Intermittent sea-level acceleration
Introduction
In view of their impact on coastal hazard and society, the problems of secular sea-level rise and of future sea-level trends are the subjects of extensive research (see e.g. Bindoff et al., 2007, Rahmstorf, 2007, Cazenave and Remy, 2011). There is now a general agreement about the global mean sea-level rise (GMSLR) that occurred during the 20th century (see Table 1 of Spada and Galassi, 2012). However, two related climate issues are still debated. The first is the amplitude of the global sea-level acceleration (GSLA) observed during the last centuries and the second is the possible existence of “change points” or “times of inflection” in global reconstructions or in individual tide gauge (TG) records, possibly corresponding to regime shifts of sea-level change. The importance of these issues, both on a regional and on a global perspective, is discussed in the review by Woodworth et al. (2009).
In a seminal work, Douglas (1992) estimated the GSLA by averaging the sea-level accelerations obtained from individual records of globally distributed TGs. GSLA is defined as twice the quadratic term in a polynomial regression within a limited span of time (henceforth, specific values of GSLA and their uncertainty will be simply denoted by a and ∆a, respectively). The approach of Douglas (1992), similar to that adopted by Douglas (1991) to estimate the secular GMSLR, only provided weak evidence in support of a GSLA, even for the longest period considered (namely a ± ∆a = (0.001 ± 0.008) mm/yr2 during 1850–1991). This neatly contrasted with the significant GSLA predicted to accompany greenhouse warming. The negative result of Douglas (1992) confirmed that of Woodworth (1990), who limited his attention to European records. No acceleration was observed also by Wenzel and Schröter (2010). They reconstructed mean sea level from TGs data (1900–2006) using neural networks although the dataset was then restricted to the period 1950–2006 to prevent the drastic reduction of available data during the first half of the century.
Recent studies, either based on the “virtual station” stacking method (Jevrejeva et al., 2006, Jevrejeva et al., 2008) or on a sea-level reconstruction of long TG records (Church and White, 2006, Church and White, 2011), unanimously point to the existence of a GSLA. Based on a ~ 300-year-long time series (1700–2002) obtained by combining short and long TG records, Jevrejeva et al. (2008) reported a GSLA of about a = 0.01 mm/yr2 (the uncertainty was not quantified), which apparently started at the end of the 18th century. The Empirical Orthogonal Function (EOF) approach of Church and White (2006), combined with polynomial regression, suggested GSLA of (0.013 ± 0.006) mm/yr2 in the period 1870–2001 and of (0.008 ± 0.008) mm/yr2 when the 20th century only is considered. In the follow-up paper of Church and White (2011), the acceleration (0.009 ± 0.003) mm/yr2 has been proposed for the time period 1880–2009. Sea-level curves previously presented in the literature or obtained in this study are shown in Fig. 1.
The spread of previous GSLA estimates based on tide gauge (TG) records, summarized in Table 1, is significant. The large energy of decadal sea-level fluctuations (Jevrejeva et al., 2006, Chambers et al., 2012, Houston and Dean, 2013), the poor geographical coverage of TGs, the limited number of TGs facing the open seas (hence less affected by coastal processes), and the oceans response to regional changes in the pattern of wind stress (Bromirski et al., 2011, Merrifield, 2011, Sturges and Douglas, 2011) are the main causes of uncertainty and potential sources of misinterpretation (see also the discussion in Douglas, 1992, Sturges and Hong, 2001). As recently evidenced by Gehrels and Woodworth (2013) and by a number of previous studies, the proposed GSLA value is strongly sensitive to the time span of the instrumental record considered and to additional selection criteria based on the quality of the data set. Spurious effects from gappy time series (Wenzel and Schröter, 2010), contaminating tectonic (e.g. Larsen et al., 2003, Olivieri et al., in press) or anthropogenic factors (Carbognin et al., 2010) act to further complicate the determination of GSLA.
The constant acceleration model for sea-level rise is appealingly simple and constitutes the most obvious generalization of linear models (a = 0) extensively employed to estimate GMSLR since the early determination of Gutenberg (1941) (for a review, see Spada and Galassi, 2012). However, inspection of sea-level compilations (Gehrels and Woodworth, 2013) and of individual records see e.g. (Bromirski et al., 2011), also reveals short-lived accelerations and abrupt steepness variations. These can be modeled, to a first approximation, as change points (CPs) separating periods of constant rate and/or of constant acceleration. As pointed by Church and White (2006), a CP model including an abrupt slope change at year ~ 1930, unexpectedly during a period of little volcanic activity, can indeed be invoked as a possible alternative to a constant acceleration model for the time period 1870–2001. Inflections in global and regional compilations of instrumental records at year ~ 1930 have also been proposed by Jevrejeva et al. (2008), Woodworth et al. (2009) and Church and White (2011). Based on proxy and instrumental observations from seven sites, Gehrels and Woodworth (2013) have recently proposed that year 1925 (± 20) could mark the date when sea-level rise started to exceed the long-term Holocene background rate. Inflections or CPs occurring during the 19th century could be more difficult to ascertain in view of the limited amount and sparsity of instrumental data available for that epoch. However, a major acceleration episode has been evidenced by Jevrejeva et al. (2006) during 1850–1870, though its significance was disputed.
Here we provide a new assessment of GSLA based on instrumental (TG) data alone, for the time period 1820–2010. Assuming a constant acceleration model, from a cumulative sea-level curve constructed by TG time series of sufficient length, we obtain GSLA values that are generally consistent with earlier estimates. However, by simple statistical methods, we address in a systematic manner the important role played by non-synchronous CPs at individual TGs in the assessment of the GSLA. Section 2 is devoted to the construction and to the analysis of a global sea-level curve. The results are then discussed in Section 3.
Section snippets
Building a global sea-level curve
In Fig. 1, curves (a) and (b) reproduce the sea-level time series constructed and studied by Jevrejeva et al. (2006) and by Church and White (2006), respectively. The corresponding GSLA values are given in Table 1. The figure also shows an additional curve (c) that we have built by a global stacking of the 315 Revised Local Reference (RLR) annual time series with length ≥ 50 yr, currently available from the Permanent Service for Mean Sea Level (PSMSL) for the time period 1810–2010 (Woodworth and
Discussion and conclusions
Unweighted stacking of the longest RLR annual TG time series produces a synthetic global sea-level curve (ST), which shows several features. First, ST shows a statistically significant and positive CP, implying a sudden increase in slope, within the time period 1835–1840. Second, branch ST2 of curve ST, which encompasses the time period 1840–2010, is best fitted by a quadratic polynomial (α = 95%). This confirms previous results about the existence of a GSLA for the period 1840–2010 (Church and
Acknowledgments
We are grateful to two anonymous reviewers who have provided very thoughtful comments on an earlier version of the manuscript. We have benefited of some econometrics suggestions from Barbara Petracci and Pierpaolo Pattinoni and from the insightful discussions with Fabio Raicich and Florence Colleoni. Sea-level data have been downloaded from the PSMSL (Permanent Service for Mean Sea Level) archive on August 1st, 2012 (http://www.psmsl.org/data/obtaining/). All figures have been drawn using the
References (44)
- et al.
When did modern rates of sea-level rise start?
Global and Planetary Change
(2013) - et al.
SELEN: a Fortran 90 program for solving the “Sea Level Equation”
Computers and Geosciences
(2007) - et al.
Decadal variability of sea level
- et al.
The dynamic response of reef islands to sea-level rise: evidence from multi-decadal analysis of island change in the central pacific
Global and Planetary Change
(2010) - et al.
Relative sea levels of Japan from tide-gauge records
Geological Society of America Bulletin
(1986) - et al.
Modeling regional sea level rise using local tide gauge data
Journal of Geodetic Science
(2012) - et al.
Observations: oceanic climate change and sea level
- et al.
Estimating rates of acceleration based on the 157-year record of sea level from San Francisco, California, U.S.A.
Journal of Coastal Research
(2013) - et al.
Dynamical suppression of sea level rise along the Pacific coast of North America: indications for imminent acceleration
Journal of Geophysical Research
(2011) - et al.
Global change and relative sea level rise at Venice: what impact in term of flooding
Climate Dynamics
(2010)