This paper gives an overview of several important issues concerned with the measurement and understanding of long-term sea level change, which is one of the most important topics in sea level science. The issues discussed span science, technology and our ability to fund and operate the national and global observation networks required for a proper understanding of long-term change and its impacts. Many of these topics were explored at a ‘Celebration of UK sea level science’ at the Royal Society on 16–17 February 2004, and are discussed further in a number of papers in this volume.
Sea level science spans many fields of research including oceanography, geology, climate change, geodesy, coastal engineering and the socio-economics of coastal zones. In addition, sea level data have many practical applications such as in flood warning (e.g. at the UK Storm Tide Forecasting Service at the Met Office) and navigation (via tide tables, hydrographic surveys, real-time charting, etc.). The many scientific and practical applications of sea level data have been described in detail in recent text books and review papers (Pugh 2004; Cazenave & Nerem 2004; Woodworth et al. 2004), and are demonstrated in many of the papers of this volume.
In the present paper, I shall review the status of several important issues concerned with the measurement and understanding of long-term sea level change. Issues include the magnitude of twentieth century sea level rise and the need for accurate land movement corrections to tide-gauge measurements of the rise (§§2 and 3). This leads into a discussion of what has been called the ‘enigma of twentieth century sea level rise’, which concerns how well the observed rise can be explained (§4). Accelerations in the rate of sea level change during the past few hundred years and in the last century are next discussed (§§5 and 6) leading into a consideration of potential changes in sea level during the twenty first century (§7). Research into changes in extreme sea levels, a topic which has received considerably less attention than changes in mean sea level (MSL), is reviewed in §8. The last major issue concerns the need for major investment in the global sea level network and the exploitation of the resulting data sets in combination with those of other technologies (§9). Section 10 provides the conclusions of the paper.
Most of these issues are scientific, while some are technical and others (e.g. the need for more and better data) have a financial or organizational character. Several of the issues are discussed at greater length in other papers in this volume. I have omitted discussion of the coastal impacts of long-term sea level change which are covered elsewhere (Woodworth et al. 2004; Nicholls & Tol 2006).
2. Twentieth century sea level rise
The Intergovernmental Panel on Climate Change (IPCC) third assessment report (TAR; Church et al. 2001) concluded that sea level rose at a rate of 1–2 mm yr−1 during the twentieth century. It is important to realize that this is to some extent a consensus number, with the wide range being a reflection of the rates obtained by different authors. At first sight, this wide range is surprising, given that all authors make use of the same data set from the Permanent Service for Mean Sea Level (PSMSL; Woodworth & Player 2003). However, it is a consequence of three main factors.
The first factor is that the PSMSL data set does not provide uniform geographical coverage. Consequently, authors who make use of different subsets of the data set will inevitably sample differently any genuine spatial variability in the rate of sea level rise. The second factor is that authors apply different corrections for vertical land movement to the relative (to land level) sea level data provided by tide gauges. These two factors are explored below. A third factor, which has never been properly faced (and is not addressed here either), is that data from different parts of the world are not equally as good, and therefore should not be given equal weight in global analyses.
The geographical bias of the PSMSL data set is well known (Gröger & Plag 1993; Woodworth & Player 2003). With a small number of exceptions, the long records of the data set provide information of northern hemisphere coastal sea level change. This would not matter if there were no spatial variation of sea level rise in the ocean itself. However, from TOPEX/Poseidon satellite altimetry we know that major differences in sea level trends can be observed in different regions on decadal timescales, while information from hydrographic data sets and from atmosphere–ocean general circulation models (AOGCMs) suggests that the variability can be significant also over much longer timescales. For example, the HadCM3 AOGCM run using the GSIO scenario (which includes historical greenhouse gases, the direct and indirect effects of sulphate aerosols and of ozone; Church et al. 2001) suggests that the root-mean-square (r.m.s.) of the spatial variability of sea level rise due to thermal expansion during the twentieth century was almost the same as the global-average rise (J. Gregory 2005, private communication). Similar findings can be inferred from other AOGCMs (e.g. see information in tables 2 and 4 of Gregory et al. 2001). In this case, good data would be needed from at least 10 oceanographically independent (uncorrelated sea level variability) regions to provide a standard error of global-average sea level rise for the twentieth century equivalent to 30% of the global-average.
Some authors have investigated the consistency of the sea level trends observed in different regions. For example, Peltier (2001) found trends in different areas to be in agreement at the 20% level, once corrections for glacial isostatic adjustment (GIA) had been made, rather than at the 100% level suggested by HadCM3/GSIO. This narrow range is understandable if factors contributing to sea level change, other than thermal expansion, do not add to spatial variability. Nevertheless, spatial variability within the sparse data set, and consequent potential bias in the derivation of a global-average trend, is an important issue which must always be considered.
Cabanes et al. (2001) raised this issue most strongly, by pointing out that many of the gauges with long records in the PSMSL data set were located in regions for which ocean temperatures, as represented in the hydrographic data set of Levitus et al. (1998, 2000), appeared to have been rising faster than average during the last half century. If correct, this observation would have meant that the real value of global sea level rise during the past century would have been more likely 0.5 mm yr−1, rather than the 1–2 mm yr−1 of Church et al. (2001).
Miller & Douglas (2004, 2006) demonstrated that the Levitus data set contains its own large spatial biases, primarily owing to the algorithms used for interpolation and smoothing of sparse data. They showed that the biases in the hydrographic data set led Cabanes et al. to incorrect conclusions regarding global-average sea level rise and that the best estimate for the twentieth century rise remains near to 2 mm yr−1. However, the Cabanes et al. (2001) study was extremely useful in ‘setting the cat among the pigeons’, by forcing a re-evaluation of the different data sets and by stimulating a constructive debate within the sea level community.
Spatial variations in sea level trends can be employed in a more positive way by exploiting any information content in the geographical variation. Tamisiea et al. (2001), Plag (2006) and Plag & Jüttner (2001) have devised ‘fingerprint’ methods wherein the spatial variations due to different processes might be combined to simulate the observed pattern of sea level trends. Such contributing patterns include those of GIA, recent Greenland, Antarctic and lower latitude glacial change leading to an elastic response of the land, and oceanic thermal expansion, together with a spatially independent signal. For example, a large amount of ice wastage from Greenland would result in an unloading of the land in the region and lower sea level trends being observed in Europe and the UK than the rest of the world (which is in fact the case, Woodworth et al. 1999). Results are encouraging, although estimates of, for example, Greenland wastage derived in this way are not always consistent with IPCC estimates obtained by direct observation or ice modelling (Mitrovica et al. 2001). The spatial and temporal inhomogeneities of the sea level data set, and limitations of the other historical data sets, limit the success of this form of analysis.
To summarize, consensus on twentieth century sea level rise has returned to approximately that expressed in the IPCC TAR (i.e. 1–2 mm yr−1). Within that range, most opinion considers the real twentieth century trend to have been closer to 2 than 1 mm yr−1 (each with about 30% uncertainty), with the lower value tending to be preferred by European authors, partly because of the lower rates observed in European records.
3. The need for land movement corrections
A major difficulty in determining sea level trends from tide-gauge data relates to the fact that the vertical position of the land upon which the gauge is situated (measured relative to the centre of the Earth) could be changing as much as the position of the sea itself. Vertical land movements can arise from geological processes both natural (e.g. long-term tectonics) and anthropogenic (e.g. ground water pumping or mining). However, GIA is the only geological process for which models are available capable of application to the correction of tide-gauge records (e.g. Peltier 2001). Even though the various GIA models are qualitatively similar, their present-day rates of vertical movement contain uncertainties due to imprecise knowledge of the Earth's ice history and of three main geophysical parameters (lithospheric thickness and upper and lower mantle viscosity), in addition to uncertainties due to model resolution and parameterization (e.g. see comments in Woodworth 2003).
Even if the GIA models were perfect, there is still the issue that there could be vertical motion due to other geological processes. In some locations, one can make use of geological information to estimate vertical movement, instead of a GIA model (e.g. Woodworth et al. 1999 and references therein). Peltier (2001) and Woodworth et al. (2004) commented on the reliability of the different methods. However, the correct scientific approach is clearly to measure vertical movement if one can. During the last decade, new geodetic techniques have become available which offer the promise of monitoring land levels at tide-gauge sites. The use of Continuous Global Positioning System (CGPS) receivers is the main technique, while others include Doppler Orbitography and Radiopositioning Integrated by Satellite (DORIS) and absolute gravity (AG). However, with some exceptions (e.g. the use of CGPS in Scandinavia where GIA signals are large, Milne et al. 2001), none of the proponents of the different techniques have as yet claimed them to be capable of providing the sort of accurate vertical data needed for routine combination with tide-gauge data in sea level trend analyses. In fact, in the short term, a major constraint on the effects of GIA in the UK may come from analysis of GPS data in the horizontal rather than the vertical (Milne et al. 2006).
Bingley et al. (2006) present a state-of-the-art review of the use of CGPS and AG in the UK. They show that many improvements have been made in the last few years, particularly since the introduction of GPS in continuous mode (i.e. since the late 1990s), and that interesting data sets are now being assembled. In addition, the international community has organized itself within the TIde GAuge (TIGA) project of the International GPS Service to address fundamental problems associated with the use of GPS at gauge sites. (A particular problem relates to the need for stable reference frames.) Progress in this area is likely to take several years. Consequently, that part of the 1–2 mm yr−1 range of uncertainty of Church et al. (2001) which is related to imprecision in land movement corrections will remain in the short term.
4. The enigma of twentieth century sea level rise
The next issue is concerned with what has been called the ‘enigma of twentieth century sea level rise’ (Walter Munk's words) or the ‘attribution problem’, which relates to our inability to account fully for the 10–20 cm change. This has recently been the subject of much renewed debate (Munk 2002; Meier & Wahr 2002; Douglas & Peltier 2002; Cazenave & Nerem 2004; Woodworth et al. 2004).
Church et al. (2001) discussed in detail the many possible contributors to twentieth century sea level change (thermal expansion, glacier melt, etc.) which sum to only 7 cm, only approximately half of the observed value. However, there are uncertainties associated with each contributor, of which the largest relate to terrestrial water storage fluctuations. These are dominated by anthropogenic factors such as dam construction and groundwater mining, rather than by natural changes to continental hydrological balance. The uncertainties in the various contributing terms, and particularly in terrestrial water, are so large that the 7 cm lies within its own range of uncertainty of −8 to +22 cm, and therefore could be said to be consistent with the 10–20 cm. However, that point of view is not entirely satisfactory; one would have been more content to see the central values of the ranges in closer agreement.
Since the IPCC TAR, the only significant suggestion for closing the gap between the sum of contributors and the observations has been based on the work of Antonov et al. (2002). The suggestion is that global-salinity decreased slightly during 1957–1994 with two important implications. The first is that salinity changes will have compounded the effects of temperature in producing a global-average sea level change due to density changes (‘steric changes’) approximately 10% larger than that computed by temperature alone (i.e. 0.55 compared to 0.5 mm yr−1). The second implication is that, if the salinity decrease is not due to melting sea ice, which to a first approximation does not alter sea level, then a sea level rise of 1.3±0.5 mm yr−1 due to 470±170 km3 yr−1 of fresh water input would be inferred, originating from changes in glaciers, polar ice sheets or continental water balance. Munk (2003) and Wadhams & Munk (2004) provide further discussion of this possibility (with slightly different quantitative interpretation of the Antonov et al. data).
This amount of ‘new water’ is so large that it can account completely for any discrepancies between the sum of contributors and observations reported in the IPCC TAR. However, the problem then is ‘buck passed’ to having to identify the source of this new water. Either one or more of the terms discussed by Church et al. (2001) (e.g. glacier melt at 0.3 mm yr−1) would have to have been significantly underestimated by the TAR, or the 1.3 mm yr−1 value would have to have been an overestimate, perhaps due to part of the ocean freshening being caused by melting sea ice. Wadhams & Munk (2004) discussed the possible range of contribution of sea ice. However, it seems inevitable that one is always destined to be engaged in calculations with large uncertainties.
Woodworth et al. (2004) made the important observation that the main difficulty in accounting for the observed global sea level change arises from the non-simultaneity of measurements of the various contributing terms in addition to their individual uncertainties. Almost always one finds trends quoted for different periods when one knows that one is dealing with considerable inter-annual variability in all parameters. This emphasizes the urgent need for further, and if possible more coordinated, research on each aspect of this topic.
In summary, the ‘enigma’ remains for the moment and it may well be that we shall have to await the improved understanding that comes with better data sets in the future (Woodworth & Gregory 2003).
5. Sea level accelerations on century timescales
Almost as important an issue as accounting for twentieth century sea level rise is knowing when that rise commenced. Consequently, it is essential that some idea is obtained of changes over the last few centuries or millennia, so as to place the more recent changes in a longer term historical context. Fortunately, if one is interested in variations in the rate of sea level change (accelerations) in areas without major tectonics, then one can make an assumption of vertical land movements having a linear time-dependence over century timescales. Then, any accelerations evident in the records can be explained by processes in the ocean rather than on land.
If one is interested in the sea levels of the eighteenth and early nineteenth centuries, then the amount of sea level data available is very limited. ‘Tide-gauge data’ prior to 1800 are all from northern Europe (e.g. Stockholm, Amsterdam, Liverpool), and often take the form of high water values only, often with major uncertainties over datum and timings. An analyst can employ such data as proxy-MSL information, for comparison to present-day MSL, only if there is considered to have been little change in the tide, or if the magnitude of tidal changes can be estimated by some means. In addition, the datum employed for the historical information have to be relatable to present-day benchmark heights. However, there are interesting possibilities for additional historical research. For example, it is possible that data from UK ports such as Bristol could complement those already available from London and Liverpool. In France, the sea level data set from Brest, which (with gaps) spans the entire eighteenth and nineteenth centuries, is currently being reanalysed (G. Wöppelmann 2005, private communication).
As one progresses into the nineteenth century then tide-gauge data become available from other continents (e.g. Maul & Martin 1993), and limited amounts continue to be discovered, most valuably from the southern hemisphere. For example, Hunter et al. (2003) estimated sea level change since 1841 at Port Arthur, Tasmania using tide-gauge data collected over a two year period at the penal settlement. These data were lost to science, until they were rediscovered recently after hard work in archives in the UK and Australia. It is conceivable that a small number of other valuable, historical tide-gauge data sets may eventually be found from other locations which will add to our knowledge of sea level change over the last two centuries.
Without tide-gauge data, one can resort to the ingenious use of proxy-information to infer sea level change. For example, Camuffo & Sturaro (2003) estimated changes in Venice during the last three centuries by comparison of the heights of algae evident in paintings by Canaletto and his pupils to heights observed today. The use of proxy-sea level data from archaeological sources is well established in the Mediterranean, where many Roman and Greek constructions are relatable to the level of the sea (Flemming & Webb 1986), with fish ponds from the Roman period considered to be a particularly reliable source (Lambeck et al. 2004). Historical information for the last 2000 years on water levels in coastal wells, as a proxy for sea level, has been recently investigated in Israel (Sivan et al. 2004). The use of archaeological data in the UK is limited by the smaller number of suitable Roman and other constructions and by the large tidal range. The latter introduces uncertainties regarding the ‘sea level’ relative to which any construction was constructed (MHWS, MHW, MSL, etc.), and whether there is a possibility of tidal changes over the intervening period, especially in estuaries. Akeroyd (1972) seems to have been the last person to have made a serious study of UK archaeological sea level information.
In summary, from the small number of very long tide-gauge records (e.g. Woodworth 1999), from combinations of archaeological and tide-gauge data (e.g. Lambeck et al. 2004), and from combinations of relatively recent geological and tide-gauge data (especially in NE America, see Donnelly et al. 2004; Gehrels et al. 2005), the balance of evidence suggests that the sea level rise observed during the twentieth century was significantly larger than that measured over timescales of several centuries. However, more historical information is clearly required, especially from regions outside Europe and North East America. In addition, without additional insight into climate change, it is not possible to unambiguously explain such an acceleration as due to recent, anthropogenic change, in that levels during the past several thousand years may have oscillated on time-scales of 100–1000 years by up to several decimetres (Church et al. 2001). In particular, the sea levels of the last few centuries, especially those of the northern hemisphere, may have been influenced by the Little Ice Age.
6. Sea level accelerations during the twentieth century
An inspection of any of the main climate indices, such as the El Niño Southern Oscillation (ENSO) and North Atlantic Oscillation (NAO), demonstrates variability on timescales of several years to decades. Inspection of most long sea level records also demonstrates such variations. This ‘inter-annual noise’ is a major reason why no definite long-term acceleration of sea level has been identified using twentieth century data alone (Woodworth 1990; Douglas 1992).
Church et al. (2004a) computed time series of global-average sea level change for the last half-century, with the use of tide gauge and altimeter data in combination, via an attempt to reconstruct global sea level fields in spite of the known spatial and temporal inhomogeneities in the tide-gauge data set. Their novel analysis was one of only a small number to have made maximum use of the combined data sets. Their global-average sea level time series demonstrated considerable decadal variability (e.g. that related to ENSO), in addition to a long-term change, similar to that determined from a tide gauge only study by Holgate & Woodworth (2004).
Holgate & Woodworth (2004) concluded that the 1990s experienced one of the fastest recorded rates of global-coastal sea level rise (Church et al. also found higher rates than normal in the 1990s although somewhat smaller rates than Holgate and Woodworth's). This can explain why the rate of truly global sea level change observed in the decade by satellite altimetry (Cazenave & Nerem 2004; Church et al. 2004a) is closer to 3 mm yr−1 than the 1–2 mm yr−1 reported for the twentieth century as a whole obtained from gauge data. The recent acceleration is consistent with findings of AOGCM studies for the past decade compared to those for the past century, in combination with recent observations of hydrographic and glacial changes (Woodworth et al. 2004). Holgate & Woodworth (2004) also observed that during the 1990s sea level appeared to be rising faster around continental coastlines than in the open ocean. White et al. (2005) confirmed this feature for the 1990s but concluded that over a much longer period (the last 50 years) the coastal and open ocean rates of change were essentially the same.
In summary, the 1990s experienced a higher rate of global sea level rise than normal, at a time when global air and sea temperatures achieved record highs and when high rates of glacier melt were reported. Whether the 1990s are indeed the start of a sustained higher rate of sea level rise remains to be seen.
7. Twenty first century sea level changes
Church et al. (2001) contained projections for twenty first century sea level based on AOGCM studies, as observed trends cannot be simply extrapolated into the future if climate is changing. There are many AOGCMs in use, producing different results for future global and regional climate change. Although many of them give a reasonable simulation of present-day climate and variability, they all have deficiencies, and it is not possible to simply select one as the best. Other uncertainties arise from the many scenarios for future concentrations of greenhouse gases and aerosols. In the TAR, no less than 35 scenarios were considered, implying a wide spectrum of future climate, but all producing in most models a future sea level rise primarily through thermal expansion. Global average sea level was predicted to rise by the end of the century by between 9 and 88 cm (relative to a baseline of 1990). The central value of that range corresponds to an average rate of 4.8 mm yr−1, which is considerably larger than the 1–2 mm yr−1 of the twentieth century, and is the result of a predicted continuous acceleration throughout the next 100 years.
Woodworth et al. (2004) further explored the reasons for the wide range of possible future sea levels in the TAR, the reported range being so wide as to be of little benefit to coastal planners. Of course, the greater rates are due to the projected global average temperature rise of 1.4–5.8 °C in the twenty first century. However, the uncertainty in sea level change due to choice of temperature rise (or emission scenario) is relatively small. For the next few decades, in any particular model, the different scenarios give rather similar results for sea level change; even by the end of the century, the range of possible sea level rises is only approximately 50% of the central value. Much larger is the systematic uncertainty associated with modelling, arising from the choice of different AOGCMs and uncertainty in land ice mass balance (see Gregory et al. 2001 for more information on differences between models).
This points to a major need for continued work in these areas aimed at understanding and resolving the differences among models and narrowing (or at least better quantifying) the uncertainties. For example, fig. 11.13 of Church et al. (2001) demonstrated that all AOGCMs predict considerable spatial variability in twenty first century sea level change. (Interestingly, the ratio of the r.m.s of the spatial variability of sea level rise to the mean rise decreases in many models with time, being only 33% for the twenty first century in the HadCM3/GSIO model run referred to above, demonstrating a thermal expansion component common to most regions which increases in time relative to the spatial variability.) In addition, the patterns of spatial variability differed considerably between models. Consequently, it is impossible at present to use AOGCM results in assessing coastal impacts or changes in flood risk, other than by using their findings as limits on the range of possible scenarios for impact studies such as those of the UK Foresight programme (Evans et al. 2004a,b; Hall et al. 2006).
More rapid climate and sea level change than that predicted in AOGCM studies could arise from processes which are inevitably difficult to incorporate in the models. Two major examples include Western Antarctic Ice Sheet disintegration and North Atlantic thermohaline circulation modification. The possibility of the former has been estimated to be very small over timescales of several centuries (Church et al. 2001). The latter, which could in principle occur over several decades, has been the recent recipient of significant UK research funding (Srokosz 2003). Rapid thermohaline circulation modification could significantly increase North Atlantic and UK sea levels above any global-average rise (Levermann et al. 2005).
In addition to changes in MSL, one has also to consider associated climate changes in the ocean and in regional meteorology. The strength of the prevailing westerly winds across the UK and Europe is affected by the North Atlantic meridional air pressure gradient, often measured in terms of a simple NAO index (Wanner et al. 2001). The air pressure and wind changes associated with fluctuations in the NAO are reflected in changes in UK and northern European MSL (Wakelin et al. 2003) and possibly in sea level extremes (Woodworth & Blackman 2004). The signal of the NAO can also be observed to some extent within temperature records, which leads to the intriguing question of whether the NAO also affects regional sea levels via steric (density) changes (Tsimplis et al. 2006). If MSL indeed increases during the twenty first century, and if the average winter NAO index also increases, as suggested by studies of the UK Climate Impacts Programme (Hulme et al. 2002), then the climatology of MSL variability and the return periods of extreme sea levels above MSL could also be modified significantly. North Atlantic wave heights might also be expected to increase (Bacon & Carter 1993) and break nearer the coast as a consequence of the increased depth.
As if prediction for the twenty first century (essentially the IPCC remit) was not difficult enough, it is important to keep in mind that most of the effects of anthropogenic climate change will last much longer than 100 years. Church et al. (2001) emphasized that sea level will continue to rise for many hundreds of years, no matter what the rise in the present century may be (the ‘sea-level commitment’). Hasselmann et al. (2003) have stressed the importance of policy and planning developments required with regard to the potential longer term changes in emissions, temperatures and sea levels.
8. Sea level extremes
The IPCC TAR also emphasized that far more work is required on all aspects of extreme sea levels, in addition to the comparatively much-studied MSLs. The studies of Woodworth & Blackman (2002) for Liverpool (and see other European references therein), Bromirski et al. (2003) for San Francisco and Church et al. (2004b) for Australian sites provide examples of changes in extremes at particular locations over extended periods. However, there are few studies on regional or larger scales. Zhang et al. (2000) studied the trends in extreme sea levels at stations along the US east coast and found that the rise in the level of extremes closely followed that in MSL. Woodworth & Blackman (2004) investigated data from a ‘global’ data set of 141 stations, and concluded that there is evidence for an increase in extreme high water levels worldwide since 1975, as reported frequently in the press. In addition, the variations in extremes in this period were often related to changes in regional climate (e.g. to ENSO-related indices in the Pacific). In most cases, the secular changes and the inter-annual variability in extremes were found to be similar to those in MSL, as Zhang et al. (2000) had also concluded. A similar conclusion can be drawn from a recent study of daily MSLs at Honolulu by Firing & Merrifield (2004), who found long-term increases in the number and height of extreme daily mean values if measured relative to a fixed datum (the highest ever value being due an anti-cyclonic eddy system in 2003) but no evidence for an increase relative to the underlying upward MSL trend. These are important findings with regard to studies of the impacts of coastal sea level changes. If variations in extremes were to be a consequence of a different set of processes to those in MSL (as seems to be the case for the Australian sites), then uncertainties in the occurrence of extremes in future might be expected to be even larger than those in the mean levels, which are themselves considerable.
The study of extremes has to date been much more difficult than research into MSL changes, owing to problems of access to raw sea level data (i.e. hourly values or similar, as high a recording frequency as possible being required for proper identification of the true extreme). Most countries now make their data freely available for research, although the multiplicity of data formats and sampling frequencies means that a major effort in data processing is required before scientific analysis can begin. Some countries continue to restrict access to raw data for reasons of cost recovery or national security. Gradually, these difficulties are being addressed through international programmes such as the Global Sea Level Observing System (GLOSS; see below).
The need for investment in the science, forecasting and mitigation of future river and coastal extremes in the UK has been widely recognized in recent years. For example, the Tyndall Centre funded a study of the vulnerability of the UK coastline (Tsimplis et al. 2005). The UK Government Foresight Programme has undertaken an extensive study into flood and coastal defence (Hall et al. 2006), while NERC has approved a new community-wide programme called Flood Risk from Extreme Events (FREE). There is recognition that extremes can often occur due to a combination of factors within their normal range of variability (e.g. large but not unusual tide and surge combining to produce a notable extreme, Lowe & Gregory 2005) as much as due to long-term change in one or more parameter individually. The estimation of change in risk from modifications in extremes requires simulations of future meteorology within regional AOGCMs coupled to high resolution tide-surge-wave models, taking into account possible changes in land level (e.g. Flather et al. 2001; Lowe et al. 2001). Increasing computer power will enable more realistic simulations of future coastal water levels to be made and changes in risk to be better quantified.
9. Development of sea level networks and related data sources
For good sea level science one needs excellent data sets as well as the scientific insight to exploit them. As discussed above, the historical sea level data set is unable to provide as complete a description of past global sea level trends as we would like; this is one possible reason for the ‘enigma’ being as large as it is. Any analyst of the PSMSL and other international sea level data sets recognizes immediately the possibility of geographical and temporal bias.
Consequently, the community has a responsibility to obtain the resources to conduct a thorough ‘data archaeology’ of as much unanalysed sea level information as possible from the past (archived paper charts etc.). In addition, it has to put in place the sea level observation networks for the future. This is especially urgent if one considers that the networks in many countries are actually worse than they were 10 or 20 years ago, having received inadequate investment during the 1990s.
The main driver for development of the future network has been GLOSS which is a programme of the Joint Technical Commission for Oceanography and Marine Meteorology (JCOMM) of the Intergovernmental Oceanographic Commission (IOC) and World Meteorological Organization (WMO). It aims to establish global and regional networks to provide data for both science and for operational oceanography. The two sets of data users used to be quite distinct: scientists tended to require ‘delayed mode’, fully quality-controlled information, while operational agencies needed real time (or ‘fast’) data as-good-as-you-can-make-it. However, the situation is rapidly changing. Reliable, fast data are becoming the norm thanks to modern tide-gauges and improved data communication methods, allowing their use in large-scale scientific studies (e.g. the Global Ocean Data Assimilation Experiment, Smith & Le Traon 2002) as well as in local flood warning and other operational schemes in coastal waters. Fast data availability also enables more rapid fixing of technical faults and thereby the construction of more complete data sets.
The modern tide gauges which are enabling this revolution in recording are based on different technologies, but most are affordable by national agencies and are easy to install, without compromising accuracy (IOC 2004). For example, the relatively new technique of using radar for monitoring sea levels shows great promise and should be of particular benefit to developing countries (Woodworth & Smith 2003). Many gauges can also sample rapidly in order to monitor wave conditions (Vassie et al. 2004). Web technology is already used extensively by the sea level community and will be developed further through regional GLOSS centres in order to provide data more readily to users.
GLOSS needs investment at the several million dollar level especially in the developing countries of Africa and Asia (Woodworth et al. 2003). A major start in this direction has recently been made by the government of Flanders (Belgium) which has provided funding for 3 years to establish coastal observation systems (including sea level) in Africa. Funding from USA and EU sources is also being discussed. African coasts are experiencing erosion (in West Africa) and coral degradation (in East Africa) while the continent will contain several of the world's largest coastal cities by the middle of the twenty first century. The case for sea level investment seems clear. However, such long-term measurement programmes will only work once adequate local expertise exists and responsibilities for maintaining parts of the network are accepted by each country. Developed countries have a responsibility to provide the necessary training and to contribute to maintenance of the networks in the transition period.
One would hope that, in addition to international programmes such as GLOSS, and the Climate Variability and Predictability Programme (CLIVAR) of the World Climate Research Programme, specific scientific initiatives will result in the maintenance of elements of the global network, especially in polar and other remote locations. For example, the World Ocean Circulation Experiment (WOCE) played a major role in sea level measurement during the 1990s (Woodworth et al. 2002). The International Geophysical Year (1957–1958) stimulated sea level recording in Antarctica, the UK/Ukraine Faraday/Vernadsky base as a result having the longest sea level record in the continent. One looks to the International Polar Year 2007–2008 to contribute similar stimulus to recording in both the Arctic and Antarctic, where tide-gauge data have recently been demonstrated as being essential to an understanding of the ocean circulation (Hughes et al. 2003; Meredith et al. 2004).
As a result of these developments, one expects that the scientists of the twenty first century will have access to data from a much more complete set of tide gauges. Web technology should enable data to be provided in standard (and alternative) formats on a national, regional and global basis. For example, the UK National Tidal and Sea Level Facility (NTSLF) now provides all data from the UK network freely to users via the web. The data are contributed to the European Sea Level Service (ESEAS) which is the regional collaboration of tide-gauge agencies. They are also made available to the global set of users via the GLOSS programme. We look to other countries to provide the same level of ready access to data.
In addition to tide gauges, the sea level science of the future should have access to data from a range of space-based observation systems. These data will be merged with those from the gauges to provide a coherent sea level observing system (cf. Mitchum et al. 2001). The systems will include
Continuous TOPEX-class altimetry, which is assured for the next decade through the Jason-1 and Jason-2/OSTM missions.
Temporal space gravity data. CHAMP and GRACE are now in space. GRACE has potential for measuring the mass component of global sea level change to the sub 0.1 mm yr−1 level, and data from a series of such missions should be able to discriminate between competing models of sea level change (Woodworth & Gregory 2003). In addition, temporal gravity can provide insight on hydrological and glaciological changes which may aid interpretation of past as well as future sea level changes (Chambers et al. 2004; Moore et al. 2006).
Spatial space gravity data. GOCE, to be launched in 2006, will provide a precise model of the geoid for application to a range of ocean circulation and geophysical studies from which sea level research will benefit (Balmino et al. 1999; Woodworth & Gregory 2003).
GPS measurements. In addition to monitoring vertical movements at gauges, GPS will be used for global mass loading determination, complementing space gravity insight into changes in hydrological and other loads (e.g. Blewitt & Lavallée 2002; Blewitt & Clarke 2003).
Land and sea ice data. Cryosat, to be launched in 2005, and follow-on missions will provide continuous monitoring of land and sea ice. Cryosat's primary objective is to test the prediction of thinning arctic ice due to global warming.
To these can be added a range of other ocean instrumentation (moorings, buoys) and remote sensing (ocean temperatures routine today, salinities one day) to enhance understanding of many of the contributors to sea level change.
This paper has given an overview of several of the main scientific, technical and organizational issues concerned with long-term sea level change which are currently being addressed by the sea level science community. The UK community is particularly well qualified to contribute to the scientific and technical topics. In addition, I believe that the UK can provide a global lead in obtaining investments for improving historical data sets as much as possible, as well as in securing the international funding for the measurement systems of the future. The data from those systems will provide more than just ‘monitoring’; they will be employed within a wide range of local and global operational schemes, expanding considerably the number of users of such data and benefiting society in general. The initial investments in such systems are not small and have to be convincingly argued for at many levels. However, with the ongoing appreciation of and reliance on such information, one might hope that GLOSS and its associated regional sea level networks would become well-funded permanent fixtures within global observing systems, primarily the Global Climate, Ocean and Geodetic Observing Systems (GCOS/GOOS/GGOS) but also the Global Terrestrial Observing System (GTOS) which has sea level change interests from the perspective of changing coastlines. In that way, the scientific requirements for long-term sea level data, the practical applications of those data, and the many complementary feedbacks between science and operations, will all be accommodated.
One contribution of 20 to a Theme Issue ‘Sea level science’.
- © 2006 The Royal Society