Solar forcing of Florida Straits surface salinity during the early Holocene

Authors


Abstract

[1] Previous studies showed that sea surface salinity (SSS) in the Florida Straits as well as Florida Current transport covaried with changes in North Atlantic climate over the past two millennia. However, little is known about earlier Holocene hydrographic variability in the Florida Straits. Here, we combine Mg/Ca-paleothermometry and stable oxygen isotope measurements on the planktonic foraminifera Globigerinoides ruber (white variety) from Florida Straits sediment core KNR166–2 JPC 51 (24° 24.70′ N, 83° 13.14′ W, 198 m deep) to reconstruct a high-resolution (∼25 yr/sample) early to mid Holocene record of sea surface temperature and δ18OSW (a proxy for SSS) variability. After removing the influence of global δ18OSW change due to continental ice volume variability, we find that early Holocene SSS enrichments are associated with increased evaporation/precipitation ratios in the Florida Straits during periods of reduced solar forcing, increased ice rafted debris in the North Atlantic and the development of more permanent El Niño–like conditions in the eastern equatorial Pacific. When considered with previous high-resolution reconstructions of Holocene tropical atmospheric circulation changes, our results provide evidence that variations in solar forcing over the early Holocene had a significant impact on the global tropical hydrologic cycle.

1. Introduction

[2] There is a critical need to understand the mechanisms involved in climate dynamics over the current interglacial period in order to better predict how climate may evolve over the next few centuries. Prior studies have shown that the North Atlantic has experienced a number of quasiperiodic climate cycles over the last few millennia, the most recent being the Little Ice Age (LIA) from ∼150 to 450 yrs BP and the Medieval Warm Period (MWP) from ∼700 to 1500 yr BP [deMenocal et al., 2000; Lund et al., 2006; Richey et al., 2009]. However, very little is known about submillennial-scale climate oscillations during earlier parts of the Holocene. In addition, the driver of these Holocene climate oscillations is still unknown. While some researchers argue that external forcing (solar variability) is responsible for Holocene climate cycles [Marchitto et al., 2010], others argue for an internal mechanism [Cane and Clement, 1999; Clement and Cane, 1999].

[3] The Florida Current makes up a major component of the northward flowing Gulf Stream system, and thus forms an important link between the tropics and the high-latitude North Atlantic. Lund and Curry [2006] showed that climate variability in the Florida Straits is linked to high-latitude climate change and Lund et al. [2006] found that Florida Current transport may be coupled to changes in North Atlantic Meridional Overturning Circulation (AMOC) variability over the past two millennia. In particular, the period of reduced AMOC and cooler climate in the North Atlantic during the LIA was associated with drier conditions in the Florida Straits as the Hadley circulation cells and the Intertropical Convergence Zone (ITCZ) shifted southward of its modern position.

[4] To determine if the Florida Straits experienced millennial- and centennial-scale climate oscillations similar to the LIA during the early Holocene, we generated a high-resolution record of early Holocene sea surface temperature (SST) and sea surface salinity (SSS) variability using a sediment core recovered from the Florida Margin of the Florida Straits, KNR166–2 JPC 51 (24° 24.70′ N, 83° 13.14′ W, 198 m deep) (Figure 1). By combining Mg/Ca-paleothermometry and stable oxygen isotope measurements on the planktonic foraminifera Globigerinoides ruber (white variety), we reconstruct a high-resolution (∼25 yr/sample) early to mid Holocene record (9.1 to 6.2 kyr) and a lower resolution (∼150 yr/sample) mid to late Holocene record (6.2 to 1.4 kyr) of δ18OSW variability in the Florida Straits. After correcting the δ18OSW record for global changes in δ18OSW due to continental ice volume variability, the ice volume free δ18OSW record can be used as a proxy for past changes in SSS.

Figure 1.

The location of KNR166–2 JPC 51 (24° 24.70′ N, 83° 13.14′ W, 198 m deep) on the western margin of the Florida Straits and modern annual sea surface salinity in western tropical Atlantic and the Gulf of Mexico [Antonov et al., 2010]. As discussed in the text, core KNR166–2 79 GGC is located at nearly the same location as JPC 51. Also noted is the location of the Orca Basin, where core MD02–2550 was recovered in the northern Gulf of Mexico. Black arrows indicate the major ocean currents in the region.

2. Oceanographic Setting

[5] Waters from the Caribbean are directly connected with the Florida Straits via flow through the Yucatan Channel [Maul and Vukovich, 1993; Murphy et al., 1999]. As such, waters in the Florida Straits are predominantly characteristic of the Caribbean [Brunner, 1982; Lynch-Stieglitz et al., 1999] and form an important link between waters of the Caribbean, Gulf of Mexico and North Atlantic (Figure 1).

[6] The modern seasonal SST cycle in the Florida Straits varies from 25.0 to 26.0°C from December through March to 29.0 to 29.4°C from July to mid-September, with a modern average annual SST of 27.5°C [Locarnini et al., 2006]. Mesoscale cyclonic eddies form near the Dry Tortugas during periods of strong Loop Current development [Fratantoni et al., 1998]. These frontal eddies associated with the Loop Current can persist for about 100 days and their presence acts to cool SST in the Florida Straits during winter and spring when surface temperatures in the Gulf of Mexico are cooler than in the Florida Current [Fratantoni et al., 1998].

[7] In the modern tropical Atlantic and Caribbean region, seasonal rainfall variability is primarily controlled by the annual migration of the ITCZ between 15°N and 5°S [Waliser and Gautier, 1993]. Therefore, evaporation/precipitation (E/P) ratios decrease during the boreal summer months when the ITCZ is located farthest to the north and increase during the cool, dry season as the ITCZ migrates southward during boreal winter [Stidd, 1967]. Modern annual SSS in the Florida Straits varies from 36.1 to 36.2 from January to June to ∼35.9 from August to December, with a modern average annual SSS of ∼36.1 [Antonov et al., 2006].

3. Materials and Methods

3.1. Age Model Development

[8] To develop an age model for JPC 51 across the Holocene, seven intervals were sampled for radiocarbon analysis (Table 1 and Figure S1 in the auxiliary material). Five of the intervals chosen for radiocarbon analysis span the high-resolution section of our early Holocene record. Radiocarbon analyses were conducted using 3.4 to 5.5 mg of G. ruber and Globigerinoides sacculifer shells collected from the >355 μm size fraction. These samples were analyzed at the National Ocean Sciences Accelerator Mass Spectrometry Facility of the Woods Hole Oceanographic Institution in Massachusetts. Raw 14C ages were converted to calendar age using CALIB 6.0 using the standard reservoir age of 400 years for surface waters (Table 1) (M. Stuiver et al., Calib calibration program, version 6.0, 2011). Linear interpolation between 14C-dated control points yields Holocene sedimentation rates in JPC 51 ranging from 60 to 188 cm/kyr. The upper 295 cm of JPC 51 corresponds to the time interval between 6.21 to 1.48 kyr and the lower 254.25 cm corresponds to the time interval between 9.16 to 6.21 kyr. Based on our age model, the core has maximum sedimentation rates of 141 to 188 cm/kyr from 8.0 to 7.3 kyr. Sedimentation rates range from 60 to 83 cm/kyr in the section of the core corresponding to 7.3 to 1.4 kyr. The core was sampled every 2 cm from 554.5 to 300.5 cm (corresponding to our high-resolution time interval spanning the early Holocene from 9.16–6.21 kyr) and every 8 cm from 296.5 cm to the core top.

Table 1. AMS 14C and Calendar Ages
CoreDepth (cm)CAMS Number14C AgeError (years)Calendar Age (kyr B.P.)Error (years)
JPC 515.25400171940401.49110
JPC 51159.25490094040454.07146
JPC 51312.25400185970606.40131
JPC 51394.5760766880407.3987
JPC 51480.25400197520707.99152
JPC 51494.5760777610358.0790
JPC 51632.254002096907510.56194

3.2. Stable Isotope Analysis

[9] Sediment from each core interval was dried overnight at ∼50°C, then weighed and disaggregated in ultra clean water for 6 h on a shaker table. To collect the coarse fraction, samples were wet-sieved using a 63 μm mesh. For each interval, 80 individual G. ruber (white variety) specimens were collected from the 255 to 350 μm size fraction. The 255 to 350 μm size fraction limitation was used to minimize ontogenetic and growth rate effects on shell geochemistry [Lea et al., 2000; Spero et al., 2003]. To ensure an unbiased estimation of the average δ18O value for a given interval, we used 17 to 25 G. ruber shells for each individual stable isotope analysis. Samples were sonicated in methanol for 6 s, then crushed and homogenized prior to analysis. A 100 to 150 μg split of the sample was then analyzed for stable isotopes in J. Lynch-Stieglitz's laboratory at the Georgia Institute of Technology on a Finnigan MAT253 stable isotope ratio mass spectrometer with a Kiel Device. Raw δ18O values were standardized using NBS-19 and Venato 690 (an in-house standard).

3.3. Minor and Trace Metal Analysis

[10] For minor and trace metal analysis, 45 to 60 G. ruber shells (∼580 μg) from each core interval were gently crushed between glass plates under a microscope, homogenized and then split. To maintain trace metal clean conditions, the samples were then cleaned according to the procedures of Lea et al. [2000] in a laminar flow clean bench. The cleaning process included sonication in both ultra clean water and methanol to remove clays, a hot water bath in reducing agents to remove metal oxides and a hot water bath in an oxidizing solution to remove organic matter. Finally, the samples were transferred to acid cleaned vials and leached in weak nitric acid. The samples were analyzed in duplicate at Texas A&M University on a Thermo Scientific Element XR High Resolution Inductively Coupled Plasma Mass Spectrometer (HR-ICP-MS) using isotope dilution, as outlined in Lea and Martin [1996]. A suite of elements including Na, Mg, Ca, Sr, Ba, U, Al, Fe and Mn, were analyzed and reported as metal/Ca ratios. All geochemical data is archived at the NOAA National Climate Data Center at: http://www.ncdc.noaa.gov/paleo/paleo.html.

3.4. Calculations

[11] Foraminiferal Mg/Ca:temperature relationships have been derived from a combination of experimental, core top, and sediment trap calibrations [Anand et al., 2003; Dekens et al., 2002; Nürnberg et al., 1996]. Because it is unlikely that dissolution is an issue in the shallow core selected for this study, we utilized the non-depth corrected Mg/Ca:SST calibration of Anand et al. [2003] to estimate calcification temperatures in JPC 51:

display math

This equation was derived from the measurement of Mg/Ca ratios in planktonic foraminifera collected from sediment traps in the Sargasso Sea and modern SST data. The estimated 1σ error on this relationship is between ±0.5 to 1.0°C [Anand et al., 2003]. It is important to note that equation (1) is identical to the previously published Atlantic G. ruber Mg/Ca:SST relationship developed from core tops in Dekens et al. [2002], but without the depth correction term. Given the shallow depth of JPC 51 and the lack of dissolution in the core, we did not feel a depth correction term was needed. Although a recent study by Arbuszewski et al. [2010] suggests that SSS has a major impact on foraminiferal Mg/Ca ratios, we find no evidence of this influence on our reconstructed Mg/Ca-SST record from JPC 51 (see auxiliary material).

[12] The oxygen isotopic composition of foraminiferal calcite is a function of the temperature and the ambient isotopic composition of the seawater (δ18OSW) in which it precipitates its shell. If the temperature component is accounted for, foraminiferal calcite can be used to estimate past changes in SSS because δ18OSW covaries linearly with SSS [Fairbanks et al., 1992]. Based on this concept, numerous studies have developed and refined a multiproxy geochemical approach in which calibrated Mg/Ca ratios from planktonic foraminifera shells are used to estimate past SSS variability [Carlson et al., 2008; Lea et al., 2000; Lund and Curry, 2006; Schmidt et al., 2004; Weldeab et al., 2007].

[13] Lea et al. [2000] and Schmidt et al. [2004] found that the low-light temperature:δ18O relationship determined for Orbulina universa in laboratory culture experiments [Bemis et al., 1998] yields excellent results when applied to fossil G. ruber (white variety) to calculate modern δ18OSW in the equatorial Pacific and Caribbean. Using the Bemis et al. [1998] δ18OC:SST relationship,

display math

δ18OC values and Mg/Ca SST values were combined to calculate δ18OSW.

3.5. Estimating SSS Change From δ18OSW

[14] Changes in continental ice volume also affect global δ18OSW values. If this effect can be accounted for, the resulting ice volume free δ18OSW (δ18OIVF-SW) record can be used to estimate regional salinity change. Sea level has risen by ∼25 m over the last 10 kyr due to the melting of polar ice sheets [Bard et al., 1990; Cutler et al., 2003; Edwards et al., 1993]. Because continental ice is isotopically depleted in δ18O, addition of this meltwater has changed the average global ocean δ18OSW value by ∼0.20 ‰ over the last 10 kyr. In order to correct for this continental ice volume effect, we used a compilation of sea level records for the last 10 kyr [Bard et al., 1990; Cutler et al., 2003; Edwards et al., 1993]. Assuming a one meter increase in sea level change equates to a change of −0.008 ‰ in global δ18OSW [Siddall et al., 2003], we subtracted global changes in δ18OSW from our calculated Florida Straits δ18OSW record, resulting in the regional δ18OIVF-SW record.

3.6. Error Analysis

[15] The analytical precision for the δ18O analyses is less than ±0.07 ‰. The long-term analytical reproducibility of a synthetic, matrix-matched Mg/Ca standard analyzed over the course of this study is ±0.48%, and the pooled standard deviation of the replicate Mg/Ca analyses is ±3.58% (1 SD, 186 degrees of freedom) based on 164 analyzed intervals. When combined with the error on calibration equation (1), the analytical error associated with the Mg/Ca measurements contributes an additional 0.4°C of uncertainty. To calculate the combined ±1σ error on the smoothed SST record on Figures 2b and 3, we used the following equation: math formula (where n is the number of points in the smoothing function). We estimate the ±1σ uncertainty on calculated δ18OSW values to be ∼0.25 ‰ based on the propagation of the ±1σ errors from the Mg/Ca and δ18O analyses along with the reported errors from Anand et al. [2003] for equation (1), and Bemis et al. [1998] for equation (2). Using a variety of methods, prior studies report similar error propagations for the δ18OSW residuals based on δ18OC and Mg/Ca-SSTs in G. ruber, ranging from ±0.18 ‰ to ±0.26 ‰ [Carlson et al., 2008; Lea et al., 2000; Lund and Curry, 2006; Oppo et al., 2009; Schmidt et al., 2004, 2006; Weldeab et al., 2006]. To calculate the ±1σ error on the smoothed δ18OIVF-SW record on Figures 2c and 5a, we used the following equation: math formula (where n is the number of points in the smoothing function).

Figure 2.

(a) G. ruber (white variety) δ18OC (blue line is a weighted 3-point smooth through the raw data points represented as blue circles) and (b) the G. ruber Mg/Ca-SST (red line is a weighted 3-point smooth through the raw data points represented as red circles) records from JPC 51. Measured Mg/Ca ratios (mmol/mol) were converted to SST using the planktonic relationship Mg/Ca = 0.38 exp 0.09(SST) [Anand et al., 2003]. The combined calibration and analytical error (±1σ) on the smoothed record is shown as the black bar near the y axis. (c) Calculated ice volume free δ18OSW (δ18OIVF-SW) record (green line is a weighted 3-point smooth through the raw data points represented as green circles) generated using the Mg/Ca-SST, the following δ18OC:SST relationship: T (°C) = 16.5 − 4.80 (δ18OC − (δ18OSW − 0.27‰)) [Bemis et al., 1998], and by subtracting (d) the global δ18OSW change due to continental ice volume variability based on a compilation of sea level records for the last 10 kyr [Bard et al., 1990; Cutler et al., 2003; Edwards et al., 1993] and the relationship that a one meter increase in sea level change equates to a change of −0.008 ‰ in global δ18OSW values [Siddall et al., 2003]. The estimated ±1σ uncertainty on the smoothed record is shown by the black bar on the y axis. Black triangles on the x axis indicate intervals with calibrated radiocarbon dates and their associated error based on mixed samples of the planktonic foraminifera G. ruber and G. sacculifer.

Figure 3.

Comparison of the Mg/Ca-SST records from the Florida Straits (JPC 51 in red) and the northern Gulf of Mexico's Orca Basin (MD02–2550 in gray) [LoDico et al., 2006] calculated using the Mg/Ca-SST relationship in Anand et al. [2003]. The estimated ±1σ uncertainty on the JPC 51 SST record is shown by the black bar on the y axis. Because the Orca Basin SST record is also based on Mg/Ca ratios in G. ruber and has been calibrated using the same Mg/Ca-SST relationship [Anand et al., 2003], the estimated error on this record is assumed to be similar.

4. Results

4.1. Globigerinoides ruber δ18OC

[16] The G. ruber δ18OC record shows an average Holocene value of −1.67 ‰. Based on the raw data, the δ18OC record reaches a maximum value of −1.23 ‰ at 7.7 kyr and a minimum value of −2.07 ‰ at 8.0 kyr, with a maximum variability of ∼0.8 ‰ (Figure 2a). Throughout the high-resolution section of the record during the early Holocene, there are significant enrichments in δ18O at ∼8.5, 7.7, and 6.6 kyr and depletions at ∼8.0, 7.4, 6.8, 6.5, and 6.4 kyr. The lower-resolution mid to late Holocene record shows an enrichment at ∼4.3 kyr that is followed within 100 years by an abrupt depletion in δ18O. Lund and Curry [2006] measured δ18O in G. ruber from the late Holocene in a core located near JPC 51. Their record showed millennial-scale climate oscillations of about ∼0.7 ‰ in oxygen isotope variation during the late Holocene, similar to what we find for the early Holocene. Therefore, oxygen isotope records from the Florida Margin of the Florida Straits derived from G. ruber suggest the presence of persistent millennial-scale climate cycles over the past 10,000 years.

4.2. Mg/Ca-SST

[17] The Mg/Ca-SST record indicates a core top temperature of 27.3°C, in good agreement with the modern average annual SST of ∼27.5°C in the Florida Straits [Locarnini et al., 2006] (Figure 2b). The early Holocene section of the record is marked by a warm interval from 8.1 to 7.9 kyr when temperatures reached 29.2°C. Following this warm interval, the record shows a gradual cooling trend lasting ∼1 kyr with the coolest temperatures of 27.2°C reached at 7.0 to 6.9 kyr. The high-resolution early Holocene record then shows another warming trend with peak temperatures of just less than 29.0°C reached by 6.5 kyr. The lower-resolution part of the record indicates similar temperature fluctuations in the Florida Straits over the last 6 kyr ranging from ∼29°C–27°C.

4.3. Change of δ18OSW

[18] Next, we calculated δ18OSW change using the Mg/Ca-temperatures and the measured δ18OC values in equation (2). Finally, we accounted for global δ18OSW changes due to Holocene sea level rise as described in section 3.5, thereby calculating a local record of δ18OIVF-SW (Figure 2c). The core top δ18OIVF-SW value is 0.91 ‰, in good agreement with the estimated modern average δ18OSW value in the Florida Straits of 0.95 ‰ (calculated using the modern annual mean SSS and the regional SSS:δ18OSW relationship shown below in equation (3)). During the early-mid Holocene, the record shows three clearly defined cycles with δ18OIVF-SW values ranging from 0.6 ‰ to 1.4 ‰, resulting in a maximum amplitude of about 0.8 ‰ across each of these early Holocene events. The three periods of elevated early Holocene δ18OIVF-SW values are from 8.3 to 8.0 kyr, 7.6 to 7.4 kyr and 6.6 to 6.3 kyr. In the lower resolution portion of the record, there are two prominent enrichments at 4.2 and 1.8 kyr.

4.4. Conversion of δ18OIVF-SW to SSS

[19] Using the modern tropical Atlantic δ18OSW:SSS relationship calculated from regional SSS and δ18OSW data in the global database of Schmidt et al. [1999],

display math

our core top δ18OIVF-SW value corresponds to a salinity of 35.8. This agrees well with the modern average annual salinity of ∼36.1 for the Florida Straits [Antonov et al., 2006]. Based on this modern regression, a change of 0.26 ‰ in δ18OSW is equivalent to a change in SSS of 1.0. Therefore, the calculated δ18OIVF-SW values from JPC 51 suggest Holocene SSS changes ranging from 34.4 to 38.2 with an average of 36.2. Given that this range in SSS seems too large for realistic Holocene SSS variability in the Florida Straits, it is likely that the slope of the δ18OSW:SSS relationship must have changed over the Holocene. As discussed in Lund et al. [2006], it is also possible that thermocline waters influenced the entire tropical/subtropical δ18OSW:SSS relationship during the Holocene. Because thermocline waters are ventilated at high latitudes, the δ18OSW:SSS slope is steeper (0.5 ‰ per unit salinity change) than the tropical relationship. If this were the case, then δ18OSW changes of 0.5 ‰ would equate to salinity changes of 1.0, thus reducing our estimated SSS change in JPC 51 to <2.0 across the Holocene.

[20] Based on the results of a coupled GCM modeling study, Oppo et al. [2007] found significant tropical hydrologic cycle variability across the Holocene associated with orbital changes in solar insolation. Their results suggest a decrease in water vapor transport across the Central American Isthmus during the mid Holocene associated with a northward shift in the position of the ITCZ and a decrease in the strength of the Pacific easterlies. Because the trade winds transport isotopically depleted water vapor from the Atlantic to the Pacific, modeling results suggest that a decrease in the net water vapor transport across the Central American Isthmus results in a decrease in δ18OPRECIPITATION values in the western tropical Atlantic during the mid Holocene [Oppo et al., 2007]. More negative δ18OPRECIPITATION values in the circum-Caribbean region would result in a steeper δ18OSW:SSS relationship, thus reducing the magnitude of SSS change estimated from our δ18OIVF-SW record for the early to mid Holocene. Therefore, SSS estimates calculated from our mid to early Holocene δ18OIVF-SW record should only be viewed as estimates.

5. Discussion

5.1. Holocene SST Record

[21] Today, water flows more directly from the Yucatan Channel into the Florida Straits during periods of reduced Loop Current penetration [Lee et al., 1995]. Periods of reduced Loop Current penetration also result in the formation of fewer Tortugas eddies [Lee et al., 1995]. This results in a shift in the axis of the Florida Current northward and warmer SSTs at our study site due to an increase in the component of warm Caribbean surface water. Although instrumental data suggests that eddy formation today is stochastic and unrelated to climate forcing [Maul and Vukovich, 1993; Sturges and Leben, 2000; Vukovich, 1988], significant changes in Loop Current penetration and Tortugas eddy formation over the Holocene could impact SSTs at our site. A reconstruction of Loop Current penetration into the Gulf of Mexico based on foraminiferal faunal changes in a core from the northern Gulf of Mexico showed a gradual increase in Loop Current strength from about 9 to 6 kyr, followed by a gradual decrease into the modern [Poore et al., 2004; Poore et al., 2003]. This mid-Holocene peak in reconstructed Loop Current strength at 6 kyr is not reflected as a cooling in our Mg/Ca-SST record from JPC 51 (Figure 2b). Instead, the JPC 51 SST record indicates one of the warmest periods in our record centered at ∼6.0 kyr. Therefore, we do not find evidence that SSTs at our study site are significantly influenced by changes in Tortugas eddy formation during the Holocene.

[22] LoDico et al. [2006] published a high-resolution (∼30 yr sample spacing) SST reconstruction from the northern Gulf of Mexico's Orca Basin spanning from 10.5 to 7.0 kyr based on Mg/Ca ratios in G. ruber (white variety). LoDico et al. [2006] did not use the reductive cleaning step in their methods. Because the reductive cleaning step in the Mg/Ca cleaning protocol has been shown to lower shell Mg/Ca ratios by 10% [Pena et al., 2005], we decreased their reported Mg/Ca values by 10% to make them more suitable for comparison to the results in this study. We also updated the age model for the LoDico et al. [2006] record by recalibrating their radiocarbon dates using CALIB 6.0 (Figure 3).

[23] The LoDico et al. [2006] SST record indicates an abrupt warming at 9.4 kyr, followed by millennial-scale oscillations of ∼1.0°C (gray curve on Figure 3). In contrast, the Florida Straits SST record indicates warmer conditions during the early Holocene from 9.2 to 7.8 kyr when the two records overlap. Similar SSTs are then recorded at both sites from 7.8 to 7.3 kyr. While the Florida Straits SST record shows early Holocene centennial-scale SST cycles that are similar in magnitude to those reconstructed for the Orca Basin, the timing of these oscillations is not the same, suggesting unique forcing mechanisms on SSTs at each site.

[24] The greatest difference between the Florida Straits and Orca Basin SST records occurs between ∼8.8 to 8.6 kyr and from ∼8.2 to 7.8 kyr. It is possible that the different temperature evolution between the two sites is due to the fact that the northern Gulf of Mexico is more directly influenced by continental North American climate, whereas hydrographic conditions in the Florida Straits are more influenced by waters from the western tropical Atlantic [Schmitz and Richardson, 1991]. This may explain why the Orca Basin records a cooling after about 8.2 kyr, just as SSTs in the Florida Straits indicate a warming trend over this same interval. The 8.2 kyr event is thought to be the largest climate anomaly of the entire Holocene [Alley and Agustsdottir, 2005]. It is possible that the high-latitude cooling associated with the 8.2 kyr event and the proposed reduction in AMOC that lasted several hundred years after the event [Ellison et al., 2006] resulted in cooler surface temperatures over the North American continent, but had less of an impact in the tropical Atlantic. The significant cooling recorded in the Orca Basin after 8.2 kyr most likely reflects the influence of cooler temperatures over continental North America, while the corresponding warmer SSTs in the Florida Straits suggest the tropics may have been less affected by the high-latitude cooling.

5.2. Holocene δ18OIVF-SW and Salinity Variability

[25] First, we compare the JPC 51 early Holocene record of δ18OIVF-SW with the previously published late Holocene δ18OIVF-SW record from Florida Straits core 79 GGC [Lund and Curry, 2006] (Figure 4a). Core 79 GGC is located near JPC 51 and the δ18OIVF-SW record was calculated using the same methodologies and calibration equations. Comparison of the two records shows that both the early and late Holocene were characterized by similar magnitude δ18OIVF-SW oscillations with almost the same average value (the two red curves in Figure 4a). In their study, Lund and Curry [2006] observed fresher surface conditions in the Florida Straits during the MWP and saltier surface conditions during the LIA. They argued that the development of elevated δ18OSW values (and higher SSS) in the Florida Straits during the LIA resulted from increased E/P ratios in the northern tropical Atlantic associated with a southward shift in the Hadley cell circulation and the ITCZ, as reflected in the percent titanium record from the Cariaco Basin [Haug et al., 2001] (Figure 4b).

Figure 4.

(a) Comparison of δ18OIVF-SW records from Florida Straits cores 79 GGC (red, late Holocene) [Lund and Curry, 2004] and JPC 51 (red, early Holocene) with the δ18OIVF-SW record from Orca Basin core MD02–2550 (gray, early Holocene) [LoDico et al., 2006]. (b) Record of percent titanium in Cariaco Basin core ODP 1002 [Haug et al., 2001], indicating more arid conditions and a southerly shift in the ITCZ during intervals of reduced % Ti values. The % Ti values remain elevated during the early Holocene, suggesting a more permanent northward position of the ITCZ and wetter conditions over northern Venezuela at this time [Haug et al., 2001]. (c) Local spring and summer insolation values at 24°N calculated from Paillard et al. [1996].

[26] In order to determine if meridional shifts in the ITCZ correspond with early Holocene δ18OIVF-SW variability, we compare our new JPC 51 record with the Cariaco Basin titanium record (Figures 4a and 4b). The titanium record from the Cariaco Basin was interpreted to suggest that the southern Caribbean was characterized by an enhanced hydrologic cycle during the early Holocene [Haug et al., 2001], a period marked by increased spring-summer insolation when precession was at a maximum in the northern tropics (Figure 4c). Ostracod δ18O records from Haitian lake cores also indicate wetter conditions in the northern Caribbean during the early Holocene [Hodell et al., 1991]. Therefore, rainfall amounts in the circum-Caribbean region were most likely at a maximum during the early Holocene, probably reflecting a more intense summer ITCZ located farther to the north. Although the JPC 51 record indicates large δ18OIVF-SW oscillations during the early Holocene that are similar in magnitude with those recorded for the MWP – LIA transition of the late Holocene [79 GGC record on Figure 4a from Lund and Curry, 2006], the JPC 51 δ18OIVF-SW oscillations from 9.2 to 6.2 kyr do not correlate with changes in the Cariaco Basin percent titanium record during this interval (R = 0.20, R2 = 0.04). Therefore, early Holocene δ18OIVF-SW cycles in the Florida Straits cannot simply be explained by meridional shifts in the ITCZ.

[27] Second, we compare our early Holocene δ18OIVF-SW record with a similar resolution δ18OIVF-SW reconstruction from the northern Gulf of Mexico's Orca Basin (gray curve on Figure 4a) [LoDico et al., 2006]. In order to directly compare the JPC 51 δ18OIVF-SW reconstruction with the Orca Basin record, we recalculated their data using equations (1) and (2) and then corrected the resulting δ18OSW values for continental ice volume variability using the same global δ18OSW we used to calculate our δ18OIVF-SW values. Overall, the calculated δ18OIVF-SW values from both locations show similar scale oscillations, most notably between 9.2 to 8.6 kyr and 7.8 to 7.1 kyr. However, the Orca Basin δ18OIVF-SW record indicates two significant freshening events between 8.4 and 8.0 kyr that are not recorded in the JPC 51 record. Just as the SST comparison between the Orca Basin and the Florida Straits indicates the development of a steep temperature gradient across the Gulf of Mexico after the 8.2 kyr event (Figure 3), the δ18OIVF-SW comparison suggests large differences in SSS between the two sites as well. The cool, fresh conditions in the northern Gulf of Mexico associated with the 8.2 kyr event apparently did not extend into the Florida Straits.

[28] Several studies suggest the primary driver of early Holocene climate change on centennial to millennial time scales is variability in solar forcing [Bond et al., 1997, 2001; Lund and Curry, 2004; Roth and Reijmer, 2005]. Bond et al. [2001] found increases in ice rafted debris (IRD) in North Atlantic sediments about every 1500 years and showed that intervals associated with increased IRD correspond to periods of elevated 14C production in the atmosphere. More 14C is produced in the atmosphere during periods of reduced solar winds, so periods of increased IRD correspond to intervals of reduced solar forcing (Figures 5d, 5e, and 5f).

Figure 5.

(a) Early Holocene δ18OIVF-SW from JPC 51 smoothed with a 250-year running mean. The estimated ±1σ uncertainty on the smoothed record is shown by the black bar on the y axis. (b) Planktonic foraminiferal Mg/Ca-SST anomalies from the Soledad Basin (Baja California Sur, Mexico) smoothed with a 250-year running mean [Marchitto et al., 2010]. SST changes in the Soledad Basin are interpreted to reflect shifts in the mean ENSO state in the tropical Pacific, with warmer temperatures associated with periods characterized by more permanent El Niño-like conditions [Marchitto et al., 2010]. The age model for the Soledad Basin record is based on four calibrated radiocarbon dates across this interval. (c) Stalagmite δ18O record from Dongge Cave (southern China) indicating changes in the strength of the Asian monsoon [Wang et al., 2005]. The age model for the Dongge Cave record is based on absolute-dated U-Th ages. (d) Stacked ice-rafted debris record from the North Atlantic [Bond et al., 2001] indicating times of cooler North Atlantic climate when IRD was greater. (e) IntCal04 14C production rate based on tree ring-derived Δ14C [Reimer et al., 2004] and (f) 10Be flux from the GRIP-GISP2 ice cores [Finkel and Nishiizumi, 1997; Vonmoos et al., 2006]. Note the development of more El Niño-like conditions in the eastern equatorial Pacific and saltier conditions in the Florida Straits during times of reduced solar output. Also indicated with the shaded gray bar is the 8.2 kyr event. Note the 8.2 kyr event occurs at a time of increasing solar activity, but that SSS remained elevated in the Florida Straits and SSTs in the EEP remain warm. This suggests the cool conditions in the North Atlantic caused by freshwater forcing prevented the tropics from responding to the increase in solar activity at this time. Black triangles on the x axis indicate intervals with calibrated radiocarbon dates in JPC 51 (see Table 1 for associated errors on the radiocarbon dates).

[29] In addition, several ocean-atmosphere modeling studies point to the tropical Pacific as playing a major role in driving sub orbital-scale climate variability [Bush and Philander, 1998; Cane and Clement, 1999; Cane, 1998; Clement and Cane, 1999; Kukla et al., 2002]. These studies suggest that the El Niño-Southern Oscillation (ENSO) is controlled by changes in seasonal cycle forcing at the equator, correlating periods of reduced summer irradiance in the tropical Pacific (decreased seasonality) with stronger El Niño forcing. According to the Zebiak-Cane model of ENSO dynamics, relatively small changes in total irradiance can cause the background state of the tropical Pacific to oscillate between El Niño/La Niña states [Mann et al., 2005].

[30] A recent study by Marchitto et al. [2010] showed that changes in solar forcing during the early Holocene had a large impact on ENSO variability in the eastern equatorial Pacific (EEP). Marchitto et al. [2010] reconstructed Holocene SST anomalies off the coast of Baja California Sur, Mexico in the Soledad Basin using Mg/Ca ratios in the planktonic foraminifera Globigerina bulloides. Marchitto et al. [2010] first smoothed their Mg/Ca-SST anomaly record using a 250-year running mean to remove high-frequency climate variability that cannot be reliably correlated with other proxy records (Figure 5b). They then smoothed the Holocene records of 14C [Reimer et al., 2004] and 10Be production [Finkel and Nishiizumi, 1997; Vonmoos et al., 2006] using the same 250-year running mean and then removed the long-term drift in the nuclide records resulting from slow variations in the Earth's geomagnetic field [Wagner et al., 2000] by performing a high pass filter at 1/1800-year (Figures 5e and 5f). This same high pass filter was used by Bond et al. [2001] to generate their stacked IRD record shown on Figure 5d. As a result, Marchitto et al. [2010] identified five warm intervals between 11 to 7 kyr (more El Niño-like conditions) separated by roughly 1 kyr that corresponded to periods of increased 14C and 10Be production (times of reduced solar activity). Marchitto et al. [2010] argued that centennial-scale changes in solar output resulted in stronger El Niño forcing in the tropical Pacific during times of reduced solar output. Comparison of the Soledad Basin SST record with Bond et al.'s [2001] IRD record also suggests a strong correlation between cool periods in the North Atlantic and more El Niño-like conditions in the EEP. Previous research also showed that the Asian monsoon system weakens during periods of reduced solar output (Figure 5c) [Wang et al., 2005]. Comparison of the Dongge Cave (southern China) oxygen isotope record with the Soledad Basin SST anomaly record shows that a weaker Asian monsoon correlates with stronger El Niño forcing in the tropical Pacific [Marchitto et al., 2010].

[31] In order to determine how early Holocene ENSO variability in the Pacific may have affected hydrologic changes in the western tropical Atlantic, we compare the Soledad Basin Mg/Ca-SST anomaly record (Figure 5b) with our new JPC 51 δ18OIVF-SW record from the Florida Straits (also smoothed with a 250-year running mean) (Figure 5a). As warm anomalies developed in the EEP associated with a shift to more permanent El Niño-like conditions, δ18OIVF-SW values in the Florida Straits increased, suggesting an increase in tropical Atlantic E/P ratios and more arid conditions. The strong correlation between these two records suggests that ENSO variability in the EEP is linked to hydrologic changes in the western tropical Atlantic. It is not surprising that more permanent El Niño-like conditions in the EEP are associated with a more arid circum-Caribbean climate. El Niño events are associated with reduced rainfall, warmer SSTs, and weaker trade winds in the western tropical North Atlantic [Alexander and Scott, 2002; Alexander et al., 2002; Giannini et al., 2001a, 2001b; Poveda and Mesa, 1997]. Based on coupled ocean-atmosphere general circulation model results [Schmittner and Clement, 2002; Schmittner et al., 2000] and reanalysis of two historical data sets [Schmittner et al., 2000], El Niño events may even result in enhanced water vapor transport out of the tropical Atlantic.

[32] Furthermore, proxies for solar variability also covary with the JPC 51 δ18OIVF-SW record. Intervals of increased 14C and 10Be production (Figures 5e and 5f) correspond to intervals of more positive δ18OIVF-SW values in the Florida Straits. In addition, periods of increased IRD also correlate to elevated δ18OIVF-SW values in the Florida Straits, indicating that cool periods in the high-latitude North Atlantic are associated with drier conditions in the tropical North Atlantic during the early Holocene. Finally, drier conditions in the Florida Straits correlate to increased oxygen isotope values in speleothem records from Dongge Cave, indicating a weakening of the Asian monsoon (Figure 5c). Taken together, correlations between the Florida Straits δ18OIVF-SW record, the Soledad Basin Mg/Ca-SST anomaly record from the EEP, IRD changes in the North Atlantic and variations in the strength of the Asian monsoon all suggest dramatic reorganizations of atmospheric circulation patterns around the globe driven by variability in solar forcing during the early Holocene.

[33] Although it is thought that increased solar irradiance generates a stronger interplanetary magnetic field that shields Earth from the cosmic rays that form 14C and 10Be in the atmosphere, this relationship is not well understood over long time periods. Atmospheric 14C content can also be influenced by the global carbon cycle [Hughen et al., 1998] and local climate variability can affect the 10Be flux to ice sheets [Finkel and Nishiizumi, 1997]. Furthermore, the atmospheric concentration of both nuclides is influenced by long-term changes in Earth's geomagnetic field. Nevertheless, Marchitto et al. [2010] argued that long-term changes in Earth's geomagnetic field can be filtered from each nuclide record using a high-pass filter. Therefore, shared variability between the high-pass filtered 14C and 10Be records shown on Figure 5 (filtered data from Marchitto et al. [2010] suggests that changes in solar irradiance are the most likely cause [Marchitto et al., 2010].

[34] Although we cannot determine whether forcing from the tropical Pacific (ENSO) or the high-latitude North Atlantic had the strongest impact on the tropical Atlantic hydrologic cycle, the strong cooling in the North Atlantic associated with the 8.2 kyr event seems to have locked the tropical hydrologic cycle into a ‘cold phase.’ During the 8.2 kyr event, the final drainage of large pro-glacial lakes into the North Atlantic is thought to have resulted in a meltwater-induced reduction in AMOC that caused widespread cooling in the circum-Atlantic region [Alley and Agustsdottir, 2005; Barber et al., 1999; Clarke et al., 2004; Ellison et al., 2006]. Although the JPC 51 δ18OIVF-SW record covaries with changes in solar forcing across two complete cycles (∼7.4 to 6.2 kyr and 8.0 to 7.4 kyr), the correlation decreases during the interval around the 8.2 kyr event (gray bar on Figure 5). Although the IRD record does not indicate cold conditions in the North Atlantic during the 8.2 kyr event, this record is relatively low-resolution and may not fully resolve the 8.2 kyr event. Nevertheless, several other proxy studies indicate the high-latitude North Atlantic remained cold during the 8.2 kyr event [Alley and Agustsdottir, 2005; Barber et al., 1999; Clarke et al., 2004; Ellison et al., 2006]. Because the 8.2 kyr event was forced by an internal driver rather than by external changes in solar variability, it is not surprising that the JPC 51 δ18OIVF-SW record does not match the solar output proxies at this time. Instead, the JPC 51 δ18OIVF-SW record shows a much better correlation with the Soledad Basin SST-anomalies and the Dongge Cave record during the 8.2 kyr event because these records probably reflect a teleconnected response to the high-latitude cooling. The 8.2 kyr event occurred at a time of increasing solar activity, but SSS remained elevated in the Florida Straits, SST's in the EEP remained warm and the Asian monsoon remained weak. This suggests the cool conditions in the North Atlantic caused by freshwater forcing prevented the tropics from responding to the increase in solar activity at this time.

[35] Saenger et al. [2009] used a coupled GCM to isolate the impact of high-latitude cooling on the tropical Atlantic hydrologic cycle. Their results show that a 2°C cooling in the North Atlantic results in increased wind stress and negative precipitation anomalies in the tropical North Atlantic. The high-latitude cooling increases the strength of the northeast trade winds, resulting in increased evaporation rates in the tropical Atlantic. Therefore, the elevated δ18OIVF-SW values in JPC 51 associated with increased North Atlantic IRD (and a cooler North Atlantic) are also consistent with this high-to-low latitude climate teleconnection, reflecting increased E/P ratios in the Florida Straits during periods of high-latitude cooling.

5.3. Spectral Analysis

[36] To determine the frequency of sub-orbital cycles in our high-resolution early Holocene δ18OIVF-SW record, we performed multitaper spectral analysis [Ghil et al., 2002] on our record from 9.1 to 6.2 kyr using a 25-year interpolation of the data (the average sampling resolution for our early Holocene record is 25 years) (Figure 6). The results of the spectral analysis indicate strong spectral power above the 95% confidence level at periodicities of ∼1,500, 93, 87 and 60-years. Spectral power at the 1500-year period suggests a strong correlation between solar-driven Bond cycles and hydrologic cycle variability in the Florida Straits. Nevertheless, the length of our record limits our confidence in the peak at 1500 years. Although uncertainty in our age model (up to 194 years during the early Holocene) also limits our confidence in the higher frequency peaks, it is worth noting that spectral power at the 87-year period is common in proxies forced by solar variability, possibly reflecting an influence of the Gleissberg solar cycle [Peristykh and Damon, 2003]. Spectral power at the 60-year period may reveal a connection between SSS variability in the Florida Straits and the Atlantic Multidecadal Oscillation (AMO) (30–80 year period) [Dima and Lohmann, 2007], which some modeling studies suggest is the result of high-frequency AMOC variability during the Holocene [Delworth and Mann, 2000; Enfield et al., 2001; Heslop and Paul, 2011]. Knight et al. [2006] showed that the warm AMO phase is associated with a northward displacement of the ITCZ over the tropical Atlantic, suggesting a possible influence of the AMO on the tropical Atlantic hydrologic cycle.

Figure 6.

Multitaper spectral analysis results of the high-resolution JPC 51 δ18OIVF-SW record from from 9.1 to 6.2 kyr showing high spectral power above the 95% confidence level at periods of ∼1500, 93, 87 and 60 years.

6. Conclusions

[37] Our new Mg/Ca-SST record from the Florida Straits indicates only a small (∼0.8°C) cooling trend through the Holocene. While the Florida Straits record shows early Holocene centennial-scale SST cycles that are similar in magnitude to those previously reconstructed for the northern Gulf of Mexico in the Orca Basin over the same time interval, the timing of these oscillations is not the same. The greatest difference between the Florida Straits and Orca Basin SST records occurs from ∼8.8 to 8.6 kyr and from ∼8.2 to 7.8 kyr. We conclude that the contrasting temperature evolution between the two sites is due to the fact that the northern Gulf of Mexico is more directly influenced by continental North American climate, whereas hydrographic conditions in the Florida Straits are more influenced by the waters from the western tropical Atlantic traveling through the Caribbean and Yucatan Channel to our core site.

[38] After correcting our calculated δ18OSW record for global changes in δ18OSW due to continental ice volume variability, the δ18OIVF-SW record from the Florida Straits reveals millennial-scale oscillations on the order of 0.8 ‰. Comparison of our early Holocene δ18OIVF-SW record with the late Holocene δ18OIVF-SW reconstruction from the same location in the Florida Straits [Lund and Curry, 2006] shows that both periods were characterized by similar magnitude δ18OIVF-SW variability. The periods of increased SSS (more positive δ18OIVF-SW values) during the early Holocene are associated with North Atlantic cooling and the development of El Niño-like conditions in the EEP. As solar output decreased on millennial time scales during the early Holocene, proxy records from the EEP suggest the development of more permanent El Niño-like conditions in the tropical Pacific, and Chinese speleothem records indicate intervals of a weakened Asian monsoon. At the same time, these events are associated with elevated SSS in the Florida Straits, providing evidence for a dramatic reorganization of atmospheric circulation patterns around the globe driven by Holocene changes in solar output.

[39] Finally, we show that the strong cooling in the North Atlantic associated with the 8.2 kyr event seems to have locked the tropical hydrologic cycle into a ‘cold phase.’ Because the 8.2 kyr event was forced by an internal driver rather than by changes in solar variability, the JPC 51 δ18OIVF-SW record reflects a teleconnected response to the high-latitude cooling at 8.2 kyr. Although nuclide records suggest solar irradiance was increasing at the start of the 8.2 kyr event, SSS remained elevated in the Florida Straits for several hundred years after the meltwater-induced collapse of AMOC. This suggests a strong coupling between high-latitude North Atlantic climate and the tropical Atlantic hydrologic cycle.

Acknowledgments

[40] We thank the National Science Foundation (grant OCE-0823498 (MWS)) for supporting this research. We also thank Jennifer Hertzberg for input on the manuscript and for technical help.