistocene, size changes of~100 μm are evoked by a 10°C change in temperature at Site GeoB1105 (1°39.9′S, 12°25.7′W, 3,225 m water depth; Schmidt et al., 2003) and~25 μm by a 2.5°C temperature change at the subtropical gyre Site GeoB1413 (15°40.8′S, 9°27.3′W, 3,789 m water depth; Schmidt et al., 2003). From the relationship between assemblage size and temperature shown in Figure 5, a size change of~80 μm is associated with a 5°C change in temperature at Site 999 and~70 μm with a 5°C change at Site U1313. This relationship indicates that the link between foraminiferal size and palaeoclimatic fluctuations holds throughout the last 3 Myr despite changes in composition of assemblages over time, although not at the same level: over time, the increase in size is lower with the same change in temperature.
A comparison of the Pliocene planktic foraminiferal size and temperature relationship to the Holocene ( Figure 5) is tentative due to the low number of sites in the Pliocene. Our dataset points to a significant difference between these time intervals, however, as Pliocene samples are significantly smaller than those in the Pleistocene. In the tropics Pliocene foraminifera are~50 to 100 μm smaller than modern specimens at the same SST (~490 to 550 μm in the late Pliocene to~540 to 600 μm today). This offset of around 100 μm agrees with a much lower resolution record covering the entire Cenozoic by Schmidt, Renaud, et al. (2004). The large species still dominating today's assemblages in the tropics are the same as those at 3 Ma. The observed increase in assemblage size today compared to the Pliocene is either an increase in relative abundance of the largest species such as Globorotalia menardii and Globorotalia tumida, due to increased temperature (Mary & Knappertsbusch, 2013), an increase in their size through time as documented for the latter species (Malmgren et al., 1983;Schmidt et al., 2016), or a combination of both.
The size of foraminifera at Site U1313 is smaller than expected based on the general temperature size relationship. A similar negative divergence is also observed at modern sites with high-temperature variabilities ( Figure 5). While we would have expected a smaller difference in size at Site 610 given the stability of the higher latitude size records over the Cenozoic, we are reluctant to interpret the data further due to the low number of sites analyzed. There is evidence that the Pliocene equator to pole temperature gradient was lower as a result of the warmer world Salzmann et al., 2008)) and as such might be an underling reason for this difference combined with other factors such as stronger exchange of species at the tropical site and a relative increase in size of existing species combined with a polar amplification of warming, resulting in relatively warmer temperatures in midlatitudes and little warming in the tropics (Haywood et al., 2016).
Similar to the size data, the SNW data need to be considered within a global and a local framework. In both modern and geological field studies of a variety of species, multiple drivers of weight change are indicated. In general, calcification is energetically more difficult outside optimum CO 2 conditions for each species (de Nooijer et al., 2009;Foster et al., 2013;Henehan et al., 2017) and at lower carbonate ion concentrations (Kroeker et al., 2010). In laboratory experiments, a clear reduction of weight in response to lower carbonate ion changes is documented (Beer et al., 2010;Bijma et al., 1999;Bijma et al., 2002;Henehan et al., 2017;Lombard et al., 2010;Russel et al., 2004). A reduced shell mass by up to 50% in response to increased CO 2 , and therefore decreased CO 2 3 − , was related in both recent field studies (Henehan et al., 2017;Moy et al., 2009;Osborne et al., 2016) and over a broad geological timescale (Barker & Elderfield, 2002;Gonzalaz-Mora et al., 2008;Naik et al., 2010). The higher Pliocene CO 2 may therefore have resulted in lower calcification, thinner and/or smaller tests.
Yet Weinkauf et al. (2016Weinkauf et al. ( , 2013 and de Moel et al. (2009) found that temperature, upwelling, productivity, and optimum growth conditions influence weight, regardless of changes in CO 2 3 − . Weinkauf et al. (2016) suggests that temperature and productivity control SNW in modern Gs. ruber with temperature increasing weight and productivity decreasing it. In contrast to Barker and Elderfield (2002), Weinkauf et al. (2016) did not find any environmental control on SNW in G. bulloides from the sediment core NEAP 8K in the Northeast Atlantic (59°48′N 23°54′W; water depth, 2,360 m). Gonzalaz-Mora et al. (2008) found lower SNW in Gs. ruber at low CO 2 and low temperature during MIS 7 in the Mediterranean, while for G. bulloides in the same samples, a high weight is correlated with low CO 2 and low temperatures, supporting Barker and Elderfield (2002). This led to the suggestion that the relationship between atmospheric CO 2 and weight of foraminiferal tests is species-specific and varies with location and environment, with some species showing no sensitivity at all (Davis et al., 2013;Henehan et al., 2017). Overall, there appears to be no consensus on foraminiferal calcification response to changes in CO 2 , and many species responses are unknown.
These local interpretation and species-specific interpretation of drivers of calcification complicate the interpretation of SNW. SNW of Gs. ruber at Site 999 is heavier during the glacial interval. There is a large stepwise increase in SNW during MIS M2, parallel to warming which is interpreted as a reduction of influx of low saturation waters from the Pacific, as there is no correlation between SNW and temperature. Thus, our data may be suggesting that calcification is being controlled by a factor other than pCO 2 or that this location is not in equilibrium with the atmosphere. Today, the location reflects atmospheric conditions (Takahashi et al., 2014), but an influx of Pacific waters in the earlier part of the record, which is subsequently cut off due to eustatic sea level change at the peak of MIS M2 (De Schepper et al., 2013), would result in higher CO 2 and lower carbonate ion concentration due to the influx of upwelled waters in the earlier part of the record. Before MIS M2, we observe an overall lower SNW in Gs. ruber than after MIS M2 (Figure 4c), which would support an impact of carbonate ion on SNW in Gs. ruber in response to the local conditions.
In contrast, the SNW of G. bulloides at Site U1313 shows a large drop during the initial stages of MIS M2, which is a period of warming SST at U1313, although there is no correlation between temperature and SNW. The lower SNW during MIS M2 could therefore either point toward lower pH and carbonate ion concentrations and higher CO 2 during MIS M2 or link to higher energetic needs at warmer temperatures impacting calcification in the species as suggested by Davis et al. (2017) for the same location. Weight increases strongly at 3.27 Ma following the SST rise and therefore infers a decrease in calcification during warmer temperatures and thus potentially higher CO 2 in G. bulloides. These findings agree with SNW for Pleistocene glacial-interglacial by Barker and Elderfield (2002) on the same species in the same region.
In our investigation of CV changes in Gs. ruber from Site 999, we found lower calcite volume, but higher test volume, during interglacial periods. This suggests that intervals with the coldest temperatures, more poorly ventilated waters, and higher productivity at this site (De Schepper et al., 2013;Haug & Tiedemann, 1998a;Osborne et al., 2019) are the least suitable conditions for calcification. We interpret this as a higher production of carbonate in Gs. ruber at lower CO 2 . During MIS M2, calcite increased while volume decreased compared to samples before and after MIS M2 despite narrowly selecting for size. Furthermore, no holes or internal changes to the inner shell were observed in the CT images, which are expected in specimens that have undergone the first stages of dissolution. These findings suggest that growth and calcification are affected in different ways. Globigerinoides ruber has significant morphological variability, and morphotypes are known to have differing ecological preferences (Bonfardeci et al., 2018;Steinke et al., 2005). While the narrowly defined sensu stricto morphotypes were picked (see section 2.2.3), there is still some morphological variation within the specimens, which may impact the overall volume of the specimen analyzed.
Assuming similar drivers for size and weight changes, we compare amplitudes of environmental change and biotic response. Over the last 50,000 years, pCO 2 varied by 100 ppm, and SNW in G. bulloides ranged between 10 and 21 μg (Barker & Elderfield, 2002). In the Pliocene, SNW ranges between from 14 to 21 μg with a 40-ppm change in pCO 2 (Martínez-Botí et al., 2015), suggesting that given the error with the reconstruction and our lack of mechanistic understanding, the responses to multiple stressors are comparable.

Conclusions
Planktic foraminiferal assemblage size data during the late Pliocene are smaller than current sizes as expected from previous low-resolution work. Size does not respond to pCO 2 at the available resolution. Local environments are fundamental to size changes as indicated by assemblage size at Site 999, which is impacted by changes in surface ocean circulation and the intermittent influx of fresher, cooler, and more productive Pacific waters into the region in the final stages of the closure of the CAS. While a general increase in size tracks warming throughout the Pliocene, there is little response to the cooling across MIS M2 at Site U1313, whereas assemblage size at Site 999 increases. We interpret our preliminary data of overall smaller assemblage size during the Pliocene and the flatter size gradient between the tropics and poles in the late Pliocene as combination of size increases within species, changes in species composition and environmental gradients during the Pliocene.
Globigerinoides ruber and G. bulloides SNW show the complexity of drivers of foraminiferal weight. While there is a clear correlation between temperature and weight, our results show the importance in interpreting weight records in light of the local environments. Combining size and computer tomography analysis will have the potential to link physiological processes to changes in surface and volume of the organism and might shed light on the multiple drivers of SNW. The low glacial-interglacial variability in size-normalized weight during the Pliocene suggests that the rate or amplitude of change over glacial-interglacial periods may have a pronounced impact on the response of planktic foraminiferal calcification.
Our results would tentatively suggest that larger increases in pCO 2 result in a greater decline in calcification; higher resolution pCO 2 records across the late Pliocene would be needed to full corroborate this.