Support for tropically-driven Pacific decadal variability based on paleoproxy evidence

Two independent proxy reconstructions of sea surface temperature reflect a common pattern of Pacific decadal sea surface temperature variability over the past two centuries. Since the pattern extends to both the northern and southern hemispheres, this result supports the idea that Pacific decadal variability is a basin-wide phenomenon originating in the tropics. The causes of Pacific decadal climate variability (here termed PDV) are the subject of much debate. Most work describing PDV has focused on the North Pacific, where the leading EOF of low-pass filtered sea surface temperature (SST) poleward of 20S was labelled the Pacific Decadal Oscillation (PDO) [Zhang et al., 1997; Mantua et al., 1997]. But is PDV generated in the North Pacific? Or is PDV a basin-wide pattern of variability? If so, then it is most likely to originate in the tropics [Garreaud and Battisti , 1999]. Resolving the processes controlling PDV is important for predicting socially important downstream effects on Pacific fisheries and water resources in the arid western United States [Mantua et al., 1997; Dettinger et al., 2001]. Paleoclimatologists are extending observations of PDV into the past using proxy observations to help address such questions [Minobe, 1997; Wiles et al., 1998; Linsley et al., 2000a; Biondi et al., 2001; Gedalof and Smith, 2001]. These studies have identified large variations in the frequency of and interval between reconstructed PDO regime shifts prior to the observational period. However, the physical mechanism for PDV remains unknown. Two recent sets of proxy observations support the contention that PDV is a tropically-driven basin-scale phenomenon. One is a time series of Strontium:Calcium (Sr/Ca) ratio measurements on a reef coral growing offshore of Rarotonga (21S, 160W) [Linsley et al., 2000b]. On centennial time scales, sea water Sr/Ca ratio is approximately constant, and coral aragonite Sr/Ca composition varies linearly with temperature [Beck et al., 1992]. The Sr/Ca data resolve the seasonal cycle and cover the period from 1726 to 1997; resolution of the seasonal SST cycle at Rarotonga keeps proxy 1Dept. of Earth and Planetary Sciences, Harvard University, Cambridge, Massachusetts. 2Lamont-Doherty Earth Observatory, Palisades, New York. 3Dept. of Earth and Atmospheric Sciences, State University of New York, Albany, New York. 4Laboratorio de Dendrocronologia e Historia Ambiental, IANIGLA, Mendoza, Argentina. 5Dept. of Biology and Biochemistry, University of Houston, Houston, Texas. Copyright 2001 by the American Geophysical Union. Paper number 2001GL013223. 0094-8276/01/2001GL013223$05.00 data age model error to a few months. Linsley et al. [2000b] argued that their time series resembles the PDO index. As illustrated in Figure 1a, the pattern of SST variability associated with Rarotonga Sr/Ca on decadal time scales is similar to that observed by Garreaud and Battisti [1999] in 1958–1996 reanalysis data [Kalnay et al., 1996]. Citing symmetry arguments [Garreaud and Battisti , 1999], Linsley et al. [2000b] suggested a tropical source for basin-scale PDV. Recent independently-derived dendroclimatological estimates of PDV support this idea. Evans et al. [2001] used a set of 15 tree ring indicators from Pacific-influenced regions of extratropical North and South America to reconstruct the Pacific basin SST field (70W 110E, 70N-60S). The tree ring indicators are sensitive to air temperature, precipitation or drought [Villalba et al., 2001]; each indicator is itself based on cross-correlation and standardization of many individual tree-ring data series from each site. Evans et al. [2001] found that these 15 tree-ring indicators could verifiably resolve one SST pattern, through presumed downstream influences on tree-ring site precipitation and/or temperature. This pattern is illustrated in Figure 1b; again, the pattern is similar to that illustrated by Garreaud and Battisti [1999]. We now compare these two paleo-PDV estimates (here called “Sr/Ca” and “TrSSTa”) to examine their consistency. We defer analysis of the earliest part of the Sr/Ca record (1727–1768) until the large observed shift in this interval can be replicated. Coherency between records (Figure 2) varies over time and frequency (Figure 3). To illustrate this we used singular spectral analysis (SSA; Vautard and Ghil [1989]) to divide each time series into decadal, interannual and biennial frequency components (Table 1). Coherence significance was determined by Monte Carlo simulations of correlations between time series having the red noise characteristics of the actual paleoproxy frequency components (Table 1). Coherence between decadal frequency components is significant at the 90% level in the 20th century, but varies between 70-90% for the 18th century (Figure 3a,d,e). Coherence between interannual components is ≥95% significant over the entire comparison interval (Figure 3b,d,e), while coherence significance of biennial components is low throughout the record (Figure 3c,d,e). The temporal similarity of these proxies corroborates recent modeling and observational studies linking tropical and basin-scale phenomena in the Pacific. The approximate symmetry of the spatial pattern about the equator and the simultaneous coherence between northern and southern Hemisphere proxy observations suggest a tropically mediated mechanism. Garreaud and Battisti [1999] suggested that basin-wide PDV originates in the tropical Pacific. Karspeck and Cane [2000] showed that a wind-driven tropical ocean model can reproduce a well-known feature of PDV, the 1976–77 decadal shift in thermocline depth anomaly. In a coupled atmosphere-ocean general circulation model