Mg/Ca ratios of G. sacculifer vary between 2.50 and 3.88 mmol/mol for the studied time interval (4.8–2.4 Ma) (Fig. F2A), providing SSTMg/Ca of 21.4°–27.2°C (Fig. F2B). Short-term SSTMg/Ca fluctuations are on the order of 2°–2.5°C. A best fit to reveal long-term trends was calculated to show only variations on timescales longer than 185 k.y. (Fig. F2). The general trend in SSTMg/Ca shows warming (24.5°–25.5°C) between 4.8 and 3.7 Ma (Fig. F2B). After 3.7 Ma and more pronounced after 3.2 Ma, a cooling trend is apparent, paralleled by a continuous increase in 
18Obenthic (Tiedemann et al., this volume). This cooling after 3.2 Ma reflects irreversible intensification of NHG (Fig. F2D). Lowest SSTMg/Ca are recorded during marine isotope Stages (MIS) 96 and 100 (~2.5 Ma), when SSTMg/Ca dropped to values below 22°C. Highest SSTMg/Ca are recorded in samples older than 4.0 Ma, when maxima approached 27°C (Fig. F2B).
Applying spectral analysis to the Mg/Ca record reveals that the apparent 400-k.y. cycle in the older part of the record, which is suggested by the best fit, shows no relationship to the eccentricity cycle of 400 k.y. (Fig. F3A). Hence, we filtered the data sets to remove variations on timescales longer than 185 k.y, like intensification of NHG. In order to do so, we subtracted a best fit with a Gaussian window of 185 k.y. and a time step of 3 k.y. from the different records. Therefore, further results from spectral analyses are based on the detrended data sets.
The Mg/Ca record is dominated by a 41-k.y. cycle between 3.7 and 2.4 Ma (Fig. F4B), as expected, because of its increasing similarity to the 18Obenthic record (Figs. F2,
F3B). The older part of the record (4.8 to 3.7 Ma) shows strong power at 100- and 23-k.y. periodicities and significantly reduced variance at 41-k.y. periodicity (Fig. F5B; Table T1). This suggests that regional low-latitude processes significantly contributed to the oceanography at Site 1241 during this period, as the presence of the 41-k.y. cycle, a typical high-latitude signal, is strongly reduced.
The 18O values of G. sacculifer vary between –0.8
and –2.18 with short-term fluctuations of ~0.5 (Fig. F2C). 18OG.sacculifer values decrease from –1.5 to –1.85 from 4.8 to 3.5 Ma. A rapid increase toward MIS M2 at 3.3 Ma with a maximum of –0.8 and a subsequent decrease to –1.8 terminate the relatively warm interval of the Pliocene record. After 3.2 Ma, isotopic values increase, according to the global trend of intensification of NHG, and culminate at MIS 96, 98, and 100 (greater than –1.0) (Fig. F2C).
Spectral analysis of the 18OG.sacculifer record between 3.7 and 2.4 Ma reveals dominant variability at the obliquity-related 41-k.y. cycle with significant power also in the 100-k.y. eccentricity band and little variability at precession-related cycles (Figs. F4C, F5C). Spectral variability in the older interval (4.8–3.7 Ma) is characterized by increased variance at precession-related cyclicities, similar to the Mg/Ca spectrum (Fig. F5C; Table T1).
We combined the 18O and Mg/Ca records of G. sacculifer to calculate variations in 18Owater (Shackleton, 1974). In the next step, we subtracted the Pliocene ice volume signal from the 18Owater record to assess relative variations in local SSS. Benway and Mix (2004) showed that the modern relationship between 18Owater and salinity in the Panama Bight is primarily controlled by variations in precipitation and its isotopic composition. An important source for the freshwater supply into the Panama Bight is the Caribbean, from where relatively 18O-depleted rainfall is transported over the Isthmus of Panama. As the Pliocene configuration of the area was significantly different than today with the Panamanian Gateway still open, the relationship between 18Owater and salinity is not known, and, accordingly, we did not calculate absolute salinities. We used the 18Obenthic record of Site 1241 (Tiedemann et al., this volume) to approximate Pliocene ice volume. The 18Obenthic difference between the Last Glacial Maximum (LGM) and the Holocene for the east Pacific is 1.6 (Shackleton et al., 1983), whereas the difference in global ice volume between the LGM and the Holocene is 1.0–1.2 (Fairbanks, 1989; Schrag et al., 1996), ~75% of the 18Obenthic signal for the east Pacific. We assumed that the relative contributions of ice volume, salinity, and temperature are also valid for the Pliocene. This would mean that the 1.1 amplitude for MIS 100 (Fig. F2D) reflects an ice volume signal of ~0.8, two-thirds of the LGM–Holocene ice volume difference (Raymo et al., 1989). When we take into account that it is likely that the isotopic composition of continental ice sheets has changed significantly since the Pliocene, a sea level change of 40–80 m can be expected for MIS 100 (Miller et al., 1987; Pekar et al., 2002). For our calculations, we used the maximum sea level change of 80 m (10 m sea level = 0.1 18O). The 1.1 amplitude in the 18Obenthic record for MIS 100 (Fig. F2D) was then normalized to an amplitude of 0.8 and accordingly extended to the entire 18Obenthic record. This normalized record was then subtracted from the 18Owater record, leaving a 18Osalinity record, which provides a first approximation of relative changes in local SSS (Fig. F2E). We are aware of the fact that this approach to assess SSS might be an oversimplification, but nevertheless it gives a reliable first approximation.
Prior to 3.7 Ma, the 18Osalinity record remains relatively constant around an average value of 0.55, with the exception of a decrease to 0.4 between 4.4 and 4.2 Ma (Fig. F2E). Between 3.7 and 3.6 Ma, average values decrease from 0.6 to 0.35. Between 3.6 and 2.7 Ma average values further decrease to 0.2, followed by a more pronounced decrease to –0.25 until 2.4 Ma. Spectral analysis of the 18Osalinity record indicates no power in the 41-k.y. frequency band, suggesting a decoupling from the global signal of the intensification of the NHG and high-latitude climate forcing (Fig. F4D). A weak precession signal for the entire time period indicates that local SSS is mainly controlled by low-latitude variations in the precipitation-evaporation budget instead of global, high-latitude forcing (Figs. F4D, F5D). This is corroborated by Benway and Mix (2004), who showed that surface water salinities in the Panama Bight are primarily determined by variations in rainfall, both originating from local sources as from the Caribbean.
