Changes in Mean Relative Sea Level around Canada in the Twentieth and Twenty-First Centuries

Abstract

Trends in regional mean sea levels can be substantially different from the global mean trend. Here, we first use tide-gauge data and satellite altimetry measurements to examine trends in mean relative sea level (MRSL) for the coasts of Canada over approximately the past 50–100 years. We then combine model output and satellite observations to provide sea level projections for the twenty-first century. The MRSL trend based on historical tide-gauge data shows large regional variations, from 3 mm y⁻¹ (higher than the global mean MRSL rise rate of 1.7 mm y⁻¹ for 1900–2009) along the southeast Atlantic coast, close to or below the global mean along the Pacific and Arctic coasts, to –9 mm y⁻¹ in Hudson Bay, as indicated by the vertical land motion. The combination of altimeter-measured sea level change with Global Positioning System (GPS) data approximately accounts for tide-gauge measurements at most stations for the 1993–2011 period. The projected MRSL change between 1980 and 1999 and between 2090 and 2099 under a medium-high climate change emission scenario (A2) ranges from −50 cm in northeastern Canada to 75 cm in southeastern Canada. Along the coast of the Beaufort Sea, the MRSL rise is as high as 70 cm. The MRSL change along the Pacific coast varies from −15 to 50 cm. The ocean steric and dynamical effects contribute to the rise in MRSL along Canadian coasts and are dominant on the southeast coast. Land-ice (glaciers and ice sheets) melt contributes 10–20 cm to the rise in MRSL, except in northeastern Canada. The effect of the vertical land uplift is large and centred near Hudson Bay, significantly reducing the rise in MRSL. The land-ice melt also causes a decrease in MRSL in northeastern Canada. The projected MRSL change under a high emission scenario (Representative Concentration Pathway 8.5) has a spatial pattern similar to that under A2, with a slightly greater rise in MRSL of 7 cm, on average, and some notable differences at specific sites.

Résumé

[Traduit par la redaction] Les tendances des niveaux moyens régionaux de la mer peuvent différer substantiellement de la tendance moyenne mondiale. Nous utilisons d'abord les données de marégraphe et de satellites altimétriques, afin d'examiner les tendances du niveau marin relatif (NMR), pour les côtes du Canada, et ce, sur les 50 à 100 dernières années, approximativement. Nous combinons ensuite des données modélisées et les observations de satellites, afin d'obtenir les niveaux de la mer prévus pour le XXI^e siècle. La tendance du NMR qu'indiquent les données historiques de marégraphe montre de grandes variations régionales, allant de 3 mm an⁻¹ (supérieur au taux moyen d’élévation mondial du NMR de 1,7 mm an⁻¹, entre 1900 et 2009) le long de la côte sud-est de l'Atlantique, à une valeur semblable ou inférieure à la moyenne mondiale le long des côtes du Pacifique et de l'Arctique, et jusqu’à une valeur de −9 mm an⁻¹ dans la baie d'Hudson, comme l'indique le mouvement terrestre vertical. La combinaison des changements de niveau de la mer mesurés par altimètre et des données du système mondial de localisation (GPS) explique approximativement les mesures du marégraphe à la plupart des stations, pour la période de 1993 à 2011. Les changements prévus du niveau marin relatif, entre 1980 et 1999, et entre 2090 et 2099, selon un scénario de changements climatiques avec émissions moyennes à élevées (A2), vont de −50 cm dans le nord-est du Canada à 75 cm dans le sud-est du pays. Le long du littoral de la mer de Beaufort, la hausse du niveau marin relatif atteint jusqu’à 70 cm. Le long du littoral du Pacifique, ce niveau varie de −15 à 50 cm. Les effets stériques et dynamiques de l'océan contribuent à la hausse du NMR le long des côtes canadiennes et dominent sur la côte sud-est. La fonte des glaces terrestres (glaciers et nappes glaciaires) contribue 10 à 20 cm à l’élévation du NMR, sauf dans le nord-est du Canada. L'effet du mouvement terrestre vertical ascendant est considérable et centré près de la baie d'Hudson. Il réduit la hausse du NMR de façon significative. La fonte de la glace terrestre cause aussi une diminution du niveau marin relatif dans le nord-est du Canada. Les changements prévus du NMR selon un scénario d’émissions élevées (8.5 des profils représentatifs d’évolution de concentration ou RCP) présentent une configuration spatiale similaire à celle qu'a produite le scénario A2, sauf pour une hausse légèrement plus grande du NMR, de 7 cm en moyenne, et quelques différences notables à des sites précis.

KEYWORDS:

• sea level rise
• climate change projections
• steric and dynamic ocean adjustments
• vertical land motion
• land-ice melt
• Canada

1 Introduction

A rise in mean sea level is not only a key climate change indicator but also an important societal issue that affects coastal communities. In many parts of the world, mean sea level rise is the major reason for more frequent flooding (UNESCO/ICO, 2010). The rate of rise in global mean sea level is estimated to be 1.7 ± 0.2 mm y⁻¹ for the 1900–2009 period based on tide-gauge data (Church & White, 2011). The Intergovernmental Panel on Climate Change (IPCC) projected global sea level rise for the twenty-first century (Solomon et al., 2007). Under the medium-level A1B scenario from the Special Report on Emission Scenarios (SRES; Nakićenović et al., 2000), the rise in global sea level by the end of this century is projected to be 0.21–0.48 m, and when accelerated ice-sheet mass loss is considered, the rise is projected to be 0.20–0.61 m (Meehl et al., 2007). The IPCC Fifth Assessment Report (AR5) projected rises in global mean sea level of 0.36–0.71 m and 0.52–0.98 m during the twenty-first century under the Representative Concentration Pathways (RCP) RCP4.5 and RCP8.5, respectively (Church et al., 2013). Slangen et al.’s (2014) latest work shows a global mean sea level rise of 0.54 ± 0.19 m and 0.71 ± 0.28 m during the twenty-first century under RCP4.5 and RCP8.5, respectively.

Local sea level changes can differ significantly from the global mean (Cazenave & Nerem, 2004). Local sea level is measured relative to land (hereafter called relative sea level) and thus is influenced by local vertical land motion (VLM). In addition to seasonal (e.g., Han et al., 2002) and high-frequency (e.g., Han et al., 2012) variations, there is significant interannual sea level variability off eastern North America (Han, 2002, 2004, 2007; Sallenger, Doran, & Howd, 2012; Yin, Schlesinger, & Stouffer, 2009), which is likely associated with the large-scale ocean circulation. Sea level and circulation variability appear to be influenced by the North Atlantic Oscillation (NAO; Han, 2002), the dominant mode of atmospheric variability over the North Atlantic. Coastal sea level variability may also be associated with the large-scale wind stress curl offshore (Hong, Sturges, & Clarke, 2000). The rate of mean relative sea level (MRSL) off eastern Canada has ranged from −2 to 4 mm y⁻¹ over the past 50 to 100 years based on tide-gauge data (Han, Ma, Bao, & Slangen, 2014). Off western North America, it has been found that the coastal sea level variability is influenced by the Pacific Decadal Oscillation (PDO) and the El Niño–Southern Oscillation (ENSO; Subbotina, Thomson, & Rabinovich, 2001; Thomson, Bornhold, & Mazzotti, 2008). There are suggestions that the MRSL in this region has not risen since the 1980s, and wind pattern changes may be a major cause (Bromirski, Miller, Flick, & Auad, 2011). Mazzotti, Jones, and Thomson (2008) and Thomson et al. (2008) found very strong spatial variations in relative sea level change along the west coast of Canada. As in the past, local sea level changes in the future may significantly differ from the global mean (Slangen et al., 2012). Local sea level change projections of Slangen et al. (2012) vary from −3.91 to 0.79 m under SRES A1B in the twenty-first century compared with a global mean of 0.47 m. For eastern Canada, projected sea level rise in the twenty-first century varies from nearly 0 to 70 cm (Han et al., 2014). Thomson et al. (2008) also projected highly variable coastal sea level changes for western Canada. Changes in infrastructure elevation required to keep flooding risks unchanged under a changing climate have been estimated based on historical tide-gauge data, projected sea level rise, and associated uncertainties for eastern Canada (Zhai L., B. Greenan, J. Hunter, T.S. James, G. Han, R. Thomson and P. MacAulay, unpublished manuscript, 2015).

Secular MRSL change is not only influenced by steric and dynamic ocean adjustments, as well as glacier and ice-sheet melt in response to the present climate change, but also by glacial isostatic adjustment (GIA) to the last glacial maximum (Han, 2004; Slangen et al., 2012; Thomson et al., 2008), tectonic movement (Mazzotti et al., 2008; Thomson et al., 2008), and groundwater depletion (Slangen et al., 2014). The GIA can induce both VLM and a change in sea surface topography (Peltier, 2004). It is the net GIA effect that really matters to the relative sea level measured by a tide gauge.

As an extension to the work of Han et al. (2014) on MRSL along Atlantic Canada, historical tide-gauge data, Global Positioning System (GPS) measurements, satellite altimetry, GIA model output, and Atmosphere-Ocean General Circulation Model (AOGCM) output are combined to analyze MRSL trends at various coastal locations across Canada during approximately the last 50–100 years and to project MRSL changes during the twenty-first century. We aim to show regional differences in past MRSL trends and in projected twenty-first century MRSL changes, as well as identify major contributors to the regional differences in MRSL trends at various coastal locations across Canada.

2 Data and method

a Tide-gauge Data

Tide-gauge data along the Canadian coast are collected by the Canadian Hydrographic Service. Annual-mean sea level anomalies are used at most of the selected tide-gauge stations (Fig. 1); these are obtained from the Permanent Service for Mean Sea Level (PSMSL; http://www.psmsl.org/). At Nain, monthly-mean PSMSL data are used. At Saint John, daily mean water level data from the Integrated Science Data Management of Fisheries and Oceans Canada are used to produce the annual means if data exist for more than 180 days in any one year. At Point Atkinson, monthly-mean data from the Institute of Ocean Sciences are used to produce the annual means if data exist for six or more months. There are considerable differences in the duration of the tide-gauge records. The data records that are used begin in different years but most end in 2011 (Table 1).

Fig. 1 Map of the study region and the location of tide-gauge stations: Baffin Bay (BB); Charlottetown (CH); Hudson Bay (HB); Harrington Harbour (HH); Point Atkinson (PA); and Saint John (SJ).

Table 1. Locations of tide-gauge stations and associated data periods, along with the MRSL trend ±one standard error from the tide-gauge data over the periods specified if sufficiently long data records are available. If serial correlation is significant at the 95% confidence level then its influence on the standard error of the MRSL trend is included (bold).

	Station	Longitude (°E)	Latitude (°N)	Start	End	MRSL Trend (mm y^−1)
Atlantic	Argentia	−53.983	47.300	1972	2011	2.6±0.7
	Charlottetown	−63.117	46.233	1938	2011	3.0±0.2
	Halifax	−63.583	44.667	1896	2011	3.2±0.2
	Harrington Harbour	−59.483	50.500	1940	1988	−0.6±0.4
	Nain	−61.700	56.550	1963	2011	−2.0±0.4
	North Sydney	−60.250	46.217	1971	2011	3.9±0.5
	Port-aux-Basques	−59.133	47.567	1959	2010	2.6±0.3
	Rimouski	−68.517	48.483	1985	2011	0.3±1.9
	Quebec	−71.167	46.833	1911	2011	−0.3±0.4
	Saint John	−66.067	45.267	1896	2011	2.2±0.2
	Sept-Iles	−66.400	50.200	1973	2011	0.9±0.8
	St. John's	−52.717	47.567	1957	2011	1.9±0.3
	Yarmouth	−66.133	43.833	1967	2011	3.5±0.7
Pacific	Bella Bella	−128.133	52.167	1962	2011	−0.6±0.6
	Point Atkinson	−123.250	49.333	1914	2011	0.9±0.2
	Prince Rupert	−130.333	54.317	1909	2011	1.1±0.2
	Queen Charlotte City	−132.067	53.250	1966	2011	−1.0±0.5
	Tofino	−125.917	49.150	1910	2011	−1.8±0.2
	Victoria	−123.367	48.417	1910	2011	0.6±0.2
Arctic, Baffin Bay, Hudson Bay	Alert	−62.317	82.490	1968	2007	−1.1±1.1
	Churchill	−94.183	58.767	1940	2011	−9.4±0.9
	Resolute	−94.883	74.683	1958	1976	N/A
	Qikiqtarjuaq	−64.117	67.867	2009	2010	N/A
	Tuktoyaktuk	−132.967	69.417	1962	2011	1.9±0.9

Linear trends for MRSL and associated standard errors are derived from tide-gauge data using a least squares fit. The trends are statistically different from zero at the 95% confidence level if their magnitudes are equal to or greater than two standard errors. Time series of sea level residuals are tested for serial correlation using the Durbin-Watson statistic (Durbin & Watson, 1950; Han, Ikeda, & Smith, 1993; Sallenger et al., 2012). If the test results indicate autocorrelation at the 95% confidence level, we correct error estimates for the influence of serial correlation by modifying the number of degrees of freedom. An effective number of degrees of freedom (N_e) is estimated to replace N (N being the number of data points used in the linear regression model). A method for estimating N_e using lag-1 autocorrelation is(1) where r is the lag-1 autocorrelation, which is taken as the coefficient of a first-order autoregressive model fitted to the residuals (Sallenger et al., 2012). The lag-1 autocorrelation computed directly from the residual time series is also tested, showing only small differences in the estimated trend uncertainties.

b Satellite Altimetry Data

Following Han et al. (2014), weekly sea surface height anomalies from Archiving, Validation and Interpretation of Satellites Oceanographic data (Aviso; http://www.aviso.oceanobs.com/en/altimetry/index.html) are used for 1993 to 2011. The Aviso data are an objectively mapped product of TOPEX/Poseidon, Jason-1, Jason-2, European Remote Sensing-1 and -2 (ERS-1 and ERS-2) satellites, Geosat-Follow-on, and Envisat altimeter data on a 1/3^o Mercator projection grid (Ducet, Le Traon, & Reverdin, 2000). All standard corrections were made by Aviso to account for wet troposphere, dry troposphere, and ionosphere delays, inverted-barometer responses, sea state bias, as well as ocean, solid earth, and pole tides. Linear trends and associated standard errors are derived from the altimetric data using a least squares fit.

c VLM Estimation

The GIA causes land subsidence (uplift) that increases (decreases) the MRSL and, at the same time, a decrease (increase) in gravitational attraction that decreases (increases) sea surface topography and thus the MRSL. Three types of VLM rates are used: the GIA model (ICE-5G and VM2) results of Peltier (2004), which include present-day VLM; the net MRSL change associated with the GIA in a 1° longitude by 1° latitude grid; and the VLM derived from GPS data (Craymer, M.R., personal communication, 2013; Sella et al., 2007). The GPS data with an average duration of about 10 years are available at some of the tide-gauge sites.

The VLM rate is inferred by subtracting the sea level trend derived from altimetric data from that derived from tide-gauge data (hereafter referred to as the ATG VLM) during the same period. The standard error associated with the ATG VLM is estimated as the root-sum-square of the standard errors associated with the altimetric and tide-gauge rates. Note that GPS and ATG data include any VLM caused by other factors in addition to the GIA. We also estimate the MRSL change rate by subtracting the GPS VLM from the altimetric sea level trend for 1993 to 2011. The standard error associated with the estimated MRSL rate is calculated as the root-sum-square of the standard errors associated with the altimetric and GPS rates. See Fig. 2 of Han et al. (2014) for the relationship among MRSL, tide-gauge data, GPS data, VLM, and altimetric data.

Fig. 2 Rate of historical MRSL change. Filled circles indicate that the rate (red indicates positive and blue indicates negative) is statistically different from zero at the 95% confidence level; open circles indicate that the rate is not different from zero. Note that the data duration varies from station to station (see Table 1).

d Computation of Sea Level Projections

The IPCC AR4 low (B1), medium (A1B), and medium-high (A2) SRES emission scenarios are considered, as in Slangen et al. (2012). On the global scale, the B1 scenario provides the lower bound of sea level rise, A1B the average condition, and A2 the upper bound. Simulated global sea level rise for the past two decades under A2 closely follows the observed trend. Four contributing components are included: steric and dynamic ocean adjustments, land-ice melt, GIA, and groundwater depletion. The dynamic ocean adjustment is mainly the change in sea level associated with changes in ocean currents through the geostrophic relation. The steric and dynamic ocean component is from an ensemble average of sea surface height change between 1980–1999 and 2090–2099 from eight global AOGCMs for IPCC AR4 scenarios (Slangen et al., 2012). These eight global AOGCMs are Bjerknes Centre for Climate Research Bergen Climate Model, version 2; Canadian Global Climate Model, version 3.1 (T47); Meteorological Research Institute Coupled Global Climate Model, version 2.3.2; Geophysical Fluid Dynamics Laboratory Climate Model, version 2.0; Geophysical Fluid Dynamics Laboratory Climate Model, version 2.1; United Kingdom Met Office Hadley Climate Model, version 3; Model for Interdisciplinary Research on Climate (high resolution); European Centre/Hamburg Atmospheric Model, version 5/Max Planck Institute Ocean Model. Values of the land-ice melt component are also from Slangen et al. (2012). Values of GIA are from Peltier's (2004) model output of net RSL change, including the results for the VLM and for the sea surface topography. The GPS-based VLMs, where available, are used for comparison at tide-gauge stations. Values of groundwater depletion are from Slangen et al. (2014). These authors do not provide information for Hudson Bay or the Canadian Arctic Archipelago (CAA). Because this effect is quite uniform along Canada's coasts, linearly extrapolated values are used in the projection.

We have also considered the IPCC AR5 RCP4.5 (median) and RCP8.5 (high) emission scenarios to provide the average condition and the upper bound. Slangen et al.’s (2014) regional MRSL projections from 1986–2005 to 2081–2100 are used, including an ensemble mean of the output from 21 Coupled Model Intercomparison Project, phase 5 (CMIP5) climate models (see Table 1 in the online supplementary material of Slangen et al., 2014), contribution of land-ice melt, and Peltier's GIA model output, as well as the effect of groundwater depletion. We correct Slangen et al.’s (2014) projections with GPS-based VLMs where available.

3 Past trends

a Historical MRSL Trends from Tide-gauge Data

There are significant spatial changes in the MRSL trend across Canada (Fig. 2). As described by Han et al. (2014), the historical long-term MRSL rates are 2–4 mm y⁻¹ along southeast Atlantic Canada, −2 mm y⁻¹ at a site in Labrador (Nain), and close to zero along the northern shore of the Gulf of St. Lawrence and in the St. Lawrence Estuary. At Churchill on the shore of Hudson Bay, the MRSL rate is −9.4 mm y⁻¹, the lowest of all the locations. Along the Arctic coast, the MRSL rate at Tuktoyaktuk is 1.9 mm y⁻¹, while that at Alert is −1.1 mm y⁻¹. For the Pacific coast, the MRSL patterns are complicated, showing no dominant spatial pattern. The MRSL rate changes from negative at Tofino (−1.8 mm y⁻¹) to positive at Prince Rupert (1.1 mm y⁻¹).

In addition to the secular trend, interannual variations are evident at these six stations (Fig. 3). There are also substantial decadal and multidecadal variations at some stations. These fluctuations could be related to the NAO, the Arctic Oscillation (AO), the Atlantic Multidecadal Oscillation (AMO), the PDO, or ENSO. Because of strong decadal and multidecadal variability, the estimated trend is highly sensitive to the data period and duration. This low frequency variability is responsible for significant serial correlation in annual-mean tide-gauge data, which amplify standard errors associated with the estimated trend, especially for the satellite altimetry period (Figs 3 and 4).

Fig. 3 Annual-MRSLs and linear trends (±one standard error) at selected tide-gauge stations across Canada for selected periods. See Fig. 1 for locations.

Fig. 4 MRSL rate (mm y⁻¹) at tide-gauge stations across Canada. The GIA model value is the average over a 500-year period centred on the present. The altimetry and tide-gauge values are for 1993 to 2011, except at Nain where the tide-gauge rate used is for 1963 to 2011. The tide-gauge rate is shown together with ±one standard error (black vertical line).

b Effects of GIA

From Figs 5a and 5b, the MRSL change associated with the GIA is dominated by VLM along the Canadian coast. The change in sea surface topography itself is relatively small. Along the southeast coast of Atlantic Canada, Peltier's (2004) model indicates that the GIA results in land subsidence and thus MRSL rise, but along the northern Gulf coast and the southern Labrador coast the land uplifts and the MRSL decreases. For the Canadian Pacific coast, Peltier's (2004) model shows small values of land uplift and MRSL fall, except for selected sites such as Victoria, Tofino, and Queen Charlotte. From Fig. 2 and Fig. 5, it can be inferred that the spatial differences in the MRSL trends observed by the tide gauges are, in part, a result of the GIA effect.

Fig. 5 (a) VLM rate due to the GIA based on Peltier's (2004) model. The GIA model value is the average over a 500-year period centred on the present. The rates inferred from tide-gauge and altimetry data from 1993 to 2011 are shown as coloured circles. (b) Rate of MRSL change due to the GIA, based on Peltier's (2004) model.

Peltier's (2004) model result is compared with the VLM derived from GPS measurements (Craymer M.R., personal communication, 2013; Sella et al., 2007). The model VLM rate has a mean bias of −0.9 mm y⁻¹ along the Atlantic coast of Nova Scotia, New Brunswick, and Newfoundland (Fig. 6). There is poorer agreement at Rimouski along the coast of the Gulf of St. Lawrence (Fig. 6), where the model significantly underestimates GPS uplift by 4.0 mm y⁻¹. The GPS measurement and GIA simulation are of opposite signs. There is good agreement at Churchill on the coast of Hudson Bay and Tuktoyaktuk along the Arctic coast. There are large differences for many sites along the Pacific coast, which is not surprising because of tectonic movement in this region. The discrepancy may, in part, be attributed to the fact that the GPS measurements include the VLM caused by other factors such as tectonic movement in addition to the GIA.

Fig. 6 VLM rate (mm y⁻¹) at (near) tide-gauge stations across Canada. The GPS rate is shown together with ±one standard error (black vertical line). The GIA model value is the average over a 500-year period centred on the present. The value of altimetry minus tide-gauge ±one standard error (black vertical line) is for 1993 to 2011. The tide-gauge rate used at Nain is for 1963 to 2011. The GPS rates at North Sydney and Charlottetown are from nearby sites.

Figures 2 and 6 indicate that the spatial differences in the MRSL trends observed by the tide gauges can, to a large degree, be attributed to the GPS-measured VLM. For example, at Churchill, with a land uplift of 12 mm y⁻¹, the MRSL rate is −9.4 mm y⁻¹, which is well below the global average of 1.7 mm y⁻¹; at Charlottetown,with a land subsidence of 1.3 mm y⁻¹, the MRSL rate is 3.2 mm y⁻¹, which is above the global average. The difference of 12.6 mm y⁻¹ in the MRSL trends between Charlottetown and Churchill can be accounted for by a 13.3 mm y⁻¹ trend difference in the GPS VLM.

c MRSL Trends Based on Altimetry and Other Data

The geocentric sea level change rate for coastal Canada is derived using altimetry data for 1993–2011 (Fig. 7a). This shows that the geocentric sea level increases along the Atlantic coast and in Hudson Bay and decreases along the Pacific coast. The linear trends derived from altimetric data may be sensitive to decadal variability and the 18.6-year nodal tidal oscillation, given the data length of 19 years.

Fig. 7 (a) Geocentric sea level change rate from altimetry data from 1993 to 2011. (b) Rate of MRSL change inferred from (a) and the VLM due to the GIA based on Peltier's (2004) model. Estimates from tide-gauge data for the same period are also shown as coloured circles, except at Nain where the data period is from 1963 to 2011.

The rate of MRSL change (Fig. 7b) is inferred by subtracting Peltier's (2004) GIA VLM (Fig. 5a) from the altimetric sea level change rate. The results indicate that from 1993 to 2011 the MRSL rose along the southeast Atlantic coast and decreased off Labrador. The most striking fall, up to 10 mm y⁻¹, occurred in Hudson Bay and the CAA. The MRSL also fell in most parts of the Pacific coast.

The altimetry-based MRSL change rate (the altimetric rate minus the VLM) is also compared with the rate based on tide-gauge measurements over the 1993–2011 period (Fig. 4). The MRSL change rate derived from altimetry data and the GIA model is in approximate agreement with the tide-gauge rate in southeastern Atlantic Canada, a large positive bias at Churchill and Rimouski and a small positive bias on the Pacific coast except at Prince Rupert. In contrast, the MRSL change rate from the combination of altimetry and GPS data is in approximate agreement with tide-gauge estimates in eastern Canada and has a moderately negative bias on the Pacific coast (Fig. 4).

4 Projections of MRSL change

Projections of sea level change in the Canadian oceans under IPCC AR4 A2 (Fig. 8), A1B (not shown), and B1 (not shown) are derived using eight IPCC AR4 global AOGCMs (Slangen et al., 2012), Slangen et al.’s (2012) land-ice model, Peltier's (2004) GIA model and Slangen et al.’s (2014) groundwater depletion effect, for the period from 1980–1999 to 2090–2099. The results under A1B and A2 are almost the same in the study area and are, on average, 8 cm higher than those under B1.

Fig. 8 Projected sea level changes (cm) between 1980–1999 and 2090–2099 under scenario A2, based on eight IPCC AR4 AOGCMs, Slangen et al.’s (2012) land-ice model, Peltier's (2004) GIA model, and Slangen et al.’s (2014) groundwater depletion effect.

The MRSL rise under A2 is 60–75 cm along the south and southeast coasts of Newfoundland and along the Atlantic coast of Nova Scotia (Fig. 8). The highest MRSL rise (up to 80 cm) is projected in the southern Gulf of St. Lawrence. The projected MRSL rise is relatively small in the northeastern Gulf of St. Lawrence and along the northern Newfoundland and Labrador coasts. The MRSL is projected to rise up to 70 cm in the Beaufort Sea and up to 40 cm on the Pacific coast. The projected MRSL in the CAA region is highly variable, ranging from significant decrease (up to about 80 cm) to substantial increase (up to 30 cm).

Figure 9 shows the projected total MRSL changes at selected tide-gauge sites across Canada between 1980–1999 and 2090–2099 under A2. We include contributions by steric and dynamic ocean effects based on the eight IPCC AR4 AOGCMs, by land-ice melt based on Slangen et al.'s (2012) land-ice model, by the GIA based on Peltier's (2004) model, and corrected for the VLM with Craymer's (personal communication, 2013) GPS data, and by groundwater depletion. On the Atlantic coast, the steric and dynamic ocean effects are the dominant factors causing MRSL rise. The land-ice melt contributes 10–15 cm. On the Pacific coast, both the steric and dynamic effects and the land-ice melt contribute to the MRSL rise, while the VLM compensates for it. The projected MRSL rise is about 20–50 cm at 8 of the 16 sites shown in Fig. 9. At some sites, the VLM effect is dominant, leading to an MRSL fall of up to 15 cm (e.g., Tofino and Bella Bella). In the Beaufort Sea, all four components (of which the steric and dynamic ocean effects are dominant) contribute to the MRSL rise, with a total of 70 cm at Tuktoyaktuk. At Churchill on the coast of Hudson Bay, large steric and dynamic ocean effects are compensated for by the GIA uplift, leading to a small net MRSL rise. In the CAA region (Alert, Resolute, and Qikiqtarjuaq), the joint effect of land-ice melt (because of proximity to the Greenland ice-sheet melt region) and the GIA dominates the steric and dynamic ocean effects, leading to an MRSL fall (up to 50 cm). The effect of groundwater depletion is nearly the same (about 8 cm) at all locations.

Fig. 9 Projected total MRSL changes between 1980–1999 and 2090–2099 under A2, together with contributions by steric and dynamic ocean effects (Ocean Dynamics in the figure legend) based on eight IPCC AR4 AOGCMs, by land-ice melt based on Slangen et al.’s (2012) land-ice model, by the GIA based on Peltier's (2004) model and corrected for the VLM with Craymer's (personal communication, 2013) GPS data (GIA+VLM in the figure legend), and by Slangen et al.’s (2014) groundwater depletion effect.

Sea level projections under IPCC AR5 RCP8.5 (Fig. 10) and RCP4.5 (not shown) are derived using data from Slangen et al. (2014) from 1986–2005 to 2080–2100. The results under RCP8.5 in the study region are, on average, 11 and 7 cm greater than those under RCP4.5 and AR4 A2, respectively. Figure 11 shows the projected total MRSL changes at selected tide-gauge sites across Canada under RCP8.5, together with contributions by steric and dynamic ocean effects, by land-ice melt from Slangen et al.'s (2014) land-ice model, by the GIA based on Peltier's (2004) model, and corrected for the VLM with Craymer's (personal communication, 2013) GPS data, and by groundwater depletion. The overall spatial pattern and the dominant contributing factors under RCP8.5 are essentially the same as those under A2. At Tofino and Bella Bella, the MRSL projections for change from negative to positive between RCP8.5 and A2, are mainly due to the differences in the steric and dynamic ocean effects. Note that there is some difference in the A2 and RCP8.5 projection periods.

Fig. 10 Projected sea level changes (cm) between 1986–2005 and 2081–2100 under AR5 RCP8.5, based on Slangen et al. (2014).

Fig. 11 Projected total MRSL changes between 1986–2005 and 2081–2100 under RCP8.5, together with contributions by steric and dynamic ocean effects (Ocean Dynamics in the figure legend) based on 21 IPCC AR5 AOGCMs (Slangen et al., 2014), by land-ice melt based on Slangen et al.’s (2014) land-ice model, by the GIA based on Peltier's (2004) model and corrected for the VLM with Craymer's (personal communication, 2013) GPS data (GIA+VLM in the figure legend), and by Slangen et al.’s (2014) groundwater depletion effect.

5 Conclusions

We have used tide-gauge and satellite altimetry data to investigate historical MRSL change for the coasts of Canada. Results from different sources that account for ocean–atmosphere interaction (from global AOGCMs), GIA effects (from model or GPS observations), and land-ice melt effects (model) are combined to generate MRSL projections.

The coastal MRSL change over the past 50–100 years shows large regional variations, from 2–4 mm y⁻¹ along the southeast Atlantic and Beaufort Sea coasts to well-below zero along Hudson Bay and Labrador. Along the Pacific coast, the MRSL also varies from −1.8 mm y⁻¹ at Tofino to 1.1 mm y⁻¹ at Prince Rupert. Therefore, MRSL trends across Canada deviate substantially from the global mean of 1.7 mm y⁻¹ (Church & White, 2011). The VLM from the GPS data explain the spatial difference in the MRSL changes better than the GIA model. The combination of altimeter-measured sea level change with the GPS VLM approximately accounts for the tide-gauge measurements at most stations.

For the coasts of Canada, the projected MRSL change between 1980–1999 and 2090–2099 under IPCC AR4 A2, adjusted for the GPS-measured vertical land subsidence or uplift, varies from −50 cm in the northeast to 75 cm in the southeast, significantly different from the global mean sea level rise. The VLM is the main factor resulting in the spatial difference in the MRSL change. The MRSL rise is about 50–75 cm along the southeast Atlantic coast of Canada, with smaller or no rise at some locations along the coast of Labrador, the northern coast of the Gulf of St. Lawrence, and along the St. Lawrence Estuary. The steric and dynamic effects are the dominant factors for the MRSL rise. The MRSL change on the Pacific coast varies from −15 to 50 cm. Both the steric and dynamic ocean effects and the land-ice melt contribute to the MRSL rise, while the VLM compensates for it. At some sites, the VLM is dominant leading to an MRSL fall. In the Beaufort Sea, there is a total MRSL rise of 70 cm at Tuktoyaktuk, dominated by the steric and dynamic ocean effects. At Churchill, on the coast of Hudson Bay, the large steric and dynamic ocean effects are compensated for by the GIA land uplift, leading to a small MRSL rise. In the CAA region, the joint effect of land-ice melt and GIA dominates the steric and dynamic ocean effects, leading to an MRSL fall. On average, the change in MRSL from IPCC AR5 RCP8.5 is 7 cm larger than that from AR4 A2 in the study region. The overall spatial pattern of MRSL changes and the dominant contributing factors are similar under A2 and RCP8.5, with notable changes at some sites.

Acknowledgements

This work was funded by the Aquatic Climate Change and Adaptation Services Program (ACCASP) of Fisheries and Oceans Canada. Tide-gauge data were obtained from the PSMSL, UK. Altimeter data are from Aviso in France. We thank Michael Craymer for providing the GPS data. We also thank two anonymous reviewers for their helpful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the authors.