Portal de Periódicos da CAPES

Comment on “Equatorial Pacific coral geochemical records show recent weakening of the Walker circulation” by J. Carilli et al.

2015; American Geophysical Union; Volume: 30; Issue: 5 Linguagem: Inglês

10.1002/2014pa002753

1944-9186

Kristopher B. Karnauskas, Anne L. Cohen, Elizabeth J. Drenkard,

Geology and Paleoclimatology Research

Carilli et al. [2014] present new geochemical proxy records of sea surface temperature (SST) and salinity spanning 1959–2010 from coral cores collected from Butaritari Atoll, Gilbert Islands, Republic of Kiribati (3.07°N, 172.75°E). The continuous measurement of SST by in situ and satellite platforms generally began in the early 1980s (e.g., the Tropical Atmosphere-Ocean array of moorings [McPhaden et al., 1998] and the Advanced Very High Resolution Radiometer infrared satellite [Reynolds et al., 2002]), and instrumental reconstructions extending back to the midnineteenth century [e.g., Smith et al., 2008] rely on sparse observational sampling both in space in time. The Carilli et al.'s [2014] records thus represent a potentially valuable addition to estimates of past climate variability in the important western tropical Pacific region, confirming the dominant role of El Niño–Southern Oscillation (ENSO). However, the conclusion that the Walker circulation has exhibited a long-term weakening trend over the last several decades is based on an interpretation of the proxy record that does not consider well-documented natural interdecadal variability and is based on trends in a proxy record that are opposite from trends in the observational record over the overlapping portion of the satellite era. In addition, Carilli et al. [2014] misinterpret the relationship between their proxy record of SST and trends in ocean circulation due to incorrect assumptions about the physical oceanographic context of Butaritari. These two issues are addressed below. There are justifiable alternatives to interpreting the Butaritari Sr/Ca-based proxy SST record (Figures 3b and 8b of Carilli et al. [2014]) as a long-term trend. The "small but significant increase in SST (0.39°C) from 1959 to 2010" is entirely consistent with the well-documented Pacific climate shift of 1976/1977 [Trenberth, 1990; Miller et al., 1994; Guilderson and Schrag, 1999], often ascribed to the Pacific (inter) Decadal Oscillation (PDO) [Mantua et al., 1997]. The Butaritari Sr/CaSST time series is reproduced here in Figure 1 to reveal that the trend can be fully explained by a single jump (by more than 2.5°C) that occurred over the course of just 6 months in 1976. Rather than a long-term warming trend, the two individual periods before and after 1976 show statistically significant cooling trends (−0.52 ± 0.22°C/decade and −0.18 ± 0.09°C/decade, respectively, where error bars represent 95% confidence bounds) that are even larger in magnitude and more robust than the warming trend computed over the full period (0.11 ± 0.05°C/decade). Despite these robust cooling trends, the difference in mean SST (postshift minus preshift) is 0.7°C, which is why a trend line drawn through these data will have a positive slope. For these reasons, we find a more apt characterization of the Butaritari Sr/CaSST record to be an abrupt warming superimposed upon a long-term cooling trend, which is inconsistent with a long-term weakening of the Walker circulation. A difference of linear trends in proxy SST records at the Gilbert Islands and the Line Islands (to the east) is presented as evidence of a long-term weakening trend in the zonal SST gradient and thus Walker circulation. However, those additional sites also appear to be influenced by the 1976/1977 interdecadal shift (Figure 8b of Carilli et al. [2014]), so the difference in trends computed over this period between the Gilbert and Line Islands may be indicative of differences in the local amplitude of the SST response to a single pan-Pacific event, rather than of differences in real long-term linear trends. Indeed, the SST spatial loading pattern associated with the PDO is such that a higher-amplitude response is expected at the longitude of the Line Islands than at the Gilbert Islands [Mantua et al., 1997; Deser et al., 2010]. Further, it is not inherently clear that the Line Islands are far enough east of the Gilberts to characterize the Walker circulation vis-à-vis the basin-scale zonal SST gradient. The Gilbert Islands chain crosses the equator at 174°E (although they are incorrectly marked east of the date line in Figure 1 of Carilli et al. [2014]). At ~157°W, the Line Islands are ~29° longitude east of the Gilbert Islands, which is equivalent to 24% (31%) of the distance (decrease in climatological SST) between the center of the warm pool (~150°E) and the heart of the cold tongue near the Galapagos (90°W) [Reynolds et al., 2002]. In general, the closer two sites lie along a broad spatial gradient, the smaller the signal-to-noise ratio to be expected. As Carilli et al. [2014] do not show explicitly that the SST difference between the Gilbert and Line Islands is a sufficient representation of the basin-scale SST gradient or Walker circulation, there is a possibility that SST gradients at this spatial scale simply represent local or regional scale variability rather than a physically meaningful proxy for the state of the coupled ocean-atmosphere system. Fortunately, such an analysis of observed SST [Reynolds et al., 2002] and atmospheric circulation [Kanamitsu et al., 2002] reveals that the difference in SST between the Gilbert and Line Islands (ΔSST) does serve as a faithful proxy for the strength of the Walker circulation (Figures 2a and 2b), but the slope of this dependence makes the magnitude of Walker circulation trend suggested by Carilli et al. [2014] quite worrisome. Based on linear least squares regression, we should expect a 0.70°C change in Butaritari-Kiritimati ΔSST for a 1 standard deviation (σ) change in the strength of the Walker circulation (Figure 2c). The Butaritari-Kiritimati proxy ΔSST trend presented by Carilli et al. [2014] of −1.28°C per 26 years suggests a nearly 2σ weakening of the Walker circulation—roughly equivalent to the breakdown of the Walker circulation at the height of the 1982–1983 El Niño event. Interpreting this as a long-term trend implies that the Walker circulation is weakening at a rate of 7σ per century, bearing in mind that σ here includes ENSO variability. Using observed regression slopes and proxy ΔSST trends for Butaritari-Tabuaeran and Butaritari-Palmyra yields Walker circulation trends of –9σ and –15σ per century, respectively. Such long-term trends are probably unrealistic. Finally, for the period of 1982–2013, over which we have very high confidence in SST observations, the sign of the trend in the observed Butaritari-Kiritimati ΔSST (and thus Walker circulation strength) is opposite from that suggested by Carilli et al. [2014]. For the exactly overlapping period of high-quality satellite and in situ observations [Reynolds et al., 2002] and proxy-based ΔSST (1982 through 1997), the observed trend is +0.03°C/decade, while the proxy-based trend is −0.19°C/decade. Since the proxy-based ΔSST record fails to capture even the sign of the observed trend over a period during which we have the highest confidence in the observational record, caution should be used when using Sr/Ca-based proxies to interpret trends in earlier periods. In summary, while the Carilli et al. [2014] record may serve as valuable confirmation of existing instrumental estimates of certain features of the climate record over the past half century such as the 1976/1977 shift and major ENSO events, their presentation as evidence of a long-term slowdown of the Walker circulation is in fact the opposite of what the data show when taking into consideration previously documented natural, interdecadal variability, and opposite of what well-constrained observations indicate has taken place over the satellite era. Also, at 3.07°N, Butaritari does not lie within the EUC, so coral records from that atoll cannot inform us of this key aspect of tropical ocean circulation. Moreover, a recent study using the extended SODA ocean velocity data set revealed a robust strengthening of the EUC over the historical period of 1871–2008 [Drenkard and Karnauskas, 2014], which is significant in the vicinity of the Gilbert Islands. By the best available in situ measurements and historical estimates, the EUC has indeed been strengthening at a rate comparable to that predicted for the future in the latest fully coupled climate model simulations. It remains quite reasonable to expect important consequences thereof for near-shore temperatures, seawater chemistry including nutrients, coral geochemistry, and the overall marine ecosystem response to climate change.