Royal Society Publishing

Towards a vulnerability assessment of the UK and northern European coasts: the role of regional climate variability

M.N Tsimplis , D.K Woolf , T.J Osborn , S Wakelin , J Wolf , R Flather , A.G.P Shaw , P Woodworth , P Challenor , D Blackman , F Pert , Z Yan , S Jevrejeva


Within the framework of a Tyndall Centre research project, sea level and wave changes around the UK and in the North Sea have been analysed. This paper integrates the results of this project. Many aspects of the contribution of the North Atlantic Oscillation (NAO) to sea level and wave height have been resolved. The NAO is a major forcing parameter for sea-level variability. Strong positive response to increasing NAO was observed in the shallow parts of the North Sea, while slightly negative response was found in the southwest part of the UK. The cause of the strong positive response is mainly the increased westerly winds. The NAO increase during the last decades has affected both the mean sea level and the extreme sea levels in the North Sea. The derived spatial distribution of the NAO-related variability of sea level allows the development of scenarios for future sea level and wave height in the region. Because the response of sea level to the NAO is found to be variable in time across all frequency bands, there is some inherent uncertainty in the use of the empirical relationships to develop scenarios of future sea level. Nevertheless, as it remains uncertain whether the multi-decadal NAO variability is related to climate change, the use of the empirical relationships in developing scenarios is justified. The resulting scenarios demonstrate: (i) that the use of regional estimates of sea level increase the projected range of sea-level change by 50% and (ii) that the contribution of the NAO to winter sea-level variability increases the range of uncertainty by a further 10–20 cm. On the assumption that the general circulation models have some skill in simulating the future NAO change, then the NAO contribution to sea-level change around the UK is expected to be very small (<4 cm) by 2080. Wave heights are also sensitive to the NAO changes, especially in the western coasts of the UK. Under the same scenarios for future NAO changes, the projected significant wave-height changes in the northeast Atlantic will exceed 0.4 m. In addition, wave-direction changes of around 20° per unit NAO index have been documented for one location. Such changes raise the possibility of consequential alteration of coastal erosion.


1. Introduction

Global sea-level rise is a major threat to the coastal environment and is expected to accelerate with global warming (Church et al. 2001). Past measurements of sea level are based on tide gauge observations, which cover more than 150 years. These indicate a value for the global sea-level rise in the range of 1–2 mm yr−1 (Church et al. 2001). Nevertheless, most of these tide gauges are located in the northern hemisphere and it has been questioned whether they truly represent the global ocean. Cabanes et al. (2001) raised further doubts by observing that the thermal expansion in the areas in which these long-term tide gauges are located is faster than the global average. This view is disputed (Tsimplis & Rixen 2002; Miller & Douglas 2004), but nevertheless the spatial bias in the oldest sea-level records is real. A second point made by Cabanes et al. (2001)was that after 1993, sea level has been rising fast owing to warming. This is consistent with air-temperature observations from countries experiencing the warmest decade (Hulme et al. 2002).

Sea levels around the UK, as recorded from tide gauges, show remarkable consistency at decadal and inter-decadal time-scales—marking the importance of meteorological forcing at these scales—while the estimated sea-level trends vary widely between different locations (Woodworth et al. 1999). Projections for future sea-level change around the UK have been developed by the use of regional climate change models.

In addition to the regional rise in mean sea-level, changes in wind and wave climate also affect the vulnerability of various coastlines to global change. Storm surges and set-up associated with waves contribute to the sea level in coastal waters and especially at the coast. Waves are also extremely powerful and, in extreme cases, may damage sea walls or beach defences (Draper & Bowness 1983), in addition to commonly ‘overtopping’ defences. Wind and waves can also disrupt shipping both at sea and in harbours. Wave heights in the northeast Atlantic have increased since the 1960s (Bacon & Carter 1993; Woolf et al. 2002, 2003a), but little is known about wave direction changes. It is not clear whether climate change will affect the global distribution of waves, and little is reported within the Intergovernmental Panel on Climate Change Third Assessment Report (IPCC TAR) (Church et al. 2001).

The severity of the impact of sea-level rise at any location will depend on whether the land is locally lifting or subsiding, and on changes in wind and wave factors. The relative importance of the various forcing mechanisms varies from site to site. In order to assess the impact of global climate change on a particular coastal environment, therefore, it is important to identify and estimate the contribution of regional climatic changes. When the regional changes are linked with global change, downscaling from the global models is the issue. Where the regional variability is not reproducible in global models and represents a response to forcing mechanisms absent from or unreliably represented by the models, serious problems with regional prediction arise. Moreover, the impact of climate change on coastal societies depends both on the physical characteristics of the coasts and on whether the local economy relies strongly on sectors vulnerable to sea-level rise and extreme weather/wave conditions. Thus, in addition to physical processes, socio-economic factors need to be considered in deciding the management of vulnerable coastal areas.

The North Atlantic Oscillation (NAO) dominates atmospheric variability over the North Atlantic and Europe (Hurrell 1995), especially in winter. The focus of this study is, therefore, on the NAO, which is essentially a measure of the atmospheric pressure difference between the Icelandic Low and the Azores High. Winters exhibiting high NAO index values have increased temperature, precipitation and westerly winds over northern Europe and drier conditions over southern Europe (Hurrell 1995). In addition to its dominance of inter-annual variability, a multi-decadal positive trend occurred in the NAO index from the 1960s onwards, raising the possibility that climate change might be strongly linked with changes in the NAO (Hurrell 1996), a possibility that remains uncertain (Osborn et al. 1999). Regardless of whether global climatic change will be regionally expressed over the British Isles and Europe by changes in the NAO, or whether the NAO only represents strong natural variability that is superposed on global signals (obscuring recent and future changes) the impacts of the NAO need to be quantified, interpreted and assessed. The effects of the NAO on various meteorological parameters have been well established (Hurrell 1995; Osborn et al. 1999). The impacts of these effects on fishes and bird populations, agricultural production and the Arctic Ocean (see Dickson et al. 2000 and references therein) have been documented. The effects of the NAO on the sea level of the Mediterranean Sea have also been identified: the increased atmospheric pressure associated with the NAO and the changes in the evaporation–precipitation balance caused sea levels to drop after 1960 (Tsimplis & Baker 2000; Tsimplis & Josey 2001). Similar analyses of sea-level variations around the British Isles associated with the NAO are required. Wave heights in the North Atlantic also respond to the NAO (Cotton & Challenor 1999), but this relationship requires further investigation, particularly for the coastal regions of the British Isles.

As outlined by the TAR of the IPCC, regional scenarios of sea-level rise are limited by the lack of climate model simulations with reliable regional-scale sea-level information. Most scenarios and impact studies, therefore, resort to using global-mean sea-level scenarios coupled to estimates of local land movements. For similar reasons, regional scenarios of future wave climate are not readily available. Improved and more detailed modelling studies will eventually overcome these problems. Nevertheless, the models will probably not be able to account for NAO changes and their impacts if the NAO changes are not caused by climate change. Thus, alternative approaches that can be readily applied using existing climate model simulations are needed.

The Tyndall Centre project ‘Towards the vulnerability assessment of the UK coasts’ aimed to resolve the contribution of the NAO and its variations to mean sea level, sea-level extremes, wave height and wave height extremes and also to assess their present and future impacts on coastal communities. This paper will consolidate some of the results of this project concerning sea level, and wave heights around the UK and European coasts. In addition, a methodology for assessing the implications of scenarios for the UK coasts will be developed. To this effect, empirical relationships are combined with scenarios of future regional atmospheric circulation changes (and changes in variability) to improve regional sea-level scenarios and to introduce wave-climate scenarios. Some of the results have already been reported in the scientific literature, while others are under peer review. This paper provides an overview describing the state of knowledge regarding the vulnerability of coastal regions to sea level and wave height changes and their application to planning for the future.

2. Data and methodology

(a) Sea-level data

Sea-level data from a variety of sources were used:

  1. Mean monthly values from tide-gauge data included in the Permanent service for mean sea level (Woodworth & Player 2003) were used (Wakelin et al. 2003; Woolf et al. 2003b; Yan et al. 2004). For the analysis of sea-level extremes, hourly values from a number of tide-gauges in the research quality dataset of the University of Hawaii Sea Level Centre were used (Caldwell 2000; Woodworth & Blackman in press). In addition, a reconstructed time-series for the Liverpool tide gauge dating back to 1768 (Woodworth & Blackman 2002) has been analysed. Generally, tide-gauge time-series provide the most accurate way of measuring relative sea-level change, but they may have gaps.

  2. Satellite altimeter data from the Topex/Poseidon mission were taken from the Southampton Oceanography Centre database, ‘GAPS’, ( Monthly sea surface height anomalies on a 1° by 1° grid for 108 consecutive months from September 1992 to August 2001 were calculated (Woolf et al. 2003b; Yan et al. 2004).

  3. Model sea-level data from a two-dimensional tide+surge model for the period 1955–2000 were used (Flather et al. 1998; Wakelin et al. 2003). The model resolution is 1/3° latitude by 1/2° longitude and is forced by 6 hourly wind and atmospheric pressure data and by tides (Flather 1981). Mean monthly values were formed from the hourly model values.

  4. Sea-level values from nine general circulation models (GCM; Gregory et al. 2001) were used to explore the relationship between future UK and global sea-level change.

The monthly values of sea-level data were separated into a seasonal signal (calculated by averaging over the monthly values for each of the 12 calendar months), and a residual anomaly. The set of residual anomalies represent inter-annual variance about the mean seasonal cycle. The hydrostatic response to local pressure changes (‘inverse barometer correction’) is generally considered to be of −1 cm mbar−1. We analysed the data of both ‘actual’ sea level (excluding tides) and inverse-barometer-corrected sea surface heights. The results presented here are of ‘actual’ sea levels unless otherwise stated.

A number of standard statistical techniques have been applied to these data, including principal component analysis and ordinary (non-canonical, least-squares) regression analysis. The entire monthly dataset has been studied but as the NAO influence is more pronounced during the winter, which is also the period where storms cause most damage, we concentrated on the period December–March. We mainly discuss anomalies from the mean seasonal cycle unless stated otherwise.

(b) Wave data

  1. Satellite altimeter measurements of significant wave height were used. For details on data processing and methodology, see Woolf et al. (2002) and Woolf et al. (2003a).

  2. Wind and waves from model output were also analysed.

(c) NAO data

The NAO index of Jones et al. (1997) was used. Note that analysis is limited to an immediate linear response to the NAO; nonlinear and lagged effects are not considered. Broadly speaking, we have seen a period of predominantly ‘positive’ NAO in the last few decades (1977–2001), preceded by a period of predominantly negative NAO (1935–1977), preceded by another period of predominantly positive NAO (1899–1935). There is also evidence that the ‘centres of action’ of the NAO shifted eastward in the most recent period (Hilmer & Jung 2000), which some authors link to greenhouse gases (Paeth et al. 1999; Ulbrich & Christoph 1999). Therefore, we investigate the stability of the correlations of sea level with NAO by examining different periods where possible.

3. Results

(a) The NAO influence on sea level

The sensitivity of the NAO index with the mean monthly sea-level anomalies from altimetry is shown in figure 1. Most of the northern European shelf is positively correlated, while parts of southwest England and the Atlantic coasts of France, Spain and Portugal are negatively correlated. The sensitivity to NAO varies, with the highest values at the southeastern part of the North Sea and in the Baltic, while around the UK the sensitivity is small (Wakelin et al. 2003; Woolf et al. 2003b).

Figure 1

Correlation and sensitivity of winter sea level to NAO around northern Europe from satellite altimetry (1993–2001).

The data from the two-dimensional tide+surge model for the period 1955–2000 confirm the strong correlation between sea level with the NAO and the spatial patterns revealed by the altimetric data. The percentage of variance explained by the regression of sea level on the NAO from the altimetry data (covering the last decade) and the model data are shown in figure 2a,b. Most of the variance in the east and south of the North Sea is linked with the NAO variability. The correlations between the NAO and simulated sea level when wind and atmospheric pressure variations are both included as forcings to the model (figure 3a) are compared with those obtained when the atmospheric pressure variations are excluded (figure 3b; from Wakelin et al. 2003). The sensitivity over the North Sea is mainly due to increases in westerly winds, while atmospheric pressure changes also have an effect. The strongest sensitivity occurs at the southeastern part of the North Sea. Outside the shallow North Sea, almost all the effect of the NAO are owing to pressure changes with small residual non-hydrostatic effects. Thus the tide+surge model results for 1955–2000 confirm the TOPEX/POSEIDON data for 1993–2001, extend the results over a longer period of time and resolve the contribution of two of the forcing mechanisms.

Figure 2

Percentage variance ‘explained’ by the linear regression of the NAO Index on sea level. (a) Altimetry data (1993–2001); (b) model data (1955–2000).

Figure 3

Sensitivity of the sea-level elevation to the winter-mean NAO index in mm/(unit NAO index). (a) The winter-mean tide+surge model elevations and (b) the winter-mean tide+surge model elevations corrected for hydrostatic pressure.

The agreement between tide gauges and models, as well as tide gauges and altimetric observations in respect to sea-level correlation with the NAO is in general very strong; they reveal essentially the same features and confirm the results from altimetry and the models for 1955 onwards (Wakelin et al. 2003; Woolf et al. 2003b). Nevertheless, the sensitivity of the model data to NAO index changes was found to be somewhat lower than that from the tide gauges (Wakelin et al. 2003). This could be due to an additional influence of the NAO on sea level through temperature (steric) effects as it is known that high NAO corresponds to higher than average winter temperatures (Hurrell 1995) or it could just be a consequence of the reduced standard deviation the model has in relation to the tide gauge data.

The length of the tide-gauge records allows an investigation into the stability of the relationship between the NAO and sea level. This is found to change over the period of observations (table 1). The results of the analysis indicate that the correlation varies with time and that the correlation has become stronger during the last decades. Finally, it was investigated whether there is a frequency band, which correlates best with the NAO index. Wavelet analysis (Yan et al. 2004), cross-spectrum analysis and standard-band pass filtering have been used. The results using standard-band pass filtering (figure 4) for three stations (Helsinki, in the northern Baltic; Cuxhaven 2 in the southeastern North Sea; Newlyn in the Atlantic margin) confirm that the correlation at 0–2 year periods is significant in all three periods for Helsinki and Cuxhaven 2. The role of the westerlies has become more important in recent years because the correlation coefficient increases after 1978. It also extends to a larger part of the spectrum, with the NAO becoming dominant at the longer periods. In contrast, at Newlyn where the atmospheric pressure is the dominant NAO forcing mechanism, the correlation becomes significant for 0–2 years after 1936 and becomes significant in the 2–5 year period in the 1978–2001 period. Relationships on time-scales longer than 10 years are difficult to resolve because the independent sample sizes become very small. Nevertheless, we note the change in sign in the correlation of Cuxhaven between 1899–1935 and the other two periods.

Figure 4

Band pass data analysis for three of the tide-gauge stations: Newlyn, Cuxhaven and Helsinki. The correlation for the whole dataset is shown in (a) while the results for three subsets (1899–1935), (1936–1977) and (1978–2001) are shown in (b), (c) and (d), respectively. Statistical significance was estimated by a bootstrap method in which the NAO index was shuffled randomly 1000 times before band passed and correlated with the band passed tide-gauge data. Where less than 5% of the absolute values of the resulting coefficients were larger than the observed correlation, the coefficient was considered statistically significant. Statistically significant correlations are shown as filled symbols. The periods larger than 20 years are not reported for (b–d) as the records are too sort.

View this table:
Table 1

The location of the stations used and the correlation and sensitivity of their winter values with the winter NAO index (A test of the difference between the correlation coefficients is significant at the 95% level only for Ijmuiden and Vlissingen. Nevertheless, the fact that all correlations strengthen in magnitude implies a real change in the relationship. The records in the periods were not detrended and at least 30 of years of data were contained in each period).

The correlation and sensitivity of the sea level in Cuxhaven and NAO index are shown in table 2 for particular winter NAO index ranges and periods. It appears that a major change in the low NAO regime took place for this station over the last decades. While in the past, the NAO index was strongly anti-correlated with sea level for strongly negative NAO winters, the relationship changed sign during the latest decades. Low NAO index corresponds to weak atmospheric pressure gradients over the north Atlantic. Thus, our observation appears consistent with the results of Plag & Tsimplis (1999) who found changes in the annual and semi-annual component of sea-level dependent on the predominance of the maritime or continental system over Scandinavia and the southeast North Sea. As Cuxhaven is located within this region it appears that weakening of the NAO allows the continental system to influence the region. In the last four decades though, it appears that the NAO extends further eastwards and the significance of the continental meteorology has become of secondary importance for the region even at very low NAO values. Whether this is owing to eastward shifting of the Icelandic low and/or the Azores high is beyond the scope of this study. In addition, examination of 200 years of mean annual sea-level data in the Baltic confirm that the influence of both the NAO and the Arctic Oscillation have increased during the past 30 years albeit the percentage of variance explained by the NAO index in the non-detrended time-series over the two centuries is only between 10 and 35% (Jevrejeva et al. submitted).

View this table:
Table 2

Sensitivity (correlation) number of samples with the NAO for different epochs and different NAO indices for Cuxhaven 2 (The number of samples with NAO less than −1 is very small. As a result, the difference between the correlation coefficient before 1950 and the long-term correlation coefficient is statistically significant only if both pre-1950 periods are joined together. Then the change is statistically significant at the 95% level).

(i) Sea-level trends

The trend values of the winter sea-level time-series at tide gauges are shown in table 3. The trends were derived (i) through linear regression on time and (ii) through linear regression on time and the NAO index for the 1909–1953 and 1954–2000 periods. The resulting values demonstrate significant increases in the winter sea-level trends in all the North Sea stations, which are reduced when the NAO index is also used in the regression. Lagos and Newlyn are the exceptions in agreement with (Tsimplis & Baker 2000) as they are within the area of negative influence of the NAO (Wakelin et al. 2003). The winter data from the two-dimensional tide+surge model include trends of the order of −1 mm yr−1 at the southwest corner of the grid to over 2 mm yr−1 in the eastern North Sea. The features of the spatial distribution of the trends (figure 5) correspond well to those in the variance plot (figure 2b). When the NAO influence is removed from the model data through linear regression, the residual signal includes only small trends in the range −0.4–0.4 mm yr−1. Thus the influence that the decadal NAO variability related has on sea-level trend estimation becomes clear. These results also explain the variability in trends in the tide+surge model identified by Tsimplis et al. (1994) and which were attributed to trends in the meteorological forcing.

Figure 5

(a) Trend of model sea level, (b) trend of model sea level with NAO regression removed.

View this table:
Table 3

Linear trends from tide gauge records from linear regression (The values in columns 5 and 7 are the regression coefficients when the regression included the NAO index).

(ii) Extreme sea levels

A very important question for coastal protection is whether in addition to the mean sea-level rise there are changes in the shape of the sea-level distribution and, especially, whether extreme storm surges have become or will become larger or more frequent. Past changes in hourly sea-level data for tide gauges covering the past 20–30 years have been examined (Woodworth & Blackman in press). Hourly values were used to calculate the percentiles of the distribution of sea level over the year. The 50 percentile is a good approximation of the mean sea level. The 99.9 percentile corresponds to the highest 88 hourly values in a year (Woodworth & Blackman in press). By looking at the changes in the 99 percentile values for each year, statistically significant trends were identified in several tide-gauge records around the globe. Nevertheless, these trends were removed for many stations when the 50 percentile value for each station was subtracted demonstrating that the changes in the extremes were associated with mean sea-level changes. This applies particularly to regions where extremes are determined by ENSO (EI Niño southern oscillation) variability. Woodworth & Blackman (in press) also pointed to the importance of astronomical tidal variations to extremes in a large part of the world where they are ‘tidally dominated’. In some European locations (e.g. eastern North Sea and Baltic), the changes in the 99 percentile correlated well with the NAO even after the 50 percentile was subtracted (Woodworth & Blackman in press; table 4). This implies that the NAO contributes both to changes in the mean and additional changes in the upper part of the distribution. When the analysis was restricted to the 99.9 percentile, however, the significance of the correlation with the NAO was reduced, owing to the considerable noise in the time-series of the highest percentiles. Their conclusion was that, northern Europe aside and considering the globe overall, and with due regard to the limitations of the restricted datasets available for study, no statistically significant difference in the extreme sea levels owing to changes in storminess is identifiable anywhere around the globe over the past 25–30 years. Therefore, the changes in the extremes are mainly owing to the changes in the mean sea level and thus follow the same physics (Woodworth & Blackman in press). This has important implications for predicting the range of uncertainty of changes in extremes for the future, given that predictions of mean sea-level rise are themselves very uncertain (Church et al. 2001), even without additional uncertainty owing to possibly different behaviour of extremes with respect to the means.

View this table:
Table 4

Correlation coefficients between winter NAO and winter (December, January and February) percentiles in various northern European stations for the period 1975 onwards. (R-50% is the correlation of the median (a good approximation of the mean sea level). R-99% is the correlation with the NAO of the highest 1% of the sea-level values in each year. R(99–50%) is the correlation of the time-series formed when the median for each year is subtracted from the 99 percentile. n is the number of years used in the analysis. * denotes results that are not statistically significant with 95% confidence.)

These results were extended back in time by the examination of the annual maximum high water (AM×HW), annual maximum surge-at-high-water (AM×SHW) and surge at annual maximum high water (SAM×HW) from Liverpool for the period 1768–1999 (Woodworth & Blackman 2002). No long-term change over the 232 years in AM×HW and SAM×HW was found, although significant inter-annual variability was observed. The values of AM×SHW were found to be larger in the late eighteenth, late nineteenth and late twentieth centuries than for most of the twentieth century, qualitatively consistent with knowledge of temporal variations in storminess in the region based on meteorological data and anecdotal information (Woodworth & Blackman 2002). However, no correlation with the NAO was established.

Figure 6 shows one aspect of the temporal development of SHW climatology at Liverpool (Woodworth & Blackman 2002). It shows the 3, 10, 20, 50, 80, 90 and 97 percentile levels of SHW values for each year, relative to the median surge for the year, so considers how very high and very low surges are distributed about the mean. One can see that a small negative trend exists for the 97 percentile level, which is absent for the lower percentile levels. The 3, 10 and 20 percentile levels, with respect to the median values, are essentially constant while those for the 80, 90 and 97 percentiles suggest either negative trends throughout the record or, alternatively, a negative trend in the first half of the record and essentially constant values thereafter. In recent decades, especially large values for the high percentiles are observed in 1974 and 1990, evidence for recent storminess which has been confirmed from more detailed analysis of residuals of the complete (hourly and every 15 min, respectively) sea-level record. The evidence for an evolving statistical distribution of surge levels, with 97 percentile levels (i.e. the levels exceeded by approximately the 20 largest SHW each year) and other high-percentile levels larger in the late eighteenth and mid- and late nineteenth century than in the mid-twentieth century, supports the conclusions obtained from the sparser AM×SHW dataset also discussed by Woodworth & Blackman (2002) with regard to the overall levels of storminess in each period.

Figure 6

Percentile levels (3, 10, 20, 80, 90 and 97), bottom to top, for SHW in Liverpool for each year relative to the corresponding median values for SHW for each year (from Woodworth & Blackman 2002).

(b) Waves

(i) Regional wave climate and the NAO

The regional wave climate and the influence of the NAO on the winter wave climate has been studied in detail, primarily using satellite altimeter measurements of significant wave height. These studies have been described in detail elsewhere (Woolf et al. 2002, 2003a), and here we will only summarize some key findings.

Wave climate around northern Europe is strongly seasonal with mean wave heights peaking in January, but with a high risk of both high monthly mean wave heights and extreme wave heights throughout autumn and winter (October–March). There is also high inter-annual variability in monthly mean wave heights, particularly in an ‘extended winter period’ from December to March. These months are those primarily associated with the NAO (Jones et al. 1997). Analysis of altimeter data has demonstrated that a large part of the inter-annual variability in monthly mean wave heights during these months can be described by a linear relationship of wave height anomaly to a NAO index (Jones et al. 1997). The sensitivity of mean monthly wave height to NAO index—estimated by linear regression analysis of an altimeter-based climatology—offshore of northern Europe is shown in figure 7. To the west of Scotland, the relationship is particularly strong, describing approximately 70% of the variance and implying monthly mean wave heights varying from 3 to 7 m for extreme ‘NAO negative’ and ‘NAO positive’ winter months, respectively. The relationship is weaker elsewhere—vanishing on the east coast of Britain—but is a major feature of the region as a whole.

Figure 7

Sensitivity of winter monthly mean significant wave height to NAO around northern Europe.

(ii) Influence of NAO in northwest Scotland

The seas to the west of Scotland are among the roughest in the world. This is also the region where waves are most strongly dependent on the NAO (Woolf et al. 2002, 2003a). The communities of the Western Isles of Scotland and the west coast of the Scottish mainland are highly dependent on the sea and marine resources, both for communications and wealth generation. Therefore, it is interesting to consider whether the sensitivity of waves to the NAO extends to the shores of the islands and mainland and to what extent this impinges on the economic and social activity of the region.

The methods described by Woolf et al. (2002, 2003a) for analysing climatologies of wave heights are primarily useful far offshore where climatological properties of waves typically vary only over large length-scales (aproximately 100 km or more). In this case, climatologies built on grids of one or two degrees of latitude and longitude are appropriate. Within approximately 300 km of the coast, wave heights will vary markedly over much smaller distances and gridded climatologies cannot describe the situation. Instead, satellite altimetry can only be used as a tool if a fairly high spatial resolution is maintained. In figure 8, we show a map of northwest Scotland with ground tracks of Topex/Poseidon superposed. Each symbol along the track represents the central location of a 1 s sample of altimeter returns; these samples are used for calculating averages of each geophysical variable (Snaith 2000). The ground track moves approximately 7 km in 1 s and the footprint of the altimeters is 7–10 km (sea state dependent), therefore along-track analyses of 1 s samples represent retrieval of geophysical variables with a spatial smoothing of approximately 10 km. Ground tracks pass from west to east and reliable data can be retrieved to within 20 km of land; but on leaving land, the altimeter electronics can require a few seconds to ‘lock in’ to the sea surface and there may be a longer hiatus in adequate data. There is a substantial temporal sampling issue with along-track data. In the case of Topex/Poseidon, each ground track is repeated every 9 days and 22 h. Thus, although high frequency analysis cannot be performed on these data, there are sufficient data to perform seasonal analyses. For example, we have estimated the wave height along track 189 in each December–March period from December 1992 to March 2001 and estimated the relationship to NAO index (Jones et al. 1997; by linear regression. The results are summarized in figure 9; these and similar results from other Topex/Poseidon tracks strongly suggest that the sensitivity of wave climate to NAO extends to the exposed coastlines of northwest Scotland and even some relatively sheltered waters. Track 189 passes into the Sea of the Hebrides and the results show that while this sea is less rough than the open ocean, it is still rough by most standards and is sensitive to NAO. This is an important commercial seaway, so this result suggests a significant commercial impact of the NAO.

Figure 8

T/P ground tracks west of Scotland.

Figure 9

Statistics on track 189.

By along-track analysis we have extended the utility of altimeter data to coastal regions. However, there are still major limitations to satellite altimetry as a tool, not least that tracks are far too widely spaced to reveal the entire spatial distribution in a coastal region. Numerical modelling of waves can fill this gap and can also describe the distribution of wave properties and erosion forces over a detailed bathymetry in specific conditions. Satellite altimeter data can be used to provide offshore boundary conditions (Hargreaves et al. 2002) or waves can be hindcast for the entire ocean with a high resolution model of the coastal region nested within the oceanic model. In this study, we have used three nested models for the Sea of the Hebrides: (i) NE Atlantic, 1° resolution, (ii) Malin Shelf (7.5 km) and (iii) Sea of the Hebrides (1.85 km) the first two using the wave action model (WAM) third-generation wave model (Komen et al. 1994) and the inner model using the simulating waves nearshore (SWAM) model (Booij et al. 1999). We show here some output from the latter model for three runs, each in conditions suggested by the broader studies of wave climatology and the influence of the NAO. The first two cases—figure 10, (a) Offshore significant wave height, Hs=5 m and wind=10 m s−1 from west; and (b) Hs=5 m and wind=10 m s−1 from NE—illustrate the different impacts of similar offshore wave heights but different directions. In the case of waves and winds from the west—figure 10a, a fairly typical positive NAO scenario—the more southerly of the Western Isles are exposed to fairly large waves (ca. 4 m). These conditions will be less common in negative NAO winters; in these winters the conditions typified by figure 10b will occur more often. Such conditions are not a threat to the Southern Isles, but do produce fairly rough seas in the northern channels (‘Minch’ and ‘Little Minch’). In this sense, a change in the phase of the NAO simply shifts the region of vulnerability. However, it is also the case that storms from the west (primarily a positive NAO phenomenon) are generally more intense than northerly or northeasterly storms and also have the fetch required to build seas to their maximum height (Carter 1982). Figure 11 shows the case of an intense storm quite common in positive NAO winters—Hs=20 m and wind=20 m s−1 from SW—and illustrates the high exposure of the Sea of the Hebrides and surrounding isles to these events. These case studies help to explain the climatological variability apparent in figure 9, and also extend knowledge to other wave properties and to the areas between satellite tracks.

Figure 10

SWAN model output in Sea of the Hebrides. Five metre significant wave height offshore, 10 m s−1 winds: (a) westerly, (b) northeasterly.

Figure 11

SWAN model output in Sea of the Hebrides. Twenty metre significant wave height offshore, 20 m s−1 winds, southwesterly.

(iii) Effects of the NAO on wave direction

The NAO affects the wind field by increasing the strength of the westerlies and by shifting the storm track further north when in a positive state (Hurrell 1995). Thus it is likely that in addition to the changes in wave height described above, the NAO will be affecting the wave direction at the coast. Nevertheless, this effect has not been documented to date because most historical wave height measurements are not directional. As part of this project, the impacts of climate change at Christchurch Bay in the south of England were studied. Measurements of ground meteorological parameters, in addition to wind and waves offshore from the Meteorological Office model output, were analysed. The influence of the NAO on many parameters was identified. Specifically, the wave direction data around the southwesterly direction show a marked correlation with the NAO index (figure 12). However, the correlation with the wave height data is not that clear owing to the large scattering of the values. The dependency of the wave direction data on the NAO is documented for the first time. This result, which is supported by the general understanding of the shift in wind direction and strength with which the NAO changes, needs to be verified by further in situ studies.

Figure 12

Wave height and wave direction change versus the NAO index outside Christchurch Bay. Data come from the Meteorological Office wave model (

(iv) Effects of the NAO on extreme wave heights

Changes of extreme wave height (90 and 99 percentiles of significant wave height) in the northeast Atlantic have been identified with the NAO previously by analysis of model hindcasts (Wang & Swail 2001). Confirmation of this link from analysis of satellite altimeter or in situ data has so far eluded us. However, the (statistically weak) evidence, such that it is, suggests that the intra-seasonal variability in wave heights is generally correlated with the NAO index in the same sense as the monthly and seasonal means. Thus, for example, to the west of Scotland the high percentiles of significant wave height (e.g. 90 and 99%) are likely to increase even more than the mean wave height for a given increase in NAO index.

4. Scenarios for future changes around the UK

Having quantified the spatial and temporal dependency of sea level and wave heights on the NAO, the next step is to develop improved scenarios of sea level and wave height changes. The approach followed here is to consider scenarios of global-mean sea-level rise, and then to modify these according to regional patterns of sea level simulated by climate models. The variability and/or changes in sea level driven by variability and/or changes in the winter-time NAO will need to be superposed on these scenarios. In addition, differential vertical land movement around the UK (Shennan & Horton 2002) should be taken into account when sea-level trends at particular sites are estimated. These have been discussed in detail in Woodworth et al. (1999)and are not discussed further here. In order to develop a scenario, the following questions must be answered:

  1. What is the relationship of global and UK sea-level change?

  2. To what extent are past NAO changes described by climate models?

  3. To what extent are sea-level changes in the climate models due to NAO forcing?

  4. Is the NAO variability independent of climate change?

Answers to the above questions may be unknown, but assumptions can be made based on the limited evidence available. What follows is the development of a consistent approach for doing this.

(a) Global-mean sea-level scenarios

Here, we follow some UKCIP02 conventions (Hulme et al. 2002), in looking out to three time horizons (‘2020s’, ‘2050s’ and ‘2080s’), expressing all changes relative to the 1961–1990 baseline, and using four scenarios (low, medium-low, medium-high and high), though note that these are not equivalent to the UKCIP02 scenarios. For global-mean temperature change by 2100, the four UKCIP02 scenarios fall at locations that are 14, 24, 60 and 74% of the way between the low and high ends of the IPCC TAR's full range of predictions (based on emissions and scientific uncertainties). Here, similar percentiles within the range of global-mean sea-level changes are chosen for the four sea-level scenarios (table 5).

View this table:
Table 5

The four global-mean sea-level scenarios for the coastal vulnerability study (low, medium-low, medium-high and high) (These global sea-level changes (relative to the 1961–1990 mean) fall within the IPCC uncertainty range for each scenario, but they span a wider range than just the HadCM3 model results. Results from simulations with coupled ocean-atmosphere general circulation models (GCMs) and from a simple climate model (MAGICC; Hulme et al. 1994; Wigley & Raper 2001) are utilised, all of which were presented in IPCC TAR).

The global-mean sea-level scenarios combine the two main components of sea-level rise: thermal expansion of the oceans and melting of land ice in the form of glaciers and ice sheets. The models indicate that the thermal component of sea-level rise is between 52 and 78% of the total rise, while the rest is due to ice melting (Church et al. 2001). It is likely that this proportion might depart from this range at longer time-scales because the response times of the ice sheets are longer than that of the ocean, but for scenarios out to 2100, we consider it reasonable to assume that the thermal expansion component is 70% of the total sea-level rise.

(b) The relationship between global and UK sea level in climate models

The spatial pattern of the sea-level rise due to the melting of land ice has not been simulated to date. It is arguable that at least part of the sea-level change owing to ice melt will be driven by the same changes in circulation that cause differences between global mean and local rise for the thermal expansion component. Nevertheless, in the absence of any quantification we assume that, to a first approximation, the ice-melt-driven rise for the seas around the UK would be the same as for the global mean.

Comparison of the spatial patterns of sea-level rise owing to thermal expansion simulated by nine different GCMs, all driven by almost identical scenarios of increasing CO2 concentration, revealed very little consistency; the average pattern correlation between two GCMs was 0.23, indicating an even greater uncertainty in regional sea-level predictions than in the global mean (Gregory et al. 2001). The ratio of regional sea-level rise around the UK to global sea-level rise during the last decade of each simulation was in the range 1.05–1.55 for seven of the models; that is, the UK sea-level rise was, in most models, higher than the global average. Nevertheless, one model gave a value of 0.4. The ninth model was not used because of concerns over the reliability of its ocean component (Osborn 1995), though its results fall within the range of the other models. Somewhat arbitrarily, this simulated range is sampled by applying factors of 0.5, 1.00, 1.25 and 1.50 to the low, medium-low, medium-high and high scenario, respectively. Because Gregory et al. (2001) state that the patterns for each GCM are relatively stable during each simulation, we consider it reasonable to assume that these factors, that give the relative difference between UK and global-mean sea-level rise, are the same for all three time horizons.

To combine the global-mean sea-level rise with these factors, we scale 70% of the global-mean values (table 5) by the factors given above, and then add in the remaining 30% (which corresponds to the ice melt component) unscaled. Thus, the four UK sea-level rise scenarios for each time horizon are produced (table 6).

View this table:
Table 6

Estimates for sea-level change around the UK when the regional patterns of GCMs are taken into account

(c) Changes in the NAO index with climate change

The established relationships between the NAO and sea level (and waves too) apply to the present-day climate state from which they were derived. In the past, as we have shown, the relationship may have been different. If the characteristics of the winter-time variability in atmospheric circulation (including the NAO) were to change dramatically in a future climate state, then the relationships might not be directly applicable. To investigate this, simulations from seven different GCM-based climate models, under increasing CO2 forcing, have been analysed (some results appeared in Osborn (2002) and Gillett et al. (2003); full results are presented by Osborn (2004)).

This set of GCM simulations provides no firm evidence for a change in the spatial characteristics of the winter NAO under enhanced CO2. Although the analysis revealed some changes in the pattern of variability for particular models (e.g. the eastward shift of the centres of action previously noted by Ulbrich & Christoph (1999), in ECHAM4/OPYC) there is no consistency between the seven GCM changes. Therefore, we assume that the NAO–sea level and NAO–wave height relationships will remain applicable in the future climate state.

There is no firm evidence of any change in the strength of NAO variability owing to enhanced CO2 either (Osborn 2004); one model indicates an increase in the variance of the NAO index, one a decrease and five no significant change. We therefore assume that the variability in sea level (and waves) induced by the observed inter-annual variability in the winter NAO index should be superposed (unchanged) on the scenarios of mean sea-level rise for the UK.

A more fundamental problem relates to the relationship between the NAO index and climate change. At multi-decadal time-scales, the observed NAO index has exhibited some quite strong fluctuations (figure 13: note that the NAO index shown is based on principal components rather than pressure differences between locations). Should these be regarded as ‘variability’ and therefore also superposed on scenarios of mean sea-level rise, or are they part of the climate ‘change’ signal? With no changes in CO2 (or other external forcings), none of the seven GCMs exhibit NAO index trends as large as those observed from the early 1960s to the early 1990s (Osborn 2004). On the other hand, while the models' response to increasing CO2 is a gradual shift towards higher NAO index values during the twenty-first century (figure 13), under the late twentieth century forcing this trend is rather small. Thus, either the models are deficient in their internally generated multi-decadal NAO variability or in their response to increasing CO2, or some other forcing or forcings (for example, solar, volcanic or sulphate) are involved in driving late twentieth century NAO index trends (see Gillett et al. 2003 and Osborn 2004 for further discussion). None of these possibilities can be discounted (nor can any combination of them). In light of inconclusive information, we shall assume that the multi-decadal NAO variations in the observed record (and their influence on sea level and wave heights) could again be superposed on a future mean climate/sea level.

Figure 13

Observed history of the NAO (thick line). The shaded area shows the range of NAO values simulated by seven GCMs under increasing CO2 concentration, with the mean given by the dashed line.

The final aspect of the NAO that is considered here is whether the predicted gradual shift towards higher NAO index values during the twenty-first century (figure 13), which is owing to an intensification of the pressure gradient between the Mediterranean Sea and the Arctic Ocean (Gillett et al. 2003; Osborn 2004) and appears to be part of the climate change signal (rather than variability), is already reflected in the regional sea-level patterns simulated by the GCMs. Comparison of the regional sea-level patterns with the shifts in NAO index (for the five GCMs that were analysed in both Gregory et al. (2001) and Osborn (2004)) yields no clear indication that the models with the greatest increase in NAO index respond with a greater increase in regional sea level around northwest Europe, as might be expected. In the absence of a careful evaluation of each model's ability to simulate the sea-level response to NAO variations, we should assume that the sea-level implications of any shift in the NAO index must be included as an additional component.

The treatment of the NAO as an independent factor is in effect a worst case scenario because if part of its influence is contained already within climate models, then our treatment overestimates its effects.

(d) How will the NAO index change in the future?

What shift in the NAO index should be assumed for each of the four scenarios and time horizons? Given that the relationship between NAO index and sea level around the coast of the UK is mainly positive (i.e. higher sea level accompanies higher NAO index), a similar procedure to the previous sections is followed, whereby the range of possibilities (here given by figure 13) is sampled arbitrarily, taking values from the lower part of the range for the low and medium-low scenarios, and from the upper part for the medium-high and high scenarios. The values in figure 13 are based on a pattern-based index of the NAO and do not correspond directly to the Jones et al. (1997) ‘Gibraltar minus Iceland’ NAO index used elsewhere in this project. In the observations, these two methods of measuring the NAO give well-correlated results (r=0.83; Osborn 2004), and it seems reasonable to convert the values in figure 13 into their equivalent ‘Gibraltar minus Iceland’ index values by a linear scaling factor. The magnitude of the shifts in the winter NAO index indicated in table 7 can be gauged against the trend observed during the period from the mid-1960s to the mid-1990s; for the latter period, the Jones et al. (1997) NAO index increased by about 2 units. Even under the highest scenario obtained here, the shift by the 2080s (from the 1961–1990 mean) is only two-thirds of the recent multi-decadal trend.

View this table:
Table 7

Estimates for NAO change on the basis of the GCMs

Other projections for the NAO behaviour exist in the scientific literature. For example, Latif (2001) suggests a link with the ENSO with a 30-year time lag and on this basis forecasts a reduction of the NAO values. Thus, we consider necessary to present the impact a reduction in the NAO would have. We thus select a value of −2.5 for the 2030s. The mechanism of this interaction is speculative, but nevertheless the possibility of a future reduction in the NAO index is not rejected.

(e) Producing scenarios of future sea level

These scenarios, as all future climate/sea-level scenarios, are based on many assumptions, which have been documented here and justified where possible. There are four steps involved in applying these scenarios:

  1. Obtain the mean sea-level rise for the UK as a whole from table 6.

  2. Add or subtract the adjustment for local land movement (from Shennan & Horton 2002) for the location being studied.

  3. For winter-time only, add or subtract an adjustment for a shift in the winter NAO index by combining the change in NAO index given in table 7 with the local relationship between NAO index and sea level for the location being studied (figures 1 or 3).

The contribution of NAO to mean sea level in 2080s for the scenarios described in table 7 is shown in figure 14. Even for the highest scenarios the contribution of the NAO-related change is at most 10 cm at the eastern North Sea. Nevertheless, one must remember that there is an underlying assumption that the NAO variability can be described by the available climate models.

Figure 14

NAO contribution to sea-level rise (cm) in 2080s for the scenarios in table 6.

In contrast with climate models, the NAO reduction by the 2030s suggested by Latif (2001) produces a significant reduction of sea level over the North Sea (figure 15) of up to 20 cm, thus reducing the impacts of global-warming related sea-level rise.

Figure 15

NAO contribution to winter mean sea-level rise (cm) in 2030s for NAO index −2.5.

(f) Scenarios for waves

There is not much information about future wave height or direction changes available. It appears that wave height variations in the northeast Atlantic have recently been mainly due to the NAO influence. To our knowledge there is currently no substantial evidence for wave height increases as a response to ‘global warming’ per se. Thus, we relate future offshore wave climate changes to the future NAO index alone. In figure 16, the implied wave height changes for the four scenarios of table 7 for the 2080s and for the ‘Latif scenario’ for 2030s are shown. The northeast Atlantic is expected to show again the maximum changes in significant wave heights.

Figure 16

NAO contribution to significant wave height (m) in 2080s for (a) high; (b) medium high; (c) medium-low scenarios shown in table 6. In plot (d) the changes caused by the reduction NAO index reduces to −2.5 by 2030s is shown.

Nearshore in relatively shallow water, the changing sea levels will also affect the development of the wave field. Therefore, for coastal zones, it is necessary to combine the sea level and offshore wave height scenarios within wave models (e.g. Hargreaves et al. 2002).

5. Conclusions

Within the context of the Tyndall Center project, a number of new and important findings have been made. The strong influence of the NAO on sea-level variability and wave heights and directions around north European coasts has been established and documented and its spatial and temporal variability have been resolved. A remaining uncertainty is the contribution of temperature and salinity changes to sea level in the North Sea, but as this is a shallow sea, they are not expected to be large. We also described a method by which our results can be used to improve the scenarios for future sea-level rise and we have also produced scenarios for the future wave climate. We have not produced scenarios for extreme sea levels but, provided the NAO index continues to increase, extreme sea levels are likely to increase locally more than the mean sea level (Woodworth & Blackman in press). An assessment of these changes can be made on the basis of the regression coefficients of the percentile analysis described above and in Woodworth & Blackman (in press) for those stations for which data exist.

Each stage in the creation of these scenarios has uncertainties associated with it. These uncertainties have been described and justification is provided for the assumptions that are made at each stage. It should be appreciated that the uncertainties associated with future sea level are large—even more so at the regional/local scale—and are perhaps larger than for many other climate and climate-related variables. See UKCIP02 (Hulme et al. 2002) and IPCC TAR Working Group 1 (WG1) ch 11 (Church et al. 2001) for further details and many uncertainties and caveats relating to the generation of sea-level scenarios for the future, especially at regional scales.

The next step is the use of regional wave models including sea-level changes to assess the geomorphological and socio-economic impacts. A part of this project was devoted to such an effort and the results will be reported elsewhere.


The above work was funded by the Tyndall Centre for Climate Change Research. A. G. P. Shaw was funded as part of the EC EVRI-CT-2002-40025, ESEAS-RI project.


  • One contribution of 14 to a Theme ‘The Big Flood: North Sea storm surge’.


View Abstract