Tide-gauge records from the north of Ireland have been digitized to generate annual estimates of both mean-sea-level (MSL) position from Malin Head (1958–1998), and mean tidal level (MTL) from Belfast Harbour (1918–2002). Both sites exhibit substantial annual variation, but show overall long-term shallow rates of falling relative sea-level change (RSLC) that are very similar at −0.2 mm a−1 (±0.37 mm a−1) for Belfast and −0.16 mm a−1 (±0.17 mm a−1) for Malin. Using these rates as constraints, plus other constraints of inferred RSLC rates from the mid-Holocene, an approximation of the likely profile of RSLC rates for the northeast of Ireland since 6 ka ago is presented.
The development of mid- to late-Holocene coastal depositional chrono-stratigraphy around the north of Ireland (Wilson & McKenna 1996; Orford et al. 2003a; Wilson et al. 2004), has been hampered by no observations on the associated relative sea-level position (RSL), the tendency of its relative sea-level change (RSLC) and the relationship between RSLC and shoreline deposition form and volume. Carter (1982a) specified a mid-Holocene RSL (figure 1) that reached its maximum between +2–3 m Irish Ordnance Datum (IOD) at ca 6.8 ka on the north Irish coast and ca 5.5 ka ago on the northeast Ulster coast (ages in un-calibrated radiocarbon years). Field investigations of RSLC around the west and northwest (Shaw & Carter 1994) Irish coasts have since confirmed the general gradient of this mid-Holocene RSL peak position rising to the east (figure 1), across the north of Ireland. Unfortunately Carter's RSL curve was unsupported by relevant sea-level indicator positions over the last 4 ka to specify the RSLC trend, and indicated only that the RSL curve dropped to the present day mean-sea-level position (MSL) at some unknown point in the past.
Carter (1982b) identified on the basis of limited tide-gauge data from Belfast Harbour that the mid twentieth century RSLC rate was virtually zero, suggesting that the isostatic effect was by now considerably diminished. Lambeck (1996) and Lambeck & Purcell (2001) using geophysical models of earth crust deformation with, and then without local ice presence, specify a mid-Holocene high-stand for northeast Ireland peaking between 5–6 m OD, ca 6 ka ago, but is only an approximate guide to the potential RSLC envelope that brings us to the modern day. Lambeck's falling RSL from the modelled mid-Holocene high-stand persists in time beyond that of Carter's (1982a) extrapolation, and could still be a component of late-Holocene RSLC into the twentieth century. This paper analyses the twentieth century RSL signals for the north of Ireland based on tide-gauge records from Malin Head (analysed record: 1958–1998) and Belfast Harbour (analysed record length: 1918–2002), to establish a contemporary constraint on long-term RSLC rate signal (multi-decade to sub-century) and hence attempt to establish the possible structure of the Holocene RSLC rate profile post-6 ka, given other potential constraints on RSLC rate over this period.
2. Long-term tide-gauge data in the north of Ireland
The analysis of twentieth century RSLC for the north of Ireland must be based on data from long-running and well-maintained tide-gauges. There are only two gauges, operated by the Ordnance Survey of Ireland (OSI) at Malin Head (Portmore Harbour, Co. Donegal, Ireland), and the Belfast Harbour Commissioners in Belfast Harbour, that meet these requirements. The Belfast Harbour tide-gauge (BHTG) was installed in the 1880s, but the available analogue record for this study only starts in 1918. Orford et al. (2003b) identified the history and quality of the Belfast Harbour tidal gauge record up to 2002. Belfast has operated four gauge types in several positions within the harbour, causing some difficulty in data consistency due to shifts in gauge position and datum elevations. Likewise, there are many short breaks in the record due to instrument failure. The uncertainty of ground stability in the harbour area due to a sub-surface, mid-Holocene soft estuarine silty-clay has been indicated as a potential problem for establishing RSLC estimates, though Belfast's gauges have been sited within the up-estuary dock area in Belfast, where settlement has been minimal compared to the development in the lower estuary in the last century (Carter 1982b). Todd (1981) undertook a mean tidal level (MTL) analysis for 1918–1980, which was used by Carter (1982b) in his estimate of recent RSLC. Post the Partition of Ireland (1922) the Ordnance Survey Northern Ireland (OSNI) defined its Ordnance Datum (Belfast OD) relative to what was defined at that stage as Belfast MSL. This level is used as the datum for BHTGs.
A consistent stable gauge at Malin Head has been used during the whole record (1958–2002) and has been relatively free from disturbance other than short-term incidents of storm debris fouling the stilling well. Freel (2004) has outlined the history and problems of this gauge. The initial gauge record was analysed by the OSI to determine an MSL position (through MTL analysis) and defined the IOD for Ireland's terrestrial geodetic levelling.
There has not been any attempt to geodetically link the two datums, or to link either datum to UK OD, though both OSNI and OSI regard their respective datum as approximately similar in elevation to each other.
MSL in this study is defined as the mean of hourly determined still-water level elevation observed over a chronological year following Pugh's (1987) definition. An alternative, although statistically less-precise methodology, but which is quicker and less prone to errors of timing in the record, is the use of the semi-diurnal high and low water elevations as a measure of tidal excursion. The annual average of semi-diurnal tidal maximum and minimum levels (MTL) is taken as a surrogate for annual MSL, though usually there is a small difference between the two with MSL being higher than MTL. MSL and MTL differ by variable amounts and sign at different places. These two indices of sea-level position are used synonymously in some prior reporting of RSLC rates in the north of Ireland (e.g. Carter 1982b).
In this study, the original analogue signals have been used with a consistent treatment to transform data into a digital record through digitizing the marigrams. This is the first such treatment for both Malin and Belfast. In the case of Malin Head, the 1958–1998 record was digitized and transformed into hourly elevations (Freel 2004). In the case of Belfast Harbour, a MTL analysis complementing that of Todd (1981) for the period 1960–2002 was obtained. The full explanation of data digitizing transcript, data validation and data adjustments for both Malin Head and Belfast Harbour tidal records is given elsewhere (Orford et al. 2003b: §5).
The hourly elevation values for Malin Head (set to IOD) are stored in annual files. This record is available via the permanent service for mean sea-level (PSMSL) run from the Proudman Oceanographic Laboratory in Liverpool. For this analysis, each annual file's listing of observed elevations was reduced to the mean value to characterize the annual MSL.
Two data sets were produced for analysis of Belfast Harbour RSL: raw and adjusted. The raw data is as identified from the original marigram tidal levels corrected for levelling inconsistencies and gross timing errors. The adjusted data series was derived to fill-in missing data sections of the raw data, through interpolation and comparison checks with ‘buddy gauge-stations’ within a 40 km radius of Belfast, following IOC (1994) procedures: Portpatrick (Scotland), Bangor and Larne (Northern Ireland) were selected. Clearly, this is biased to what other stations might indicate in that it assumes tidal behaviour that echoes other adjacent sites, as well as implying some past stationary memory process. These data (1960–2002) were then linked up with the existing monthly values (1918–1963) as derived by Todd (1981) (obtained from PSMSL and set to a consistent revised local reference datum level) and converted to Belfast OD. The signal for adjusted data series was then reduced to the annual average of semi-diurnal high and low water elevations set to OD Belfast, defined as MTL.
Given an annual MSL or MTL time-series, then a trend line fitted via linear least-squares regression analysis specifies a value of RSLC rate and associated standard error (s.e.). In this analysis, a long-term trend rate is required which reduces the effects of multi-decade and sub-decade atmospheric and oceanographic forcing (Goodwin et al. 2000) from a range of sources that can induce substantial inter-annual variation that raises the s.e. associated with the linear trend estimate: e.g. nodal–tidal elevation (Lisitzin 1974); SST (Sutton & Allen 1997); storms (Schmith et al. 1998); and NAO (Hurrell 1995; Butler et al. 2005). An un-weighted moving average term of 19 years smooths some of the effects of these forcing periodicities. A linear regression line fitted to the resulting smoothed MSL series has been used to produce an estimate of long-term RSLC. This smoothing has the advantage of reducing extremes and the overall variability around the long-term trend. A quadratic polynomial regression has been applied to test for indications of variable rates of long-term RSLC over the record length. However, the records used in this analysis are not sufficiently long to determine genuine accelerations in the rate of sea-level change. Changes in long-term RSLC sense may be identified where statistically the use of a quadratic trend rather than the linear trend, is justified by the %R2 change associated with the quadratic term being significantly different from zero at p<0.05.
4. Results: RSLC determination
RSLC for Malin Head is determined, based on either the annual, or smoothed MSL data (figure 2). The best-fit linear regression lines are shown for both data series. The quadratic equation is only cited if it leads to a significant (p>0.05) increase in %R2 over the linear term. The Malin linear RSL change based on annual MSL is −0.2 mm a−1 but the variability of the data is witnessed by a substantial s.e. term (±0.37 mm a−1). The linear rate of change for the smoothed RSL is −0.03 mm a−1 with a greatly reduced s.e. reflecting the reduction of annual variance (±0.1 mm a−1); however the poor %R2 term (<0.8%) specifies the nonlinearity of both raw and smoothed data (table 1). Although both rates are statistically not different from zero, the overall long-term sense of RSLC for Malin Head in the last half of the twentieth century is negative implying a falling (or regressive) sea-level.
Figure 3 shows the adjusted MTL and smoothed MTL data plus associated linear regression lines for Belfast Harbour. Table 1 also identifies the associated rates of MTL change (adjusted and smoothed data) developed from the Belfast Harbour over the twentieth century. The annual adjusted MTL change rate is −0.16 mm a−1 (±0.17 mm a−1) with the smoothed rate at −0.08 mm a−1 (±0.05 mm a−1). The smoothed data (figure 3) show periodic fluctuations as in the Malin Head data, though the Belfast period appears to be twice that of the Malin trend. The Belfast inter-annual variability increases in the latter half of the twentieth century compared to the early/mid twentieth century. Both series show the early 1970s dip and 1985 peak (note that smoothing shifts the relative fluctuations on in time). The 1970s dip has been noted elsewhere (Woodworth 1987).
It is difficult to draw long-term comparisons between Malin and Belfast using the smoothed data given the difference in data window length. Both time-series show significant (p<0.05) increases in %R2 when characterizing the trend with quadratic rather than linear regression (42 and 18%, respectively). This trend identifies minimal RSL change rates positions in the mid-1960s for Belfast and late-1980s for Malin, with rising RSL rates by the end of the twentieth century.
Given that Lambeck (1996) identifies a mid-Holocene high stand of ca 5–6 m OD between 5 and 6 ka ago, then the potential average linear RSLC rate since the mid-Holocene would be −1 mm a−1 (figure 4). However, it is unlikely that such a fall was linear over time, as its rate will depend on the persistence of the high stand over time, and the shape of the falling shoulder from the high stand towards present-day MSL. From luminescence-dated, prograded gravel-dominated beach ridges at Murlough, northeast of Ireland (figure 1), an estimate of long-term falling RSL rate can be made on the basis of decreasing elevation of gravel ridges (Orford et al. 2003a,b) formed during this falling stage. These ridges were formed on average every 80–120 years and show a mean decreasing elevation of between −1.0 and −2 mm a−1, between 3 and 2 ka ago. Given an assumption that wave state was not radically different in this period (in terms of ridge building) then RSL fall rate at this time could be of a similar order. The tendency for an increasing ridge-building periodicity in the younger and seaward gravel ridges has been interpreted as indicating a reducing rate of sea-level fall associated with these younger ridges (also reducing the available onshore directed sediment volume) such that after 2 ka, the RSLC dropped to less than the 3–2 ka rate, i.e. less negative than −1.0 mm a−1.
Carter (1982a) identified a number of sea-level index points for northeastern Ulster with which he constrained his RSLC curve; however these do not identify the rate of RSLC and are not useable in this analysis.
Using these four constraints: (i) a zero RSLC at the 6 ka high-stand peak at ca 6 m OD; (ii) a RSLC rate of −1.0 to −2 mm a−1 between 3 and 2 ka; (iii) a RSLC rate of less than −1.0 mm rate between 2 and 1 ka; and (iv) the late twentieth century RSLC rate of approximately −0.1 to −0.2 mm a−1, then a first approximation of the smoothed RSLC rate over the last 6 ka can be considered (figure 4). If these estimates of mid- to late-Holocene RSLC rates are appropriate, then figure 4 suggests that the high stand may have persisted for greater than 1 ka, while the probable falling shoulder period was between 5 and 1.5 ka. This shoulder period was associated with the highest rates of falling RSL, and is the most-likely time zone (maximal deposition period: figure 4) for the formation of beach ridges and coastal dunes around northeastern Ireland, that were dependent on both onshore and long-shore sediment sources.
The falling twentieth century RSL trends in the north of Ireland are contrary to rising RSL trends in the majority of other UK tide-gauges (Woodworth et al. 1999). These falling sea-levels indicate the likelihood of some element of remnant isostatic crustal uplift occurring that is still exceeding the eustatic signal. There are difficulties including Ireland with Britain in late Holocene crustal change analyses due to North Channel tectonic separation. This means that the closest estimates of uplift rates from Holocene RSL analyses come for southwest Scotland (Shennan & Horton 2002) where late-Holocene uplift rates are between +0.5 and +1 mm a−1. Direct measurement of surface deformation due to isostatic adjustment using permanent GPS networks have yet to include the north of Ireland. Nocquet et al. (2005) have offered a latitudinal profile of vertical crustal velocities for the European permanent GPS network, which if extrapolated for the north of Ireland identifies modern crustal rise of less than +0.5 mm a−1. If the assumption is made that the current crustal uplift rates are unlikely to have shifted noticeably during the last half century, there has to be some explanation as to why the falling RSL rate might be reducing, as observed by the long-term smoothing of the gauge signals. This could reflect, in accordance with general UK trends, a possible rise in eustatic MSL rate (currently probably in the same range as uplift rate, i.e. <1 mm a−1) due to accelerating global warming that is starting to mask the near consistent twentieth century isostatic signal in the north of Ireland. It is unfortunate that although the uncertainty associated with determination of the RSLC trend is being refined per se, there are still major uncertainties with the two input parameters of isostatic and eustatic rates in the north of Ireland that need reducing, in order to support projections of RSL rise into the next decade.
Analysis of existing long-term sea-level data, drawn from tide-gauge records at Malin Head and Belfast Harbour, establish best estimates of twentieth century RSLC for the north of Ireland. Although there is now some uniformity of RSL change estimate, it is not yet feasible to specify the relative magnitude of the main components that contribute towards this RSLC signal. This trend is another constraint on an initial specification of the RSLC profile for the north of Ireland since the high-stand peak of the mid-Holocene. This specification suggests that the high stand may have persisted for longer than hitherto assumed, while a shorter period of high falling RSLC in the mid- to late-Holocene may account for a potential concentration of concomitant coastal progradation in northeast Ireland between 5 and 1.5 ka.
The significant quadratic signatures for smoothed data could suggest that recent and near-future RSLC trends may be moving from the overall falling linear trend estimate, and that long-term trends may be moving into a positive (rising RSL) mode. While at this stage there is insufficient statistical evidence to justify recent acceleration in the tide-gauge records, it is suggested that the gauge data can be interpreted as showing the long-term isostatic component rate (century-scale) may be starting to be outweighed by the eustatic component, and assuming no measurable change in the isostatic element over the next few decades, then the RSLC signal for the northeast of Ireland could be moving towards a positive trend, rather than maintaining its current negative to near zero trend.
The authors would like to acknowledge the financial support from the Environment and Heritage Service (NI) to undertake this work, and Mr Ian Enlander's (EHS) support for coastal research in general. We thank Dr Philip Woodworth and Mrs Libby Macleod of Proudman Oceanographic Laboratory, Liverpool for enabling access to the tide-gauge records for Belfast Harbour. Likewise we acknowledge access to the Malin Head analogue records supplied by the Ordnance Survey of Ireland. Further financial support is acknowledged from the EU (Environment and Climate Programme) under Research Contracts EU FPIV ENV4-CT970488 and EC5V-CT94-0455, by which Malin Head 1958–1988 data were digitized. R.F. gratefully acknowledges DEL (NI) support through a Natural Environmental Research Studentship to enable working on the Malin Head data. We also appreciate the comments of three anonymous referees.
One contribution of 20 to a Theme Issue ‘Sea-level science’.
- © 2006 The Royal Society