By Judith Curry
“We are in the uncomfortable position of extrapolating into the next century without understanding the last.” – Walter Munk
Part II provided an overview of the relatively recent geological evidence for sea level variations. Part III considers the historical observations for sea level rise in the 19th and 20th centuries, focusing on an evaluation of conclusions in the IPCC AR4 and AR5 in the context of recent research.
The tide gauge record
Tide gauges measure the variations of sea level relative to a geodetic benchmark, a fixed point of known elevation above mean sea level. Some tide gauge records extend back to the eighteenth century. Reconstruction of global sea levels since the late 1800s (until the satellite era beginning in 1993) relies uniquely on the tide gauge record.
To reconstruct equilibrium sea level changes from tide gauges, account must be made of vertical shifts of the land, caused by geological processes or land use (e.g. ground water extraction). To improve scientific utility for sea level studies, numerous modern tide gauges are being augmented with automated, continuous GPS measuring instruments which records vertical land movements. Further, account must be made of non-eustatic dynamic changes in sea level due to tides, storm surges, tsunamis and large-scale ocean currents.
Further, tide gauge technology has changed over time. Simple wooden staffs have evolved into higly sophisticated digital equipment — it is likely that the results from different equipment might not agree with each other.
A wooden staff is not going to measure with the same degree of accuracy-or under the same circumstances as a digital equipment.
Tide gauges have the following disadvantages for determining global sea level changes: uneven distribution around the world; missing data; spatial and temporal variations in ocean circulations; and land movements. Because of these disadvantages, calculating global mean sea level rise from the limited tide gauge network has proven to be difficult.
Although considerable progress has been made, further improvements to the historical record are still needed, particularly in accounting for ocean circulation changes.
A complementary method for determining and/or evaluating global sea level rise is a budget analysis that adds together the cumulative effect of the main contributors to sea level rise: thermal expansion, melting of ice in glaciers, ice loss from the Greenland and the Antarctic ice sheets, and changes in land water storage. If the budget analysis matches the tide guage analysis, then the sea level budget is closed. Any mismatch provides an estimate of the uncertainty of the estimates of sea level change.
A major source of uncertainty rests in the methodology of determining trends (velocity) and acceleration from the tide gauge records of sea level. In particular, trends determined via simple linear regression and acceleration determined through simple quadratic fits are extremely limited and likely to be unduly influenced by the particular time slice.
Visser et al. (2015) reviewed the different trend models applied to sea level data to assess sea level rise and its acceleration-deceleration. They reviewed 30 methods, each having its individual mathematical formulation, flexibilities, and characteristics. They found that just by choosing a different model, one can find decelerations where another trend model shows a linear or even accelerating pattern. They conclude that uncertainties should be taken into account to prevent biased or wrong conclusions and that removing internally generated climate variability by incorporating atmospheric or oceanographic information helps to uncover externally forced climate change signals. No scientific consensus has been reached yet as to how a possible acceleration could be separated from intrinsic climate variability in sea level records.
A recent Ph.D thesis by Phil Watson of the University of New South Wales (LINK) developed improved techniques to estimate mean sea level velocity and acceleration from long ocean water level time series. This task involved extensive time series analysis that identified Singular Spectrum Analysis (SSA) as an optimal analytic for resolving estimates of mean sea level from long tide gauge records with improved accuracy and temporal resolution, since it provides a superior capability to separate key time varying harmonic components of the time series.
Has sea level rise accelerated since the 19th century?
Part II examined the century-to millennial scale sea level record. Both Kopp et al. (2016) and Grinsted et al. (2009) found the lowering sea levels of the Little Ice Age reversed to rising sea levels around 1800, although those datasets have coarse resolution.
Global mean sea level (GMSL) has risen about 8 inches during the 20th century, although we will see in the next section there is significant uncertainty surrounding this number. Part of the argument for an impact of human caused warming on sea level rise (SLR) is an acceleration in the rate of SLR since the 19th century.
From the IPCC AR5:
AR4 concluded that there was “high confidence that the rate of global sea level rise increased from the 19th to the 20th century” but could not be certain as to whether the higher rate since 1993 was reflective of decadal variability or a further increase in the longer-term trend. It has been clear for some time that there was a significant increase in the rate of sea level rise in the four oldest records from Northern Europe starting in the early to mid-19th century. The results are consistent and indicate a significant acceleration that started in the early to mid-19th century (Woodworth 1990, 1999), although some have argued it may have started in the late 1700s. The increase in the rate of sea level rise at Stockholm (the longest record that extends past 1900) has been based on differencing 100-year trends from 1774–1884 and 1885–1985. The estimated change is 1.0 [0.7 to 1.3] mm yr–1 per century (calculated by Woodworth 1990).
From the abstract of Woodworth (1990):
In general, no evidence was found for MSL accelerations significantly different from zero over the period 1870 to the present, although non-zero accelerations were observed at individual stations. In order to extend the study to time-scales longer than a century, data from the oldest European MSL records at Brest, Sheerness, Amsterdam and Stockholm starting in 1807, 1834, 1700 and 1774, respectively, were also investigated with the result that a positive acceleration of order 0·4 (mm year−1) per century appears to be typical of European Atlantic coast and Baltic MSL over the last few centuries.
A recent paper by Phil Watson (2017) has assessed sea level rise and its acceleration from European tide guage data, using the singular spectrum analysis approach. Watson examined the four longest European records (Amsterdam, Netherlands, 246 yrs; Stockholm, Sweden, 214 yrs; Brest, France, 208 yrs; and Swinoujscie, Poland, 202 yrs). The analysis showed velocities and accelerations that vary over time and that that relative velocity is steadily increased over time, peaking at or near the recent end of the time series record. For each of the 4 records, the acceleration is predominantly confined to a narrow band within ±0.05 mm/year2 and not statistically different from zero at the 95% confidence level for most of the records, despite evidence that relative velocities are continuing to increase.
Consideration of the entire European tide guage dataset with records at least 80 years long (83 total stations) shows a broad pattern of acceleration centered in bands around 1880 to 1910, 1940, and 1976 and a strong spatially coherent signal between 1994 and 2000.
Watson concluded that at the 95% confidence level, that there is no consistent or compelling evidence (yet) that recent rates of rise are higher or abnormal in the context of the historical records available across Europe. Watson states that it is likely a further 20 years of data will distinguish whether recent increases are evidence of the onset of climate change–induced acceleration.
A challenge with the long time series methods is not only changes in measuring technology, but the gauge is rarely continuously located in the same place. The Brest tide gauge has moved several times, with the last move occurring after it was destroyed by Allied bombing in 1944 when it was dormant for some 10 years.
20th century sea level rise
Here is a summary of the status of 20th century sea level estimates from the tide guage observations at the time of the 2013 IPCC AR5. From Chapter 3 of the AR5:
Tide gauges with the longest nearly continuous records of sea level show increasing sea level over the 20th century. There are, however, significant interannual and decadal-scale fluctuations about the average rate of sea level rise in all records. Different approaches have been used to compute the mean rate of 20th century global mean sea level (GMSL) rise from the available tide gauge data: computing average rates from only very long, nearly continuous records; using more numerous but shorter records and filters to separate nonlinear trends from decadal-scale quasi-periodic variability; neural network methods; computing regional sea level for specific basins then averaging; or projecting tide gauge records onto empirical orthogonal functions (EOFs) computed from modern altimetry or EOFs from ocean models. Different approaches show very similar long-term trends, but noticeably different interannual and decadal-scale variability. The rate from 1901 to 2010 is 1.7 [1.5 to 1.9] mm yr-1, which is unchanged from the value in AR4. Rates computed using alternative approaches over the longest common interval (1900–2003) agree with this estimate within the uncertainty.
Since publication of the AR5 with its highly confident assessment of a very likely mean sea level rise rate between 1900 and 2010 of 1.7 [1.5 to 1.9] mm yr-1, or 1.5 ± 0.2 mm yr-1 from 1900 to 1990, the following global mean sea level rise estimates have been published:
- Jevrejeva et al (2014): 1.9 ± 0.3 mm yr-1 (20th century)
- Kopp et al. (2016): 1.4 ±0.2 mm yr-1 (20th century)
- Mitrovica et al. (2015): 1.2 ±0.2 mm yr-1 (1900–1990)
- Hay et al. (2015): 1.2 ±0.2 mm yr-1 (1900–1990)
- Thompson et al. (2016): 1.7 mm yr-1 ± 0.3 (20th century); less than 1% probability less than 1.4 mm yr-1
- Dangendorf et al. (2017): 1.1 ± 0.3 mm yr-1 (1900–1990)
A translation of these rates into inches per century: 1.9 mm yr-1 equals 7.5 inches; 1.5 mm yr-1 equals 6 inches; 1.1 mm yr-1 equals 4 inches.
A useful graph comparing the time series of different sea level rise analyses is provided by Klaus Bitterman in a post at RealClimate:
While there is some difference in these numbers associated with ending the period in 1990 or 1999, the major discrepancies relate to tide guage selections, vertical land motion corrections, area weighting and statistical analysis methods.
The important point here is that recently published rates of SLR of 1.1 and 1.2 mm yr-1 are outside of the very likely confidence interval in the IPCC AR5.
Closing the SLR budget in the 20th century
The very substantial differences in estimated rates of 20th century sea level rise, plus the confounding factor of internal variability, have significant implications for attribution analyses and also for projections of 21st century sea level rise.
There are two integral constraints that can be examined in assessing the credibility of sea level rise estimates:
- Analyses of the mass and steric components of the sea level budget;
- Constraints from earth rotation rate and length of day on the mass flux from ice and glaciers
Munk (2002) addressed the imbalance in estimates of 20th century sea level rise in context of these integral constraints. ‘Munk’s enigma’ refers to:
. . . the historic rise started too early, has too linear a trend, and is too large. Melting of polar ice sheets at the upper limit of the Intergovernmental Panel on Climate Change estimates could close the gap, but severe limits are imposed by the observed perturbations in Earth rotation.
Among possible resolutions of the enigma are: a substantial reduction from traditional estimates of 1.5–2 mm/y global sea level rise; a substantial increase in the estimates of 20th century ocean heat storage; and a substantial change in the interpretation of the astronomic record.
The issue of the sea level rise budget as a constraint on uncertainty in the tide guage assessments of sea level rise received little mention in AR4 or AR5. Further, the Munk (2002) paper on the enigma of the twentieth century sea level rise was not cited in either the AR4 or AR5.
Table 13.1 in the AR5 finds a global mean sea level rise budget imbalance of 0.5 [0.1 to 1.0] mm yr-1 for the period 1900-1990. The AR5 concluded that the observational sea level budget cannot be rigorously assessed for 1901–1990, due to insufficient observational information to estimate ice sheet contributions with high confidence before the 1990s; in addition ocean data sampling is too sparse to permit an estimate of global-mean thermal expansion before the 1970s.
Since publication of the AR5, the topic of balancing the budget of sea level rise during the period 1900-1990 has been the subject of several papers (note: the SLR balance is much more straightforward in the satellite era since 1993, which is the subject of Part IV).
Jevrejeva et al. (2016) have prepared a review article on the 20th century sea level budget, clarifying the outstanding issues.
Gregory et al. (2013) published an overview of estimates of individual sea level contributions over the twentieth century. The range of possible sea levels obtained by combining all individual estimates in various combinations (total 144 combinations) suggested that the observed sea levels lie at the very edge of the range and a residual trend is needed to make up for the discrepancy, selecting the largest or smallest estimates for individual contributors. Gregory et al. concluded that if the residual trend can be interpreted as a long-term Antarctic contribution, the budget can be satisfactorily closed.
Hay et al. (2015) argue that rates of sea level rise between 1.0 and 1.4 mm yr-1 close the sea-level budget for 1901–1990 as estimated in AR5, without appealing to an underestimation of individual contributions from ocean thermal expansion, glacier melting, or ice sheet mass balance.
A study by Mitrovica et al. (2015) has demonstrated that the combination of lower estimates of the GMSL rise between 1900 and 1990 (~1.2 mm yr-1), improved modeling of the GIA process and the signal due to core-mantle coupling in ancient eclipse observations resolves ‘Munk’s enigma.’
In summary: Values of GMSL on the lower end (e.g. 1.1 or 1.2 mm yr-1) come closest to balancing sea level budgets for the period 1900 — 1990. These relatively low values of GMSL are well outside the AR5 very likely range of 1.5 ± 0.2 mm yr-1 for the same period.
Internal variability and sea level rise
Assessing the impact of internal variability in ocean circulations on global and regional sea level rise is essential for detecting and attributing sea level rise to human caused climate change and predicting future change, including any acceleration in SLR that can be attributed to humans. Separating the effects of natural climate modes and anthropogenic forcing, however, remains a challenge and requires identification of the imprint of specific climate modes in observed sea level change patterns.
From the AR5:
A long time-scale is needed because significant multidecadal variability appears in numerous tide gauge records during the 20th century. The multidecadal variability is marked by an increasing trend starting in 1910–1920, a downward trend (i.e., leveling of sea level if a long-term trend is not removed) starting around 1950, and an increasing trend starting around 1980.
Although the calculations of 18-year rates of GMSL rise based on the different reconstruction methods disagree by as much as 2 mm mm yr-1 before 1950 and on details of the variability (Figure 3.14), all do indicate 18-year trends that were significantly higher than the 20th century average at certain times (1920–1950, 1990–present) and lower at other periods (1910–1920, 1955– 1980), likely related to multidecadal variability.The IPCC AR5 found that it is likely that a sea level rise rate comparable to that since 1993 occurred between 1920 and 1950.
There are several recent analyses of rates of sea level rise over the 20th century:
- Hay et al. (2015) find that average rates of sea level rise (15 year averages) circa 1940 were of comparable magnitude to values at the end of the 20th century.
- Hamlington and Thompson (2015) identified the negative rates of sea level rise during the 1950’s in the Hay et al. analysis as primarily associated with inclusion of high latitude tide gauges.
- Dangendorf et al. (2017) identified peak in rate around 1940, with values slightly exceeding 2 mm yr-1
Klaus Bitterman has compared recent analyses of rates of sea level in a post at RealClimate:
All of the analyses find a mid century peak in the rate of sea level rise, although most of the analyses find the current peak to be higher. The mid century peak in sea level rise is believed to be associated with multi-decadal internal variability.
A recent paper by Han et al. (2016) reviews our current state of knowledge about spatial patterns of sea level variability associated with natural climate modes on interannual-to-multidecadal timescales.
Over the Pacific Ocean, the PDO and NPGO are the two dominant climate modes and they are associated with global signatures and distinct spatial patterns of sea level changes. Over the Indian Ocean, sea level trend patterns since the 1960s are driven primarily by surface winds. On decadal timescales, wind stress associated with ENSO and IOD is the major cause for decadal sea level variability north of 20°S.
Over the Atlantic, the NAO-associated sea level patterns exhibit a dipole structure in the North Atlantic basin. The upward trend of the AMO SST index during recent decades coincides with the observed accelerated SLR along the US northeast coast. Over the Arctic, significant correlations between the AO index and tide gauge records have been found, but with distinct regional variations. Finally, the SAM can have a significant influence on sea level in the Southern Indian and Pacific Oceans. Zonal asymmetry in SAM-associated winds might have contributed to the asymmetry of decadal sea level variations in the southern ocean during most of the twentieth century.
The short records of available datasets limit our ability to detect the full character of decadal climate variability. Modeling studies suggest that we need ~500 years of observations to sample the full range of ENSO decadal variability (e.g., Wittenberg 2009).
Chambers et al. (2012) investigated whether there is a quasi-periodic 60 year signal in the sea level rise data:
We find that there is a significant oscillation with a period around 60-years in the majority of the tide gauges examined during the 20th Century, and that it appears in every ocean basin. Averaging of tide gauges over regions shows that the phase and amplitude of the fluctuations are similar in the North Atlantic, western North Pacific, and Indian Oceans, while the signal is shifted by 10 years in the western South Pacific. The only sampled region with no apparent 60-year fluctuation is the Central/Eastern North Pacific. The phase of the 60-year oscillation found in the tide gauge records is such that sea level in the North Atlantic, western North Pacific, Indian Ocean, and western South Pacific has been increasing since 1985–1990. Although the tide gauge data are still too limited, both in time and space, to determine conclusively that there is a 60-year oscillation in GMSL, the possibility should be considered when attempting to interpret the acceleration in the rate of global and regional mean sea level rise.
Summary and conclusions
Around the beginning of 19th century, sea levels began to rise, after several centuries associated with cooling and sea level decline. There are only a few historical tide guage records that extend back to 1800, with several along European coasts. Improved time series analysis methods do not support the statistical significance and likelihood levels of the IPCC’s conclusion that sea level rise has accelerated in the 20th century relative to the 19th century.
Recent analyses of 20th century sea level rise find significantly lower values than were cited in the IPCC AR4 and AR5. These lower values between 1900-1990 are more consistent with integral constraints from mass budget analyses. These lower rates of sea level rise have major implications for the assessment of sea level rise and its acceleration in the satellite era since 1993 and also for the baseline scenario of 21st century sea level rise.
There is substantial multi-decadal internal variability in the sea level change record, including an apparent ~60 year oscillation. This variability confounds analyses of sea level rise acceleration and attribution to human caused climate change.
In view of the multiple modes and periods of internal variability in the ocean, it is likely that we have not detected the full scale of internal variability effects on regional and global sea level change.
An assessment of the current state of understanding of historical sea level rise is provided in a recent proposal from an international team of sea level experts (including many of the scientists whose work is referenced in this report): Towards a unified sea level record: assessing the performance of global mean sea level reconstructions from satellite altimetry, tide gauges, paleo‐proxies and geophysical models Excerpts:
In contrast, before the altimetry era, direct estimates of GMSL changes rely on the coastal network of tide gauges that provide in situ observations of sea level relative to the land. These long‐term sea level observations show that GMSL rose slower during the 20th century. Depending on how tide gauge records are combined, rates range between 1.3 and 2.0 mm/year with uncertainties that have been estimated at about an order of magnitude smaller. The differences in rates at multi‐decadal and centennial time scales between the various reconstructions are likely to arise from: (i) the selection of the tide gauge stations; (ii) the corrections (or lack of) of the VLM of the Earth’s crust and of the geoid deformations induced by mass load changes that affect local tide gauge locations ; (iii) the methodologies used to merge a limited number of tide gauge stations or proxy records into a GMSL series; and (iv) the limited ability of all methods to separate the externally‐forced long‐term changes from the low‐frequency fluctuations associated with internal climate variations. However, a complete understanding of the relative importance of each of those factors is lacking. This severely limits our ability to provide a unique estimate of GMSL change at century to millennial time scales and thus restrict the historical interpretation of the altimetric GMSL record.
After considering the evidence used in the AR5 in light of recent research and the above conclusion from the international team of sea level experts, I would like to make some comments on the IPCC assessments of likelihood. From the IPCC guidance regarding treatment of uncertainties:
- Virtually certain: >99% probability
- Very likely: >90% probability
- Likely: >66% probability
- More likely than not: >50% probability
- As likely as not: 50% probability
The AR4 states:
There is high confidence that the rate of sea level rise has increased between the mid-19th and the mid-20th centuries. This conclusion is based on an extremely small amount of data, neglect of the role of internal variability, and a highly simplified analysis method. Watson finds no statistically significant acceleration in the long time series European tide gauges. The AR4’s ‘high confidence’ is completely unjustified.
Consider the following statements cited in Chapter 3 of the AR5:
It is virtually certain that globally averaged sea level has risen over the 20th century. In this case, virtually certain is defensible.
[SLR]with a very likely mean rate between 1900 and 2010 of 1.7 [1.5 to 1.9] mm yr-1. Recent estimates of SLR over this period of 1.1 to 1.2 mm/yr, that better align with integral constraints, are substantially lower that the AR5 values, make a mockery of the very likely likelihood. Taking several published estimates that show some sort of general agreement to produce a very likely conclusion, while ignoring the very substantial uncertainties in these estimates, should not form the basis for a very likely conclusion.
It is likely that sea level rise throughout the Northern Hemisphere has also accelerated since 1850. They consider three estimates, with two showing acceleration and the third showing none. This does not justify likely (>66%), especially given the very sparse data and the unknowns surrounding multi-decadal and longer ocean oscillations. Given the available information, a likelihood of more likely than not would be the appropriate conclusion. Given the uncertainties, as likely as not would be appropriate.
Very simply, the IPCC authors weigh the existing evidence and make their conclusions based on this information. They ignore the space of missing, ambiguous and uncertain information in their likelihood assessments. As a result, we see that some ‘very likely’ and ‘likely’ conclusions from the AR4 and AR5 that are ‘very likely’ to be overturned by the AR6.
Part IV (forthcoming): Sea level variations in the satellite era. This Part will address the insights and challenges of global sea level from satellites, including interpretation of the sea level record since 1993, SLR acceleration and arguments for and against detection of human-caused sea level rise.