Nineteenth and Twentieth Century Changes in Sea Level

. Following the Last Glacial Maximum (25,000–20,000 years ago), sea level rose at rates on the order of several tens of millimeters per year at times, and increased overall by over 130 m. However, melting of the great ice sheets was largely complete by 6,000 years ago, and it is believed that sea level did not rise significantly again until recently. The rates of sea level change during the last few centuries and in recent decades can be measured in units of millimeters per year and are particularly important in understanding present-day climate change. We now have a range of techniques with which sea level changes can be measured and thus studied more intensively than before, as a global average and in each region. This article introduces each of the main data sets and presents the primary research findings. It is hoped that a greater understanding of the reasons for past observed sea level change, discussed elsewhere in this issue, will lead to better estimation of the changes likely to occur in the future.

. installed by party off of Surveyor. Photo credit: C&GS Season's Report Sobieralski 1924-93;from http://www.photolib.noaa.gov/ htmls/theb2385.htm (top right) Tide gauge in Knik harbor, alaska, 1918. installed by party off of Eoline Hand. Photo credit: C&GS Season's Report Hand 1918-32;from http://www.photolib. noaa.gov/htmls/theb2366.htm (bottom left) Tide gauge in the aleutian islands, alaska, 1952. installed by party off of Explorer. Photo credit: Family of Captain George L. Anderson, C&GS;from http://www. photolib.noaa.gov/htmls/theb2394.htm S P e C i a L i S S u e o N S e a L e V e L Oceanography | Vol.24, No.2 aBSTr aCT. Following the Last Glacial Maximum (25,000-20,000 years ago), sea level rose at rates on the order of several tens of millimeters per year at times, and increased overall by over 130 m. However, melting of the great ice sheets was largely complete by 6,000 years ago, and it is believed that sea level did not rise significantly again until recently. The rates of sea level change during the last few centuries and in recent decades can be measured in units of millimeters per year and are particularly important in understanding present-day climate change. We now have a range of techniques with which sea level changes can be measured and thus studied more intensively than before, as a global average and in each region. This article introduces each of the main data sets and presents the primary research findings. It is hoped that a greater understanding of the reasons for past observed sea level change, discussed elsewhere in this issue, will lead to better estimation of the changes likely to occur in the future.

Sea LeVeL meaSuremeNTS By Tide GauGeS
The earliest extended sea level measurements were made in Europe during the eighteenth century. These data were visual observations of the heights and times of high and low waters, or sometimes high waters alone. Many entrances to docks (or to "sluices, " connections between the sea and inland waters) were equipped with what were then called "tide gauges, " graduated markings on their stone walls to indicate water depth over the dock sill. Observations of water level could be made using those markings. Alternatively, wooden measuring rods called "tide-poles" or "tide-staffs" were used.
Visual measurements could have had centimeter-level accuracy in calm weather conditions, but would have been much less accurate in the presence of waves, especially at night during winter.
A "stilling well" enables a more accurate measurement of the "still water level" (Moray, 1665). This technique uses a vertical tube with a small opening at the bottom so that the water level inside the tube is the same as that outside. The small opening dampens the water-level fluctuations due to waves. By the 1830s, automatic (or "self-registering") tide gauges had been developed that could record the full tidal curve, not just the high and low waters; the first is often credited to Palmer (1831), with the first in the United States made by Joseph Saxton for the US Coast Survey in 1851.
These instruments again employed a stilling well but also contained a float connected by a wire run over pulleys to a pen that moved up and down as the tide rose and fell, thereby drawing a tidal curve on a rotating drum of paper. All of the PSMSL tide-gauge data (and also geological and archaeological measurements of sea level) are "relative" ones (i.e., relative to the local land) unlike the "geocentric" (relative to Earth's center) ones described below for satellite altimetry. arguably the longest US record, but it contains many gaps; Maul and Martin, 1993). These records indicate rising sea Philip L. Woodworth (plw@pol.ac.uk Tamisiea and Mitrovica, 2011, in this issue). Evidence such as these two records, and many others worldwide, are the basis of our belief that sea level has risen globally during the twentieth century. However, at Sitka, Alaska, sea level is falling at a rate of over 2 mm yr -1 as a result of a combination of GIA and local tectonics. Even higher rates of sea level fall due to GIA can be observed in other records from former glaciated areas (e.g., Hudson Bay, Scandinavia). Meanwhile, the Galveston record shows an extremely high rate of rise (over 6 mm yr -1 ), which has nothing to do with natural geological processes like GIA, but is a consequence of the removal of groundwater and hydrocarbons (Emery and Aubrey, 1991).
Three methods are often used to correct tide-gauge records for VLMs. The first requires GIA to be the dominant geological process at the tide-gauge location, and employs predictions of present-day VLM due to GIA obtained from geodynamic models of Earth's continued response to deglaciation (e.g., Peltier, 2001). In the second method, geological data from the Holocene (last 10,000 years) obtained from sites near the tide gauge are used as the basis of an estimate of the present-day rate of sea level change due to geological processes of whatever cause. It is then possible to infer that any excess in the tide-gauge trend compared to the geological value is due to recent climate-change-related processes. One drawback of this method is that it is restricted to tide-gauge locations where geological information is available. In the third method, geodetic techniques are employed to measure directly the rate of any vertical crustal movement, of whatever origin, and so "correct" the tide-gauge data (e.g., Bouin and Wöppelmann, 2010). The main geodetic technique is the Global Positioning System (GPS), with Absolute Gravity and Doppler Orbitography and Radiopositioning Integrated by Satellite (DORIS) also employed. This method also involves an assumption: the vertical rate over the GPS measurement period (typically a decade) is representative of that over the entire tide-gauge record (typically a century). Measurements of crustal movement as measured by GPS will not, strictly speaking, provide a complete "VLM correction" to the tidegauge data if the main geological process is large in amplitude and spatial scale (like GIA). This is because an additional correction for the corresponding changes to Earth's gravity field is required. If the process is small in amplitude or spatial scale (e.g., submergence due to water or mineral extraction), then this additional correction will be less important.
Most analysts of twentieth-century sea level trends have employed the first method with GIA corrections applied to individual records, and corrected trends combined into regional and quasi-global averages (e.g., Douglas, 1991Douglas, , 1997. that the twentieth-century rise in global sea level was closer to 2 mm yr -1 than 1 mm yr -1 , with values around 1.7 mm yr -1 obtained recently for the past century (Church and White, 2006) or past half century (Church et al., 2004;Holgate and Woodworth, 2004).
Although there may be a consensus regarding twentieth-century globalaverage sea level change, it is clear that far less is known about century-timescale regional rates, especially in the Southern Hemisphere. Correction for Earth's long-term GIA goes some way to explaining the large differences between measured trends in each region (Douglas, 1991), though large spatial variations remain. Some could be due to quite recent changes in hydrological and glaciological loads on the solid Earth (e.g., recent melting of ice sheets and glaciers), resulting in modifications to the gravity field and a spatial "sea level fingerprint" (Mitrovica et al., 2001. Other spatial variations could be due to the way that the ocean adjusts, on many time scales, to changing heat and freshwater fluxes. Two decades of satellite altimeter data have shown that rates can vary considerably regionally on decadal time scales due to changes in the steric (density) composition of the ocean (see below and also Church et al., 2010, andMitrovica, 2011, in this issue). In addition, numerical models have demonstrated considerable differences between rates of sea level change in different regions on century time scales due to the ocean's adjustment to climate change. It is likely, therefore, that our appreciation of the true extent of regional variation in twentieth-century sea level change, due to a number of geophysical and oceanographic causes, has been underestimated because of limitations in the geographical coverage of historical tide-gauge data. Studying the acceleration of sea level rise is an easier task than studying sea level trends if the VLM rate is constant throughout the record. In that case, the second-(quadratic) and higher-order components of a record can be studied as climate signals. This situation clearly applies to a good approximation if the main geological process is GIA, but does not apply when tectonics or anthropogenic effects dominate the VLM. If one applies a simple secondorder fit (a + bt + ct 2 , where t is time) to the longest records available, which are from northern Europe, then quadratic coefficients "c" on the order of 0.005 mm yr -2 are obtained. These figures provide evidence for long-term acceleration in sea level rise and suggest that twentieth-century rise started at around the end of the nineteenth century (Woodworth et al., 2009(Woodworth et al., , 2011. Slightly shorter records, such as those from San Francisco and New York, yield similar "c" values (0.0069 and 0.0038 mm yr -2 , respectively, using data since the 1850s), consistent with the acceleration inferred by Maul and Martin (1993) (Donnelly et al., 2004), the period 1900-1920 (Gehrels et al., 2005), the start of the twentieth century (Gehrels et al., 2008), the period 1880-1920 (Leorri et al., 2008), and the   Woodworth (2004;red), and from a simple average of tidegauge data (yellow). all of these estimates are based entirely or largely on Permanent Service for mean Sea Level tide-gauge data. The satellite altimeter record is shown in black. all series are set to have the same average value over 1960 to 1990, and the reconstruction is set to zero in 1990. From Church and White (2011). Note that altimeter data in this figure are as computed by CSiro, whereas those in Figure 5 are from the university of Colorado; the two time series are essentially the same. Dating methods used to provide chronologies for salt-marsh-based sea level reconstructions include radiometric analyses ( 14 C, 210 Pb) and stratigraphic marker techniques (e.g., 137 Cs, pollen, charcoal, Pb isotopes, metal concentrations). The 210 Pb method is suitable to date sediments younger than ~ 120 years.
Measurements of 137 Cs are used to detect the 1965 level (when nuclear bomb testing was globally at its peak) or local nuclear spill events. High-precision accelerator mass spectrometry (AMS) 14 C techniques, in combination with bomb-spike AMS 14 C dating, can also be used to date recent sediments (Marshall et al., 2007;Kemp et al., 2009). Isotopic ratios of Pb and metal concentrations in the sediment (e.g., Pb, Cu) provide ages for local and regional pollution events that usefully can fill in some dating gaps in the nineteenth century (Gehrels et al., 2006;Marshall et al., 2007). Along coasts in the Southwest Pacific and Northwest Atlantic, pollen analyses can provide markers for the eighteenth and nineteenth centuries by revealing a distinct change in vegetation resulting from the first settlement by Europeans (Gehrels et al., 2005(Gehrels et al., , 2008Kemp et al., 2009).
Proxy sea level records from salt marshes have been presented in two different ways. The first way is to depict proxy records by the "traditional" method and only show established sea level index points in an age-depth plot (Donnelly et al., 2004;Gehrels et al., 2008). A sea level index point is obtained from a sediment sample with a measured age that can be related to a former sea level position. The relationship between a sediment sample and sea level is referred to as the sample's "indicative meaning" and is usually determined by analyzing microfossils (e.g., foraminifera) whose modern counterparts are found in narrow vertical niches within the intertidal zone. The second way of presenting proxy records is to make use of age models (Gehrels et al., 2005(Gehrels et al., , 2006Leorri et al., 2008;Kemp et al., 2009)   Despite these efforts to improve the accuracy and precision of satellite altimetry, there is still the possibility that the instrument might drift after it is on orbit. Tide gauges have proven to be very effective for monitoring the stability of the altimeter instruments after they have been launched (Mitchum, 2000). in the less-accurate tide-gauge record (Nerem et al., 1999. It is thought that this variability is caused by ENSOforced variations in land water storage. Global mean sea level tends to be higher during ENSO events because there is An acceleration of sea level rise has yet to be detected in the altimeter record itself, and it is estimated that 30 years of altimetry will be required before an acceleration can be detected in the presence of the ENSO-related variability (Nerem et al., 1999). However, it may be noted that the rate observed by altimetry over the last two decades, which is comparable to that from tide gauges over a similar period, is nearly double the average sea level trend for the twentieth century. Consequently, there is already some evidence for acceleration relative to the longer term. The acceleration of sea level rise has also been detected in the longer tide-gauge record (Holgate and Woodworth, 2004;White et al., 2005;Prandi et al., 2009;Merrifield et al., 2009). Over the nine years since its launch, this satellite has proven to be capable of monthly measurements of water-mass distribution with a spatial resolution of ~ 500 km anywhere on the planet.

Regional variations in
GRACE has been used successfully to monitor changes in Greenland and Antarctic ice masses (Luthcke et al., 2006;Velicogna, 2009), the melting of glaciers in Alaska (Luthcke et al., 2008) and Patagonia (Chen et al., 2007), global ocean mass , ocean bottom pressure (Morison et al., 2007), continental water storage (Rodell et al., 2009;Tiwari et al., 2009), and a host of other signals.