SEARCH

SEARCH BY CITATION

Keywords:

  • Holocene;
  • diatoms;
  • sea surface temperature;
  • solar insolation;
  • Norwegian Atlantic Current;
  • climate change

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] A high-resolution sediment core from the Vøring Plateau has been studied to document the centennial to millennial variability of the surface water conditions during the Holocene Climate Optimum (HCO) and the late Holocene period (LHP) in order to evaluate the effects of solar insolation on surface ocean climatology. Quantitative August summer sea surface temperatures (SSSTs) with a time resolution of 2–40 years are reconstructed by using three different diatom transfer function methods. Spectral- and scale-space methods are applied to the records to explore the variability present in the time series at different time scales. The SSST development in core MD95-2011 shows a delayed response to Northern Hemisphere maximum summer insolation at ∼11,000 years B.P. The record shows the maximum SSST of the HCO to be from 7.3 to 8.9 kyr B.P., which implies that the site was located in the regional warm water pool removed from the oceanic fronts and Arctic waters. Superimposed on the general cooling trend are higher-frequency variabilities at time scales of 80–120, 210–320, 320–640, and 640–1280 years. The climate variations at the time scale of 320–640 years are documented both for periods of high and low solar orbital insolation. We found evidence that the submillennial-scale mode of variability (640–900 years) in SSST evident during the LHP is directly associated with varying solar forcing. At the shorter scale of 260–450 years, the SSST during the LHP displays a lagged response to solar forcing with a phase-locked behavior indicating the existence of a feedback mechanism in the climate system triggered by variations in the solar constant as well as the role of the thermal inertia of the ocean. The abruptness of the cooling events in the LHP, especially pronounced during the onsets of the Holocene Cold Period I (approximately 2300 years B.P.) and the Little Ice Age (approximately 550 years B.P.), can be explained by a shutdown of deep convection in the Nordic Seas in response to negative solar insolation anomalies. These cooling events are on the order of 1.5°C.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] The North Atlantic exerts a major influence on European climate via heat flux and ocean atmosphere interactions that are at the heart of the North Atlantic Oscillation (NAO). Northwest Europe and the northern North Atlantic region have a climate which is some 5°C–10°C warmer than the zonal mean, in part because of the huge amount of heat transported northward by the warm North Atlantic Drift (NAD). Yet, this has not always been the case. Paleoclimatic and historical records reveal several occasions since the last glacial maximum when the climate of Europe suffered major cooling events and also show that they occurred within a decade or two (e.g., Little Ice Age, Younger Dryas). Forcing mechanisms suggested for these rapid climate fluctuations have been freshwater forcing [Hald and Hagen, 1998], solar forcing [Bond et al., 2001], and reductions in the production of North Atlantic deep water (NADW) [Bianchi and McCave, 1999; Oppo et al., 2003]. Identifying, quantifying and understanding the forcing mechanisms behind climate variability are of utmost importance for our ability to predict future climate changes.

[3] The purpose of this study is to understand the nature of climate variability during periods with higher or lower solar insolation as a result of orbital variations. Our approach is to assess the typical time scales and magnitude of low-frequency (multidecadal and longer) climate variations of the Norwegian Atlantic Current climatology during the Holocene Climate Optimum (HCO) (9.5–6 kyr B.P.) and the late Holocene period (LHP) (3 kyr B.P. to present). For this study, we generated an August summer sea surface temperature (SSST) record derived from a diatom-based transfer function with a time step of 2–40 years. The SSST time series is derived from core MD95-2011 at the Vøring Plateau. Spectral- and scale-spaced methods are applied to explore the variability present in the time series at different time scales.

2. Physical and Oceanographic Setting

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[4] Core MD95-2011 was retrieved from the Vøring Plateau, which is situated under the direct influence of the Norwegian Atlantic Current (NwAC) in the eastern Norwegian Sea (Figure 1). The NwAC is a poleward extension of the Gulf Stream, and serves as a conduit for warm and saline Atlantic Water (AW) from the NAD entering the Arctic [Hopkins, 1991; Orvik and Skagseth, 2005]. The NwAC splits into two branches in the eastern North Atlantic, which enter the Norwegian Sea close to the eastern coast of Iceland and through the Faroe-Shetland Channel. The western branch is a jet in the Polar Front that tends to feed the interior of the Norwegian Sea, resulting in a southward recirculation toward the Fram Strait. The eastern branch, the Norwegian Atlantic Slope Current (NwASC), is, however, the major link between the Norwegian Sea and the Barents Sea. Variations in the NwASC current is found to be strongly linked to the wind field [Skagseth, 2004; Skagseth et al., 2004; Orvik and Skagseth, 2003], and dependence of these ocean currents on climate renders them well situated as recorders of climate changes. These changes are likely to be of regional importance and therefore may also be registered in records at other sites in the North Atlantic region.

image

Figure 1. Map showing the main surface circulation in the Nordic Seas and the location of core MD95-2011. NAD, North Atlantic Drift; NwAC, Norwegian Atlantic Current; NwASC, Norwegian Atlantic Slope Current; IC, Irminger Current; EGC, East Greenland Current; and EIC, East Icelandic Current. Colors indicate 1994 August SSST [Levitus and Boyer, 1994]. NwAC and NwASC are drawn after Orvik and Skagseth [2005].

Download figure to PowerPoint

3. Material and Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

3.1. Material

[5] Box core JM97-948/2A (66°58.18N, 07°38.36 E) and giant Calypso piston core MD95-2011 (66°58.18 N, 07°38.36 E) (Figure 1) were collected from the Vøring Plateau at 1020 and 1048 m water depth. Box core JM97–948/2A is 31 cm long and serves as a reference for an intact sediment surface, while Calypso core MD95-2011 is 17.49 m long. Previously, the box-core was analyzed for diatoms at 0.5–1 cm intervals [Andersen et al., 2004], providing a time step of 2–20 years. The giant calypso core was previously analyzed at 5 cm intervals through the Holocene [Birks and Koç, 2002; Andersen et al., 2004], and at 1–2 cm intervals between 0.5 and 3 kyr B.P. [Andersen, 2003], and at 1 cm intervals between 9 and 12 kyr B.P. [Berner et al., 2010]. For this study, we have decreased the time step for the HCO by analyzing the giant calypso core continuously every single cm through this interval which yields a time step of 15–40 years for the HCO. We then combined these results with the unpublished results for the last 3 kyr [Andersen, 2003].

3.2. Methods

3.2.1. Chronology

[6] Accelerator mass spectrometry (AMS) dating and lead 210 (210Pb) measurements were used to establish the age model for the cores [Andersen et al., 2004]. A new age model has been reconstructed based on a total of 31 AMS dates (Table 1). All measurements were made on the foraminifer species Neogloboquadrina pachyderma (dex) or N. pachyderma (sin), and the GRIP age of the Vedde ash layer was used [Grönvold et al., 1995]. The dates were calibrated with the MARINE04 calibration curve, with ΔR of 0+/−20 in the Holocene and 300 ± 20 in the pre-Holocene [Hughen et al., 2004; Mangerud et al., 2006]. The box core and the piston core models were developed separately, and the age-depth model uses a mixed-effect regression procedure, which uses the midpoint of the 1σ ranges rather than the intercept as a point estimate of the 14C dates, as described by Heegaard et al. [2005].

Table 1. Radiocarbon Dates and Calibrated Agesa
Core Name/Depth (cm)Lab Number14C Age ± Standard DeviationCalibrated Age (cal years B.P.)
  • a

    The 14C AMS dates were measured at the Kiel (KIA), Gif sur Yvette (GifA), Trondheim (TUa), and Potzdam (Poz) labs. The measurements are made on the foraminifer species N. Pachyderma (dex) and N. pachyderma (sin). The age of the Vedde ash layer is the GRIP age from Grönvold et al. [1995]. The dates were calibrated with the MARINE04 calibration curve, with ΔR of 0 ± 20 in the Holocene, and ΔR of 300 ± 20 in the pre-Holocene. The age-depth model uses a mixed-effect regression procedure, which uses the midpoint of the 1σ ranges rather than the intercept as a point estimate of the 14C dates, described by Heegaard et al. [2005].

JM97-948/2A
0top0−46
21.75KIA 6285735 ± 40369.5
30.75KIA 4800940 ± 40537.4
 
MD952011
5Poz-82451020 ± 100544.7
10.5GifA96471980 ± 60566.9
24.5KIA56001590 ± 40637.2
30.5KIA39251040 ± 40672.8
47.5KIA56011160 ± 30797
55.5Poz-82441530 ± 90868
70.5KIA39261560 ± 501018.9
89.5KIA62861590 ± 301239.4
101.5Poz-82461790 ± 601391.9
120.5Poz-101582210 ± 601638.4
130.5KIA39272350 ± 401761.7
154KIA62872335 ± 252034.6
170.5GifA964722620 ± 602213.5
225Poz-82423000 ± 502905.2
250Poz-82413380 ± 703302.7
269.5KIA100113820 ± 353646.4
300Poz-82404080 ± 704170.8
320.5KIA4634330 ± 504529.5
361.5Poz-101594990 ± 405273.9
451Poz-82386420 ± 1606791.35
490.5Poz-101606920 ± 507483
520.5KIA4647260 ± 608098.3
528.5Poz-82377690 ± 1108161.7
533.5Poz-82368530 ± 1608277.05
541.5Poz-82358280 ± 1408334.45
570.5Poz-82348700 ± 909198.2
703.5TUa-331510,775 ± 8511,984.23
730.5Tua-331611,875 ± 14012,734.9
750.5KIA46512,220 ± 9013,279.1
703.5veddeNA11,981
3.2.2. Diatoms

[7] Marine diatom species were used to reconstruct past sea surface temperature following the concepts and procedures described by Koç-Karpuz and Schrader [1990] and Andersen et al. [2004]. Diatoms were concentrated in the sediment samples through a process in which the samples are treated with HCl and H2O2 acids to remove carbonate and organic matter, neutralized of remaining acid, and separated from clay particles with a differential settling technique. For detailed descriptions of the diatom cleaning method see Koç et al. [1993]. Quantitative diatom slides of cleaned samples were prepared as described by Koç-Karpuz and Schrader [1990]. A Leica Orthoplan microscope with 1000X magnification was used for identification and counting of the diatoms, and the counting procedures described by Schrader and Gersonde [1978] were followed.

3.2.3. Statistics
3.2.3.1. Transfer Functions

[8] For quantitative estimation of past summer sea surface temperature (SSST) the following statistical methods were used: Imbrie and Kipp (IK) [Imbrie and Kipp, 1971], maximum likelihood (ML) [Upton and Cook, 2002], and weighted averaging partial least squares (WA-PLS) [ter Braak and Juggins, 1993]. The surface calibration set is based on 52 diatom species from 139 surface samples [Andersen et al., 2004]. The IK diatom transfer function is based on eight components and has a root mean square error of 1.25°C, a coefficient of determination between observed and inferred SSST of 0.89, and a maximum bias of 0.92°C. The eight different factors defined by the diatom assemblages were established and named according to their relation to modern surface hydrography [Andersen et al., 2004]. These are the Arctic Greenland Assemblages (factor 1), the North Atlantic Assemblage (factor 2), the sub-Arctic Assemblage (factor 3), the Norwegian Atlantic Current Assemblage (factor 4), the Sea Ice Assemblage (factor 5), the Arctic Assemblage (factor 6), the East and West Greenland Current Assemblage (factor 7), and the Mixed Water Masses Assemblage (factor 8). These factors enable us to reconstruct changes in the dominance of different water masses over the study site.

[9] A key assumption of transfer function methods is that the environmental variable of interest is an ecologically important variable, or that it is linearly correlated with one that is, and that this correlation is stable through time [Birks, 1995]. The ML method should be chosen if many of the taxa show a unimodal relationship with the environmental variable of interest [Telford and Birks, 2009]. The WA-PLS method can be regarded as the unimodal-based equivalent of multiple linear regressions [ter Braak and Juggins, 1993]. Like the IK method, WA-PLS uses several components in the final transfer function, and they are selected to maximize the covariance between the environmental variables to be reconstructed and hence the predictive power of the method. In this analysis we chose to use 4 components.

[10] Observed SSST versus predicted SSST for the different transfer function methods shows that the ML method has problems in reconstructing temperature below 4°C, and the IK method has problems in reconstructing temperature in the range of 14–18°C, while the WA-PLS method seems to perform satisfactorily through the whole SSST range [Berner et al., 2008]. For more detailed descriptions of the diatom transfer function methods see Berner et al. [2008] and Justwan et al. [2008].

3.2.3.2. SiZer

[11] SiZer (Significance of Zero Crossings of the Derivative) [Chaudhuri and Marron, 1999] was applied to explore frequencies with significant oscillations in the IK and WA-PLS sea surface temperature time series. SiZer methodology assumes that the reconstructed SSSTs are independent random variables. A key idea in SiZer is that significant features are found at different scales, that is, at different levels of smoothing. The SiZer approach is motivated by scale-space ideas from computer vision [Lindberg, 1994]. It departs in two important ways from the classical approach, where focus is on inference about the true underlying curve. First, SiZer studies a very wide range of bandwidths and thereby avoids the need to choose one bandwidth. Second, SiZer shifts the focus from the hypothesized true underlying curve to the observed curve, viewed at varying levels of resolution, and thereby avoids the bias. In SiZer, the notion of scale is controlled through the bandwidth in the kernel estimator. For further description of the SiZer method and an example of its application to paleoceanographic data, see Berner et al. [2008].

3.2.3.3. SiNos

[12] SiNos (Significant Nonstationarities) was also applied to our SSST time series to identify frequencies with significant oscillations. It is designed to handle time series where there is stochastic dependence between different data points [Godtliebsen et al., 2003]. This method explores potential nonstationarities in a stochastic process. SiNos simultaneously looks for significant changes in the mean, variance, and the first lag autocorrelation of the observed time series, with a null hypothesis that the process is stationary. The plot should be interpreted in an analogous way as described for SiZer.

3.2.3.4. Wavelet Analysis

[13] We also analyzed the reconstructed SSST time series (WA-PLS and IK) using wavelet transform. Wavelet transform decomposes a time series into wavelets and allows highlighting of the variability features at different time scales. The wavelet analysis thereby provides a useful extension to the common spectral estimates showing how the frequency content of the analyzed time series varies in time [Torrence and Compo, 1998; Percival and Walden, 2000]. The statistical significance of wavelet power at each particular point of the wavelet variance was assessed at 95% confidence levels against a red noise background (see Torrence and Compo [1998] for further details). The appropriateness of the AR1 (first-order autoregressive process) model to describe the analyzed time series was tested using a nonparametric runs test [Bendat and Piersol, 1986]. This also provided a sufficient condition for weak stationarity of the time series, which is necessary for the wavelet spectrum to be well defined [Percival and Walden, 2000]. The respective lag-1 autocorrelations were estimated directly from the original time series using the RedFit method devised by Mudelsee [2002] and Schulz and Mudelsee [2002]. This avoids the potential overestimation of the autocorrelation coefficient due to data resampling and makes the analysis less conservative. Following the method proposed by Torrence and Compo [1998], the wavelet transform is also applied for band-pass filtering of the analyzed records. For a more detailed description of the method we refer to Berner et al. [2008].

4. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

4.1. Surface Ocean Productivity

[14] The diatom abundance values (number of valves per gram dry sediment) for core MD95-2011 have been converted to flux estimates as flux is equal to the number of valves times MAR, were MAR is the mass accumulation rate (in g cm−2 kyr−1), to allow inferences about surface ocean productivity in the past. Diatom fluxes are generally low through the whole record (Figure 2). Very low fluxes (<1 × 106 g cm−2 kyr−1) of diatoms are recorded in the sediments prior to 11 kyr B.P. indicating unfavorable surface conditions for diatom production. This is the period prior to the establishment of the warm NwAC, a period when the surface conditions were cold and affected by the melting of the Scandinavian ice sheet [Koç et al., 1993].

image

Figure 2. Total diatom abundance valves (in cm−2 kyr−1) at the Vøring Plateau site through the last 13 kyr B.P. shown as (a) diatom valves per gram sediment and (b) diatom accumulation flux. Flux is abundance times MAR, where MAR is the mass accumulation rate (in g cm−2 kyr−1).

Download figure to PowerPoint

[15] Low flux of diatoms (∼1–2 × 106 g cm−2 kyr−1), implying low surface ocean productivity, characterizes the Holocene Climate Optimum between 9.5 and 6 kyr B.P. Two pronounced peaks, at 8.5 and 7 kyr B.P., with fluxes increasing to around 6 × 106 g cm−2 kyr−1 are exceptions to this.

[16] The average flux increases slightly after the HCO, reaching values of ∼2 × 106 g cm−2 kyr−1 (Figure 2). Diatom fluxes show a prominent increase to 125 × 106 g cm−2 kyr−1 between 1and 0.5 kyr B.P., indicating very favorable conditions for surface productivity during this period. However, an abrupt decrease is registered at 0.5 kyr B.P., with a very low flux of diatoms (∼1 × 106 g cm−2 kyr−1) since then.

4.2. Factor Analyses

[17] The diatom species registered in core MD95-2011 are converted to 8 diatom assemblages (factors) through Q-mode factor analysis. The various diatom assemblages reflect different water masses in the area [Andersen et al., 2004]. Shifts in the loadings of these factors through the Holocene will therefore reflect changes in the dominance of different water masses over the study site. Results from the factor analyses show high communalities, generally between 0.7 and 0.9, indicating tight correspondence between the modern assemblages and the Holocene assemblages at the study site. This also provides confidence in the estimated SSST (Figure 3). The general Holocene succession of the assemblages in core MD95-2011 is described by Birks and Koç [2002]. Here we will focus on the periods between 9.5 and 6 kyr B.P. and 3 kyr B.P. to present which is analyzed at high resolution.

image

Figure 3. (a) Reconstructed SSST (by weighted averaging partial least square) from core MD95-2011 from the Vøring Plateau site. Diatom assemblages (factor loadings times 100) reflecting different water masses in the area: (b) mixed water masses (factor 8), (c) East and West Greenland Current (factor 7), (d) Arctic (factor 6), (e) Norwegian Atlantic Current (factor 4), (f) sub-Arctic (factor 3), and (g) North Atlantic (factor 2).

Download figure to PowerPoint

[18] The factor analysis indicates that a marked shift in the oceanographic conditions at MD95-2011 core site occurred at the transition from the Holocene Climate Optimum to the late Holocene. Figure 3 suggests a clear dominance of the North Atlantic Assemblage (factor 2) during the HCO with factor loadings reaching 90. This implies a persistent presence of warm North Atlantic water masses over the core site. The Norwegian Atlantic Current Assemblage (factor 4), in turn, exhibits relatively high-amplitude variations throughout the HCO, with much lower factor loadings (0–30) compared with the present. This is indicative of only sporadic presence of this type of water mass at the core site during this part of the Holocene.

[19] During the LHP the opposite situation occurs: a clear dominance (factor loadings of 90–100) of the Norwegian Atlantic Current Assemblage is evident, with the North Atlantic Assemblage (Figure 3) factor loading greatly reduced to its present value of about 20–30. This suggests a reduced influence of the warm North Atlantic waters at the core site compared with the HCO. The surface conditions during the LHP are closer to the present conditions. Two distinct peaks in the East and West Greenland Current Assemblages (factor 7) are notable, at 2 and 0.5 kyr B.P., indicating the influence of colder surface ocean conditions with an influx of cold Arctic Water during these times. The same pattern, though of lesser magnitude, is registered in the sub-Arctic Assemblages (factor 3) and Arctic Assemblages (factor 6).

4.3. Sea Surface Temperature

[20] August sea surface temperature at the MD95-2011 site were previously quantitatively estimated for the last 13 kyr B.P. using diatoms [Birks and Koç, 2002; Andersen, 2003; Andersen et al., 2004; Berner et al., 2010]. Here we present the high-resolution SSST record for the period 9.5–6 kyr B.P. and 3–0 kyr B.P. and compare these two periods (Figure 4). Figure 4 shows the reconstructed SSST derived using three different transfer function approaches utilized in this study (ML, IK and WA-PLS). Two of the methods, the ML and WA-PLS, yield similar SSST variability for the HCO interval, with the main difference between the method outputs lying in the reconstructed absolute temperature. Within the interval 9.5–6 kyr B.P., the IK method reconstructs SSST some ∼0.5–2°C warmer than the other two methods, with a smaller inferred magnitude of temperature variability. The latter can be related to a relatively poor performance of this technique at water temperature above +14°C [Berner et al., 2008]. In the late Holocene period, ML reconstructed SSST are some ∼0.5°C–1°C warmer than from the other two techniques. Yet these discrepancies lie well within the uncertainties associated with the methods applied.

image

Figure 4. (a) Holocene SSST variations at the Vøring Plateau (core MD95-2011) as reconstructed by the IK, ML, and WA-PLS techniques. The IK and WA-PLS SSST reconstructions for the selected periods of (b) 6–9.5 and (c) 0–3 kyr B.P. are shown. The color coding is identical to Figure 4a. The following notations are used in Figure 4c to designate the specific climate periods during the last 3000 years: LIA (Little Ice Age), MWP (Medieval Warm Period), HWI (Holocene Warm Period I), HWII (Holocene Warm Period II), and HCI (Holocene Cold Period I). Black triangles in Figure 4a mark the AMS 14C dates used to build a core chronology.

Download figure to PowerPoint

[21] SiZer and SiNos are applied to the IK and WA-PLS reconstructed SSST series to define the statistically significant trends and other features in the records. The results of the analyses are presented in Figures 57. Figures 5b and 5e and 7b and 7e show family plots of smoothed reconstructed SSST for the two periods of 6.0–9.5 and 0–3 kyr B.P., respectively. The feature maps showing significant features at different time scales of variability are displayed in Figures 5c, 5f, 7c, and 7f. It should be noted that when the SiZer map shows, for example, a significant increase, this corresponds to an actual rise in the proxy value, and vice versa for significant decreases. SiNos is applied to the HCO part of the record to detect the variance changes in the reconstructed SSST across the range of time scales (Figure 6). The plot is to be interpreted in an analogous way as described for SiZer.

image

Figure 5. SiZer analysis of the HCO (6–9.5 kyr B.P.) SSST from core MD95-2011 derived by the (a) IK and (d) WA-PLS techniques. Family plots of smoothed curves constructed for different values of bandwidth (h, in years) for (b) IK and (e) WA-PLS. The green dots in Figures 5b and 5e represent the raw reconstructed SSST values. The red solid curve in a family plot corresponds to a choice of h one typically would use if only one scale were to be used. This particular h is a data-driven bandwidth and a best choice from a purely statistical point of view [Ruppert et al., 1995]. (c and f) SiZer maps given as a function of location (time) and scale (h). In agreement with common practice in the geological community, we interpret curves from right to left: a significant decrease is flagged as blue while a significant increase is flagged as red. The color purple is used at locations where the derivative is not found to be significantly different from zero. The color gray is used to indicate where too few data are available to do the inference. Typically the color gray occurs at very small scales for SiZer. The horizontal black lines in Figures 5c and 5f correspond to the smoothing obtained by the solid red line in the family plot (Figures 5b and 5e). The dotted curves in the SiZer maps show “effective window widths” for each bandwidth, as intervals representing ±2h, that is, ±2 standard deviations of the Gaussian kernel used.

Download figure to PowerPoint

image

Figure 6. (a) WA-PLS reconstructed MD95-2011 SSST for the HCO period. (b) Family plot of local variances of reconstructed SSST estimated for different window widths. (c) SiNos map of the analysis of variance with black (white) coding used to designate the periods of time and time scales where statistically significant decreases (increase) in local variance is detected. Areas with no significant change in variance are depicted as gray, and areas where no inference is being performed are light gray.

Download figure to PowerPoint

image

Figure 7. Same as in Figure 5 but for the LHP (0–3 kyr B.P.) reconstructed MD95-2011 SSST.

Download figure to PowerPoint

[22] Analysis of the reconstructed SSST series suggests that during the Holocene Climate Optimum the sea surface temperature at the Vøring Plateau were significantly higher than today, with SSST ranging from 13°C to 16°C, depending on the reconstruction method considered. The magnitude of SSST variations in WA-PLS and ML reconstructions during the HCO was on the order of 0.5°C–3°C and generally not stable through time. SiNos results indicate that the highest amplitude of the variability, and hence the least stable SSST, was registered in the time intervals of 9.5–8.9 kyr B.P. and 7–6 kyr B.P. The SSST variability tends to decrease toward the middle part of the HCO, as shown by black and white colors from 7 to 7.5 and 8.5 to 9 kyr B.P. in Figure 6c, suggesting that the most stable HCO SSST in the area occurred during 7.3–8.9 kyr B.P.

[23] SiZer maps presented in Figures 5c and 5f show that the inferred patterns of climate evolution at the core location during the HCO are similar when using either the IK or the WA-PLS approach. For both reconstructions, statistically significant increasing (decreasing) trends in SSST are identified in the beginning (termination) of the Holocene Climate Optimum, at time scales comparable with the length of the considered record. At shorter time scales no statistically significant features were found, mainly due to a large magnitude of SSST oscillations as well as a relatively low sampling rate in this section of the core.

[24] MD95-2011 SSST during the last 3 kyr varied in the range of 9.5°C–11.5°C (Figure 4). All three reconstruction techniques applied here depict similar variability in the inferred temperature, although the mean SSST derived by ML is some 0.9°C warmer on average compared with the other two methods. The blue color at log10(h) > 3 in the SiZer maps of the series (Figures 7c and 7f; results for IK and WA-PLS are shown) indicates a general cooling trend in SSST at the Vøring Plateau from 3 kyr B.P. to the present.

[25] Superimposed on this general cooling, there are temperature variations at shorter time scales with at least two pronounced cooling events identified in SiZer in the broad range of bandwidths. These cold periods, spanning the intervals from 2.3 to 1.9 kyr B.P. and 0.6 to 0.3 kyr B.P. (Figure 4c), correspond to the Holocene Cold Period I (HCI) and the Little Ice Age (LIA).

[26] Three periods with SSST close to or higher than today are recorded: from 3 to 2.3 kyr B.P., 2 to 1.5 kyr B.P., and 1 to 0.6 kyr B.P. (Figure 4c). The Holocene Warm Period II (HWII) (1000–300 BC) stands out as the warmest period of the LHP, showing mean water temperature of about 10.8°C, that is, some 0.5°C above the 3000 year average of 10.3°C (WA-PLS estimate), with the largest SSST variability, on the order of 1°C–2°C. The Holocene Warm Period I (100–500 AD) has SSST close to the present, with smaller temperature variability (0.5°C–1°C). The last warm period lasted from 900 to 1400 AD and is associated with what is generally referred to as the Medieval Warm Period. Analogous to the HCI and LIA, the warm periods were not uniformly warm, but contained statistically significant, abrupt warming and cooling episodes, according to SiZer (Figure 7).

4.4. Wavelet Analysis

[27] For statistical evaluation of the variability recorded in the SSST time series at different time scales, we applied continuous wavelet transform to the IK and WA-PLS data sets. We used the data for the periods 9.5–6 kyr B.P. and 3 kyr B.P. to present, which fit the AR1 model at 95% confidence level and also have a higher sampling resolution. Figures 8a8d show the normalized wavelet power spectra for the IK and WA-PLS reconstructions for core MD95-2011. Wavelet spectra for both transfer function methods demonstrate well-pronounced millennial-scale variability; however, this is located mainly within the “cone of influence” and therefore should be interpreted with a caution. Centennial to multicentennial variability on the time scales from 80 to 120, 210 to 320, and 320 to 640 years is prominent in the HCO period (Figures 8a and 8b). It is much more pronounced in WA-PLS reconstruction, likely due to the better performance of this method at higher temperature compared with the IK. Variability at 320–640 and 640–1280 year time scales, in turn, is dominant in the LHP (Figures 8c and 8d). Integrating the wavelet spectra along the time axis provides consistent estimates of the spectral density function [Percival and Walden, 2000]. These global wavelet spectra are shown for both transfer functions in Figures 8e8h and may additionally allow inferences to be made about the typical variability in the SSST inherent to a particular time scale. The results are in general agreement with what was inferred from the wavelet spectra. This confirms the conclusion made earlier from the visual consideration of the plots that the more pronounced millennial changes in the IK-derived reconstruction tend to conceal the variability on the finer time scales.

image

Figure 8. The wavelet power spectra for the MD95-2011 sea surface temperature proxy time series for the (a and c) IK and (b and d) WA-PLS transfer functions. Solid contours enclose the regions where the wavelet power is above the red noise background at the 95% confidence level. The semitransparent areas highlight the “cone of influence” where the edge effects of the wavelet transform become important [Torrence and Compo, 1998]. (e-h) Black solid lines show the respective global wavelet spectra, while the red dashed lines indicate the 95% confidence level for a test against the red noise background.

Download figure to PowerPoint

5. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[28] Previous SSST reconstructions from the Nordic Seas described the Holocene climate as generally evolving in step with decreasing solar insolation after the climate optimum of the early Holocene [e.g., Koç et al., 1993; Koç and Jansen, 1994; Birks and Koç, 2002; Solignac et al., 2004]. The inferred amplitude of this cooling is on the order of 4°C–5°C over the Vøring Plateau, and implies both a cooling of the surface waters due to the orbitally driven changes in the northern hemisphere June insolation and a reduction in the influx of the NwAC to the eastern Norwegian Sea through the last 7000 years (Figure 9) [Birks and Koç, 2002; Andersen et al., 2004]. Our analysis of the higher-resolution MD95-2011 series demonstrates that superimposed on this general cooling trend there are pronounced (multi)centennial-scale quasiperiodic SSST variations of 80–120, 210–320, and 320–640 years during the Holocene Climate Optimum, and 320–640 and 640–1280 years during the late Holocene period (Figure 6). Due to the relatively low sampling rate in the part of the record comprising the mid-Holocene climate transition, this period is not included in the analysis that follows.

image

Figure 9. WA-PLS (red) and IK (black) reconstructed SSST of core MD95-2011 shown together with orbitally driven changes in northern hemisphere June insolation at 65°N (gray line) for the last 12,000 years.

Download figure to PowerPoint

5.1. Holocene Climate Optimum

[29] Diatom assemblages and quantitative SSST reconstructions from the Vøring Plateau indicate that the warmest surface ocean conditions in the Holocene occurred during approximately 8.9–7.3 kyr B.P. (Figures 6 and 9), with the exact timing depending on the reconstruction method considered. Before and after this time interval (approximately 9.5–8.9 and 7.3–6.0 kyr B.P.) the SSST, much higher than that observed today at the core site, exhibited variations of magnitude 1°C–3°C (Figure 6) superimposed on the warming and cooling trends (Figure 5). The period between 8.9 and 7.3 kyr B.P. is the core of the Holocene Climate Optimum in the study area when SSST stabilized at around 14°C–15°C (ML and WA-PLS estimates), or even 15°C–16°C (IK results), and exhibited low variability. The strong northward inflow of warm and saline North Atlantic Water to the Vøring Plateau during the local HCO substantially increased the east-west SSST gradient between the Norwegian Sea and the Iceland Sea [Koç et al., 1996; Andersen et al., 2004; Justwan et al., 2008] as well as the NE-SW SSST gradient between the Norwegian Sea and the Irminger Basin in the North Atlantic [Berner et al., 2008]. The area to the southwest of Iceland was under Arctic conditions with summer water temperatures varying between 8°C and 11°C, in contrast to sub-Arctic conditions which prevail at present [Berner et al., 2008]. The warm and stable surface conditions and very low fluxes of diatoms registered in core MD95-2011 during the HCO implies that the site was located in the regional warm water pool removed from the oceanic fronts and Arctic waters.

[30] Both proxy-based reconstructions and modeling studies point to a substantial regional difference in the onset and duration of the HCO in the North Atlantic. The HCO in the Norwegian Sea at MD95-2011 core site begins almost 2000 years later than the Northern Hemisphere orbitally forced summer insolation maximum at ∼11.000 years B.P., due to the presence of remnants of ice sheets from the last glacial period [e.g., Koç et al., 1993; Renssen et al., 2009]. Different timing of the HCO is observed on different sides of the North Atlantic Ocean. The Nordic Seas, as well as northwestern Europe, attained peak warmth between 9 and 7 kyr B.P. [Koç et al., 1993; Klitgaard-Kristensen et al., 2001; Koç and Jansen, 2002; Andersen et al., 2004], whereas the subpolar North Atlantic (Irminger basin) was warmest from 7 to 5 kyr B.P. [Andrews and Giraudeau, 2003; Berner et al., 2008]. However, in the Labrador Sea the HCO in summer SSST was from ∼11 to 8 kyr B.P., but with the winter SSSTs colder than at present, implying increased seasonal contrast and sea ice duration than in the late Holocene [Solignac et al., 2004]. These regional differences arise from the profound and prolonged effects of the Laurentide Ice Sheet (LIS) disintegration. This process exerted a lasting influence on the climate development via sustained meltwater discharge with attendant impacts on NADW formation in the Labrador and Greenland seas, trajectories and strengths of the NAD and the Irminger Current, and eventually the structure of the atmospheric circulation [e.g., Andersen et al., 2004; Solignac et al., 2004; Kaplan and Wolfe, 2006, Renssen et al., 2009]. These effects persisted as late as 7 kyr B.P. based on the current LIS deglacial chronology [Carlson et al., 2008]. The different timing of the HCO in the North Atlantic therefore suggests a possible out of phase relationship between the eastern and western basins of the North Atlantic [Andersen et al., 2004].

[31] We note that the timing of the thermal maximum attained during the HCO is not in line with some modeling results for this area. Renssen et al. [2009] suggested the effect of melting LIS would cause a delayed HCO, to a period of 7–6 kyr B.P., along the Atlantic seaboard of Europe, which is some 2000 years later than inferred from MD95-2011 SSST reconstruction. In the Irminger basin the modeled HCO (6–5 kyr B.P.) occurred some 1000 years later than was actually found from the proxy data [Berner et al., 2008]. The modeled persistent perennial sea ice cover in the Labrador Sea around 9 kyr B.P. is also not supported by the sediment core data [Solignac et al., 2004]. It suggests that the model does not completely catch the complexity of the area and/or the fresh water flux approximation used, as well its time variability were not sufficiently accurate.

5.2. Changes in the Time Scales of Variability and Attribution of Potential Forcing Mechanisms

[32] Superimposed on the general Holocene climate evolution, the reconstructed SSST from core MD95-2011 shows pronounced quasiperiodic sub-Milankovitch-scale fluctuations at time scales of approximately 80–120, 210–320, 320–640, and 640–1280 years (Figure 8). There are obvious differences in the character of the submillennial-scale variability during the LHP compared with the HCO. The pattern of the mid-Holocene climate transition in the characteristic scales of variability longer than 1000 years have been discussed earlier by Debret et al. [2007, 2009]. The HCO and the LHP considered in the present study encompass only ∼3000 yearlong time intervals; the variations with periods longer than ∼1500 years therefore may not be resolved.

5.2.1. Quasiperiodic Variations Observed With Both High and Low Summertime Insolation

[33] Variability in MD95-2011 SSST at time scales of 320–640 years is evident in both the HCO and the LHP (Figure 8). These quasiperiodic variations occur rather consistently through the HCO (WA-PLS reconstruction), while in the LHP this variability is only detected from 3 to 2 kyr B.P. and 1 kyr B.P. to the present. Previous investigations from core MD95-2011 have revealed variations at the same time scale in planktic δ18O from 8 to 2 kyr B.P. (417 years) and from 11.5 to 6 kyr B.P. (570 years) [Risebrobakken et al., 2003]. Other climate records have demonstrated variations at a similar time scale of ∼500 years in sortable silt and δ13C records from the Gardar Drift [Chapman and Shackleton, 2000], and in δ18O and Δ14C from the GISP2 ice core [Stuiver and Braziunas, 1993; Stuiver et al., 1995; Grootes and Stuvier, 1997]. As also suggested [Chapman and Shackleton, 2000; Schulz and Paul, 2002], observations provide a clear indication of a strongly interconnected atmospheric and oceanic variability during the Holocene in the entire North Atlantic region. Additionally, it has been suggested that the ∼500 year variability found in δ18O from the GISP2 ice core is related to variations in total solar irradiance [Stuiver et al., 1995].

5.2.2. Quasiperiodic Variations Observed Only With High or Low Summertime Insolation

[34] Figures 8c, 8d, 8g, and 8h suggest the presence of prominent SSST variations at the multicentennial to millennial time scale of 640–1280 years during the LHP. Variations in this range of time scales have been reported earlier in a number of paleoclimate studies. Evidence for Holocene millennial-scale fluctuations with periodicity of around 1000 years was detected in sortable silt and δ13C records from the Gardar Drift [Chapman and Shackleton, 2000] and in the δ18O and Δ14C records from the GISP2 ice core [Stuiver and Braziunas, 1993; Stuiver et al., 1995; Grootes and Stuvier, 1997]. In addition, quasiperiodic variations in the range of 640–950 years are documented in terrestrial temperature records from the Scandinavia [Dahl and Nesje, 1996], and in the range of 640–960 years in diatom SSST records from the Reykjanes Ridge in the period from ∼9 to ∼2 kyr B.P. [Berner et al., 2008].

[35] SSST variations of the multicentennial scale of 210–360 years are prominent only in the HCO at the Vøring Plateau site (Figures 8a, 8b, 8e and 8f). Variability at a similar time scale of 260 years are also documented in the planktic δ18O record from the same core [Risebrobakken et al., 2003], while 140–320 year variability is documented in glacier variations in southern Norway [Matthews et al., 2000] and in δ18O from the GRIP ice core [Yiou et al., 1997]. This shorter time scale of 320 years on average is close to the one found in Δ14Cresidual record [Stuiver et al., 1998; Crosta et al., 2007] and hence potentially associated with variations in solar activity.

[36] Nearly centennial SSST variations of 80–120 years are documented consistently through the HCO in the MD95-2011 record (Figure 4b). Earlier investigations from the same core document variability of ∼80 years based on the planktonic δ18O record, which appears to be consistent through time [Risebrobakken et al., 2003], and 80–120 years in diatom SSST records during the Preboreal [Berner et al., 2010]. Several papers have reported variability at the time scale of around 80 years, including tree ring records from 500 AD [Briffa et al., 1992], historical and instrumental records during the last century [Schlesinger and Ramankutty, 1994], and Holocene ice core records [Chambers and Blackford, 2001]. Friis-Christensen and Lassen [1991] discussed the variations in solar irradiance over an 80 year period, and the length of this period might indicate a possible connection to the solar Gleissberg cycle [e.g., Waple, 1999; Chambers and Blackford, 2001]. The ∼80 year variability has also been attributed to internal quasiperiodic oscillations in the atmospheric-ocean system, or to changes in the thermohaline circulation [e.g., Eddy, 1977; Ribes, 1990; Waple, 1999; Chambers and Blackford, 2001]. Changes in NADW production are suggested as an additional mechanism for internal oscillations by amplifying the solar signals and transmitting them globally [Delworth et al., 1993; Schlesinger and Ramankutty, 1994; Mann et al., 1995; Mahasenan et al., 1997].

[37] Attribution of quasiperiodic changes identified in proxy-based reconstructions to specific forcing process(es) is often hampered by a general complexity of the climate system and processes forming a particular proxy as well as by the lack of knowledge about past variations in solar activity. Additional complicating factors are frequently uneven time increments and timescale errors which lead to uncertainties in the spectral estimates [Mudelsee et al., 2009].

[38] In order to identify the possible forcing mechanisms for the quasi-cyclical variations found in the MD95-2011 SSST record we used the ice core 10Be-based Holocene record of TSI (total solar irradiance) of Steinhilber et al. [2009]. Wavelet analysis applied to this series (Figure 10) identifies quasiperiodic variations at a broad range of time scales, with most of the series variance concentrated in the higher-frequency (subcentennial) band. Analysis suggests a lack of any stationary variability evident throughout the entire record but rather the presence of intermittent variations at different timescales. The wavelet coherence approach [Torrence and Compo, 1998; Grinsted et al., 2004] performed on the reconstructed WA-PLS SSST and TSI series (not shown) revealed statistically significant coherent variations in the bands of 400–600 years during the HCO, and 260–450 and 640–900 years during the LHP. The results are visualized in Figure 11. To band-pass filter the signals in the respective frequency ranges we used the scales-averaged wavelet power following the technique described by Torrence and Compo [1998].

image

Figure 10. As in Figure 8 but for the normalized wavelet power spectra for the Holocene Total Solar Irradiance series of Steinhilber et al. [2009].

Download figure to PowerPoint

image

Figure 11. Normalized wavelet band-pass-filtered series of WA-PLS reconstructed SSST from the MD95-2011 site (black) and the reconstructed total solar irradiance (dashed gray; see text for details). For the late Holocene period (LHP), the frequency bands (a) 640–900 years and (b) 260–450 years are shown, and for the Holocene Climate Optimum (HCO), the band (c) 400–600 years is shown.

Download figure to PowerPoint

[39] Figure 11a shows nearly synchronous variations in TSI and MD95-2011 SSST at the time scale of 640–900 years, indicating that this submillennial-scale mode of variability, evident during the LHP, is directly associated with a varying solar forcing. The variations at the shorter scale of 260–450 years, similar to the 320–640 year mode of variability found in the reconstructed SSST, display a lagged response to solar forcing with a phase locked behavior (Figure 11b). The observed lag of about 100 years is rather consistent through time and points to the existence of a feedback mechanism in the climate system triggered by variations in the solar constant. Notable is that this mode of SSST variability at the core site is modulated by solar irradiance during the LHP only.

[40] Modeling studies [e.g., Renssen et al., 2005, 2006] proposed that the principal mechanism for centennial-scale cooling events in response to negative TSI anomalies involves a lasting reorganization of the oceanic circulation and deep convection shutdown in the Nordic seas, followed by sea ice expansion. A potential atmospheric mechanism amplifying cooling is related to a decrease in lower stratospheric ozone formation, resulting in amplified stratospheric cooling and leading to contraction of the Hadley Cells and expansion of the polar cells in the troposphere [Haigh, 1996].

[41] The lag in SSST cooling events, in turn, could be attributed to the thermal inertia of the oceans as well as a probabilistic character of the deep convection failure in response to a reduced TSI [Renssen et al., 2006]. It also implies that not only the magnitude but notable the duration of the TSI anomaly that increases the probability of the deep convection shutdown, so even moderate variations in TSI on longer scales are capable of generating pronounced, lasting anomalies in SSST.

[42] During the HCO the coherent variations in total solar irradiance and MD95-2011 SSST are revealed at the scale 400–600 years and show a variable phasing. We hypothesize that this could be related to the generally warmer climate of the HCO, which lowers the probability of drastic changes in the oceanic circulation in response to TSI anomalies. It leads to a less consistent response to changing solar irradiance during the periods with high orbital forcing [Renssen et al., 2005, 2006].

[43] The most prominent cooling events of the last 3000 years are recorded during the HCI at 2300 years B.P. and the LAI at 500 years B.P. These events coincide and are nearly in phase with minimum submillennial and multicentennial TSI variations. A striking feature of these events is the abruptness with which they commence, in line with the mechanism proposed by Renssen et al. [2006] involving the abrupt shutdown of the THC and associated sea ice expansion. The HCI shows a 1.5°C SSST cooling within about a half a century, whereas the LIA starts with a SSST fall of 1.5°C within a decade. During these two periods the core site was under the direct influence of cold Arctic waters. We note that the onset of the LIA recorded in the analyzed reconstruction leads by approximately 40–50 years the shift in NAO, from a consistently positive to a more variable state, similar to the one observed at present [Trouet et al., 2009]. The inferred lag is in agreement with the modeled delayed response of atmospheric NAO to changes in solar insolation and SSST [Swingedouw et al., 2010]. We note that neither the HCI nor the LIA were uniform periods, and both periods are characterized with two cold peaks at the start and end of the periods, with a warmer interval in the middle.

6. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[44] Our high-resolution SSST record from the Vøring Plateau confirms the Holocene evolution of the Nordic Seas marine climatology was primarily forced by orbitally paced solar insolation that progressively weakened after the early Holocene climate optimum of the early Holocene [Koç et al., 1993; Koç and Jansen, 1994; Birks and Koç, 2002]. Analysis of this high-resolution SSST record demonstrates that the warmest surface ocean conditions with water temperatures attaining 14–15°C occurred during approximately 8.9–7.3 kyr B.P. Before and after this time interval (approximately 9.5–8.9 and 7.3–6.0 kyr B.P.) the SSST, still much higher than observed today at the core site, exhibited variations of magnitude 1°C–3°C (Figure 6) superimposed on the warming (earlier period) or cooling trends (Figure 5). This implies that in the core of the Holocene Climate Optimum the site was located in the regional warm water pool removed from the oceanic fronts and cold Arctic waters. During the late Holocene period the SSST varied in the range of 9.5°C–11.5°C with distinct warm and cold periods associated with substantial changes in the hydrographic conditions at the core site.

[45] Superimposed on the general Holocene cooling trend are centennial- to millennial-scale quasiperiodic variations of about 80–120, 210–320, 320–640, and 640–1280 years. Using the reconstructed series of total solar irradiance we found evidence that the submillennial-scale mode of variability (640–900 years) in SSST evident during the late Holocene period is directly associated with a varying solar forcing. At the shorter scale of 260–450 years, the Vøring Plateau SSST during the LHP displays a lagged response to solar forcing, with a phase locked behavior (Figure 11b). The observed lag of about 100 years indicates the existence of a feedback mechanism in the climate system triggered by variations in the solar constant as well as the role of the thermal inertia of the ocean.

[46] The abruptness of the cooling events in the late Holocene period, especially pronounced during the onsets of the Holocene Cold period I (approximately 2300 years B.P.) and the Little Ice Age (approximately 550 years B.P.), matches the proposed mechanism of the deep convection shutdown in the Nordic seas in response to negative TSI anomalies quite well. Such a failure in convection would involve a lasting reorganization of the oceanic circulation, sea ice expansion and eventually changes in atmospheric circulation.

[47] During the HCO the SSST at the Vøring Plateau shows a less consistent response to changing solar irradiance. Although the coherent variations in total solar irradiance and MD95-2011 SSST are revealed at the scale 400–600 years, the phasing was not stable. We hypothesize that this could be related to a generally warmer climate of the HCO, associated with a lower probability of drastic changes in the oceanic circulation in response to TSI anomalies.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[48] The research for this paper is supported by the Research Council of Norway and the European Commission through the PACLIVA project, the NORPAST-II project, and the Norwegian Polar Institute. Xavier Crosta, an anonymous reviewer, and the Editor Rainer Zahn are gratefully acknowledged for their constructive comments. The University of Uppsala (Sweden) Department of Earth Sciences is acknowledged for hosting D. Divine during the preparation of the revised version of the manuscript.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Physical and Oceanographic Setting
  5. 3. Material and Methods
  6. 4. Results
  7. 5. Discussion
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information
FilenameFormatSizeDescription
palo1679-sup-0001-t01.txtplain text document2KTab-delimited Table 1.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.