|
Science 23 February 2001
DOI: 10.1126/science.1057969
|
|
Variability in the El Niño-Southern Oscillation Through a Glacial-Interglacial Cycle
Alexander W. Tudhope, Colin P. Chilcott, Malcolm T. McCulloch, Edward R. Cook, John Chappell, Robert M. Ellam, David W. Lea, Janice M. Lough, Graham B. Shimmield
|
Supplementary Material
| Supplemental Table 1. 230Th dates for corals used in this study. Letters a through d indicate replicate analyses on the same sample; a1, b1 versus a2, b2 indicate separate samples from the same coral where these samples were analyzed in duplicate; a indicates analysis by alpha spectrometry (all other dates by thermal ionization mass spectrometry, TIMS). Roman numerals under "Location" refer to previously described raised reefs on the Huon Peninsula (1, 2). A commonly used criterion for acceptance of ages is that they have d234U(T) values overlapping the range 149 ± 8 ‰. However, especially with Porites corals, even this criterion may not always be reliable. For sample H95-16, we have accepted the TIMS dates (6.51 ka and 6.57 ka) in preference to the alpha spectrometry date (mean of four analyses of same sample = 8.3 ka) on the grounds that the TIMS technique gives greater precision, and the two TIMS dates (from different samples) agree well. The high 232Th content and high d234U(T) value for H95-58 suggest that the age is likely to be erroneous.
|
| Sample | Location | 232Th (ppb) | 238U (ppm) | d234U(O) | d234U(T) | (230Th/238U)act | Age (ka) |
|---|
M93TFCa1, a
M93TFCb1, a
M93TFCa2, a
M93TFCb2, a | Madang | | 2.55
2.58
2.48
2.57 | 188 ± 39
137 ± 37
177 ± 34
110 ± 36 | 189 ± 39
138 ± 38
178 ± 34
111 ± 37 | 0.0272 ± 0.0043
0.0259 ± 0.0043
0.0273 ± 0.0046
0.0244 ± 0.0044 | 2.52 ± 0.40
2.50 ± 0.42
2.55 ± 0.43
2.42 ± 0.44 |
LaingFC2a1, a
LaingFC2b1, a
LaingFC2a2, a
LaingFC2b2, a | Laing | | 2.70
2.70
2.49
2.41 | 150 ± 49
123 ± 49
124 ± 36
126 ± 58 | 151 ± 50
124 ± 49
125 ± 36
127 ± 59 | 0.0236 ± 0.0046
0.0295 ± 0.0053
0.0272 ± 0.0043
0.0286 ± 0.0046 | 2.25 ± 0.44
2.90 ± 0.52
2.67 ± 0.42
2.80 ± 0.46 |
H95-16
H95-16
H95-16a, a
H95-16b, a
H95-16c, a
H95-16d, a | Sialum, I | 1
0 | 2.91
2.91
249
2.38
2.44
2.37 | 144 ± 1
149 ± 1
122 ± 50
152 ± 48
123 ± 38
143 ± 36 | 147 ± 1
152 ± 1
125 ± 51
156 ± 49
126 ± 39
146 ± 37 | 0.0666 ± 0.0006
0.0675 ± 0.0008
0.0809 ± 0.0094
0.0818 ± 0.0091
0.0860 ± 0.0077
0.0835 ± 0.0073 | 6.51 ± 0.05
6.57 ± 0.08
8.13 ± 0.97
8.00 ± 0.91
8.65 ± 0.80
8.24 ± 0.74 |
| H96-27 | Kanzarua, IIa | 0 | 3.64 | 122 ± 1 | 136 ± 1 | 0.3352 ± 0.0017 | 38.3 ± 0.2 |
| H96-18 | Kanzarua, IIa | 0 | 3.40 | 125 ± 1 | 140 ± 1 | 0.3444 ± 0.0022 | 39.4 ± 0.3 |
| H96-6 | Kanzarua, IIIb | 0 | 3.72 | 119 ± 1 | 134 ± 1 | 0.3596 ± 0.0010 | 41.9 ± 0.1 |
| H95-14 | Kwangam, VIa | 0 | 3.83 | 102 ± 1 | 140 ± 1 | 0.7180 ± 0.0021 | 111.9 ± 0.6 |
| H95-58 | Sialum, VIIa | 40 | 3.22 | 128 ± 1 | 186 ± 1 | 0.8087 ± 0.0020 | 132 ± 0.6 |
H95-6
H95-6 | Kwangam, VII | 0
0 | 2.97
2.70 | 107 ± 1
108 ± 1 | 152 ± 1
154 ± 1 | 0.7668 ± 0.0016
0.7725 ± 0.0036 | 124.7 ± 0.5
126.1 ± 1.1 |
AC-U11a-top*
AC-U11b-top*
AC-U11-bottom*
AC-U11a-top*
AC-U11a-bottom* | Kwangam, VII | 0
0
0 | 2.85
2.85
2.88
3.35
3.27 | 108 ± 1
100 ± 3
110 ± 1
101 ± 6
100 ± 7 | 164 ± 1
142 ± 5
158 ± 2
145 ± 9
144 ± 9 | 0.7706 ± 0.0009
0.7646 ± 0.0026
0.7790 ± 0.0008
0.7662 ± 0.0077
0.7748 ± 0.0081 | 125.5 ± 0.9
125.7 ± 0.9
127.6 ± 0.9
125.6 ± 2.6
128.5 ± 2.8 |
| *Previously reported dates. AC-U11 samples (3) came from the same coral as H95-18 of this study. |
| Supplemental Table 2. Statistical significance of differences in ENSO variance between different time periods. Values of the test statistic for the Conover test of equality of variances and, in parentheses, the probability estimates. Numbers in italics and parentheses below the age group headings are the total number of peaks and troughs in the ENSO (2.5 to 7 year) bandpass filtered time series in each age group [= n for this test (4)]. Results in bold (P < 0.01) are regarded as significant.
|
| 2-3 ka
(83) | 6.5 ka
(29) | 38-42 ka
(93) | ~85 ka
(10) | ~112 ka
(17) | 118-128 ka
(23) | 130 ± 2 ka
(70) |
| modern | -4.508
(<0.001) | -5.162
(<0.001) | -5.650
(<0.001) | -1.490
(0.136) | -3.444
(0.001) | -2.232
(0.026) | -3.042
(0.002) |
| 2-3 ka | | -2.914
(0.004) | -0.438
(0.662) | 0.868
(0.385) | -0.950
(0.342) | 0.352
(0.725) | 1.883
(0.060) |
| 6.5 ka | | | 3.026
(0.002) | 2.863
(0.004) | 1.730
(0.084) | 2.550
(0.011) | 4.381
(<0.001) |
| 38-42 ka | | | | 1.062
(0.288) | -1.095
(0.274) | 0.592
(0.554) | 2.816
(0.005) |
| ~85 ka | | | | | -1.868
(0.062) | -0.299
(0.765) | 0.317
(0.751) |
| ~112 ka | | | | | | 1.058
(0.290) | 2.578
(0.010) |
| 118-128 ka | | | | | | | 0.888
(0.375) |
Supplemental Figure 1. Cumulative probability plots of absolute amplitude in the 2.5 to 7 year (ENSO) bandpass filtered coral d18O time series. Corals have been grouped to represent 8 time periods, including the modern. m, number of corals; n, total number of years in each group. Blue lines and symbols, "glacial" periods of low global sea-level; red lines and symbols, interglacial periods of high global sea-level.
Medium version | Full size version
References
- J. Chappell, Bulletin of the Geological Society of America 85, 553 (1974).
- J. Chappell et al., Earth Planet. Sci. Lett. 141, 227 (1996).
- T. M. Esat, M. T. McCulloch, J. Chappell, B. Pillans, A. Omura, Science 283, 197 (1999).
- The approximate null distribution of the Conover test is the standard normal distribution. Because the bandpass filter used here is centered roughly on 4.8 years, only the interannual peaks (maxima) and troughs (minima) associated with these pseudo-observations of ENSO variability contribute useful information on changes in ENSO amplitude. Therefore, we restricted our equality-of-variance tests to these inter-annual maxima and minima. This decision reduced the number of values used in the Conover test to roughly 12% of the original number of bandpass filtered seasonal values. This large downward adjustment in the number of values (degrees of freedom) used in the tests still does not take into account the full effect of the bandpass filter on the independence of observations that the Conover test assumes for its null distribution. Therefore, we only consider as "significant" those differences in variance that exceed the nominal 99% significance level, and warn against overinterpretation of the exact probability values.
|
|
