Introduction

Reconstructions of Holocene relative sea level (RSL) provide important constraints for calibrating geophysical models of Earth’s rheology and glacial-isostatic adjustment (GIA) [13], estimating ice-equivalent meltwater input [47], and fingerprinting sources of ice mass loss [812].

Changes in RSL, which is defined as the height of the ocean surface relative to the land surface or ocean floor [13], are driven by the net effect of eustatic, isostatic (glacio- and hydro-), ocean dynamic, tectonic, and local (e.g., tidal regime change or sediment consolidation) factors that act over a variety of spatial and temporal scales [12, 14]. Since the Last Glacial Maximum (LGM; 26 ka), RSL changes have been driven by the melting of ~50 million km3 of land-based ice as the Earth transitioned from glacial to interglacial climatic conditions [1517]. This transfer of mass from land ice to the global ocean both increased ocean volume and triggered a large, ongoing isostatic response of the solid Earth [4, 13, 18, 19]. The present-day rate of RSL rise due to GIA varies among near-, intermediate, and far-field regions (Fig 1) [20]. In near-field regions (i.e., areas located beneath continental ice sheets at the LGM), the rate of glacio-isostatic uplift during deglaciation exceeded the rate of eustatic (land ice volume and thermal expansion) sea-level rise, resulting in RSL records characterized by continuous RSL fall (Fig 2). Rates of present-day RSL fall as great as −8 mm/a occur in near-field locations (Fig 1). Depression of land beneath LGM ice sheets caused migration of mantle material away from ice load centers, resulting in uplift of a forebulge in intermediate-field regions adjacent to ice sheets [1, 2128]. As land-based ice diminished and mantle material returned to the former load centers, the forebulge collapsed and retreated (glacio-isostatic subsidence). Thus, isostatic and eustatic effects worked in tandem to cause RSL rise. After 7 ka, the relative contribution from glacio-isostatic subsidence increased as meltwater input was reduced [4]. RSL rise slowed from the absence of [28] or minimal [5, 7] eustatic contribution during the late Holocene (last 4 ka); RSL rise in these regions resulted from continuing glacio-isostatic subsidence (Fig 2). RSL is rising at present in intermediate-field locations, the rate of which varies with distance to former ice centers (Fig 1). Transitional zones occur at LGM ice-sheet margins (e.g., Massachusetts on the U.S. Atlantic coast) where initial glacio-isostatic uplift was followed by subsidence from forebulge migration [29]. These locations experience rates of present-day RSL rise in the higher range of intermediate-field sites (Fig 1). In far-field regions at increasing distances from major ice centers, eustatic contributions to RSL change exceeded glacio-isostatic contributions [4, 5, 15, 16, 3036]. The RSL signal of many far-field locations is characterized by a mid-Holocene sea-level maximum, or highstand, at the time meltwater production decreased [37]. The fall in RSL to present is due to hydro-isostatic loading (continental levering) [18, 30] and a global fall in the ocean surface due to both hydro- and glacio-isostatic loading of the Earth’s surface (equatorial ocean siphoning) [38]. Perturbations to Earth’s rotation driven by mass redistribution also cause RSL changes in far-field regions to depart from the eustatic value [39]. These processes occur during the deglacial period but are not manifested in far-field RSL records until the early to mid-Holocene because the eustatic signal is dominant prior to this time [5]. Far-field locations are characterized by present-day rates of RSL change that are near constant or show a slight fall (<0.3 mm/a) in RSL (Fig 1).

Fig. 1
figure 1

Current rates of relative sea-level (RSL) change from glacio-isostatic adjustment (GIA) predicted using ICE-6G_C (VM6) model [20]. The color scheme denoting rates of RSL change was chosen to remain consistent with previous GIA modeling studies from the University of Toronto

Fig. 2
figure 2

Schematic representation of the glacio-isostatic adjustment processes in response to the waxing and waning of continental-scale ice sheets and the associated land-level change in near-, intermediate-, and far-field locations modified from Kemp et al. [46]

RSL Histories and Application of Gaussian Process Model

Here, we present local RSL data that are representative of regional trends to show deglacial patterns of change and identify processes that may have caused these changes. The selected recent and/or benchmark studies illustrate the full range of Holocene RSL variability. Each study has sea-level data from the early to late Holocene and minimal influence from tectonic or local factors (although if the influence from tectonic or local factors is apparent, these factors are discussed in the text). We present studies from a pole-to-pole transect across the Americas (Greenland, North America, the Caribbean, South America and Antarctica) and Eurasia/Oceania (Europe, the Mediterranean, Asia, Africa, and Oceania) that are representative of the influence of eustatic and isostatic effects on regional RSL histories (Figs. 3, 4). We note that while we have chosen sites representative of regional trends, they do no account for the full variability present in RSL histories within a region.

Fig. 3
figure 3

Selected relative sea-level (RSL) reconstructions from Southern Disko Bugt, Greenland [54, 56], Northwest Georgia Strait, Canada [49], Southern Maine, USA [71], New Jersey, USA [43], Louisiana, USA [74], California, USA [49], St. Croix [99], Suriname and Guyana [100 ], Rio Grande do Norte State, Brazil [102], and South Shetland Islands, Antarctica [112, 114]. Altitudes and interpretations of sea-level index points (including errors) have not been amended from the original publication (see section 2 for details). All y-axes are RSL (meters to present). Radiocarbon ages were converted to calibrated dates where necessary. X-axes are in 1000 calibrated years before present (ka). Limiting points that do not constrain the position of RSL are not shown. Blue curves show the Gaussian process model fit to the data, with the heavily/lightly shaded areas representing 1σ/2σ uncertainties

Fig. 4
figure 4

Selected relative sea-level reconstructions from Arisaig, Scotland [121], Baltic Coast of Sweden [122], Rotterdam, Netherlands [123], Southern France [138], Tunisia [138], Nile, Egypt [141], Langebaan Lagoon, South Africa [42], West Guongdong, China [166], Singapore [160], and North Queensland, Australia [181], and New South Wales, Australia [182] shown as in Fig. 3

We show sea-level index points that record the position of RSL over time at each location [40, 41]. Each index point contains information about its (a) geographic location; (b) age from 14C, U/Th, optically stimulated luminescence or amino acid racemization dating and the associated measurement and/or calibration error; and (c) elevation of former sea level and the associated vertical error. We also show marine and freshwater limiting data that define the lower- and upper-bound of RSL, respectively [41]. We do not change the interpretation of index points from the original studies, although in one study [42] where vertical error was not specified for four index points, we assign a conservative (±1 m 2σ) error value. We acknowledge that the error values of many of the index points may be underestimated, in particular because local factors that influence RSL, such as sediment compaction or tidal range change, are not considered [43]. Where necessary, we have converted radiocarbon ages into thousand calibrated years (ka) before present (BP), where the year 0 BP is conventionally taken to be 1950 CE. For clarity, we describe ages to the nearest 0.1 ka without chronological uncertainty. The RSL data can be found in Online Resource 1.

We expand on recent reviews showing Holocene RSL variability [13, 4447] by applying a non-parametric statistical technique (empirical hierarchical modeling with Gaussian process priors) that appropriately accounts for the vertical and chronological uncertainties of the RSL data to provide a probabilistic assessment of past RSL changes and rates of past RSL change [4850]. This method has not been previously adopted to describe spatial variability in RSL change on a global scale and represents a considerable advancement of the approach to determine rates of Holocene RSL change.

We assume that RSL over time at each site, indicated by f(t), can be characterized by a Gaussian process prior [51]. The prior mean is taken as either the maximum or the mean of all of the RSL data at the given site (maximum if near-field and mean if intermediate- or far-field). The prior covariance is given by a Matérn covariance function with smoothness parameter ν = 3/2:

$$ k\left(\varDelta t\right)={\sigma}_{m^2}\left(1+\frac{\sqrt{3}\varDelta t}{\tau}\right) \exp \left(-\frac{\sqrt{3}\varDelta t}{\tau}\right), $$

where σ is the prior amplitude of RSL variability, τ is the characteristic time scale of variability, and Δt is the temporal distance between two time points (see Online Resource 2 for further detail). The proxy records provide noisy observations y i of f(t), with geochronological uncertainty incorporated using the noisy-input Gaussian process method of McHutchon and Rasmussen [52]:

$$ \begin{array}{l}{y}_i=f\left({t}_i+{\delta}_i\right)+{\varepsilon}_i\hfill \\ {}f\left({t}_i+{\delta}_i\right)\approx f\left({t}_i\right)+{f}^{\prime}\left({t}_i\right){\delta}_i\hfill \end{array} $$

where t i is the mean reported age of observation y i , δ i is the error in the age estimate, and ε i is the error in the height estimate. Both age and height errors are assumed to be normally distributed. The variance of the age error is calculated from the error specified in columns 8 and 9 of Online Resource 1. The variance of the vertical error is calculated from the specified error in columns 5 and 6, augmented by a constant σ n 2 to correct for underestimation of data errors. The hyperparameters σ m 2, τ, and σ n 2 are tuned separately for each site to maximize the marginal likelihood of the model conditional upon the observations at that site. The reported rates of sea-level change are average rates calculated over the specified time interval based on a linear transformation of f(t).

In the following sections, we present model predictions and rates using the mean and 1σ uncertainty (Figs. 3, 4). We present average rates of RSL change for 1-ka intervals for each of the sites in Table 1 and Fig. 5.

Table 1 Average rates (m/ka) of RSL change for each site. Values listed are mean ± 1 standard deviation
Fig. 5
figure 5

1-ka average rates of RSL change calculated from the Gaussian process model from each site presented in Figs. 3 and 4 in the following regions: Antarctica/Greenland (a), Europe (b), North America (c, d), Asia (e), Oceania (f), Africa (g), Caribbean/South America (h), Mediterranean (i). Heavily/lightly shaded areas represent 1σ/2σ uncertainties

Greenland

In the near-field location of Greenland, reconstructions from isolation basins provide information on the age of the marine limit of the Greenland Ice Sheet (GrIS) and decimeter-scale changes in RSL since ~15 ka [5359]. Greenland’s RSL history has been dominated by glacio-isostatic uplift from the retreat of the GrIS since the LGM, although spatial variability exists across the region due the variations in local ice load history as well as the influence of the Laurentide deglaciation [6064].

Isolation basins from Qeqertarsuatsiaq [54] and Innaarsuit [56] located in southern Disko Bugt, west-central Greenland, indicate ice free conditions at 10.5 ka, followed by a rapid fall in RSL at a rate of −29 ± 2 m/ka to 40 ± 3 m by 8.5 ka (Figs. 3a, 5a) due to significant, rapid ice retreat, and the associated glacio-isostatic uplift [64]. The rate of RSL fall slowed to −8.6 ± 0.7 m/ka between 8.5 and 4 ka before RSL reached its present-day level at 4 ka. RSL then dropped to a lowstand of −2.6 ± 0.2 m at ~2 ka. RSL records across Disko Bugt show broadly similar RSL histories to the Qeqertarsuatsiaq and Innaarsuit record, although the timing of ice margin retreat (northwest of Disko Bugt and Disko Island became ice free before the main bay) and timing of the mid- to late Holocene lowstand (~3 ka in the east and ~ 2 ka in the west) varies, which is consistent with the progressive expansion of a belt of forebulge subsidence from ice sheet regrowth during the neoglacial [53, 64, 65].

On a broader scale, other Greenland sectors show a different pattern in RSL change. Sites in southern Greenland (e.g., Paamiut [59], Nanortalik [61, 62], Qaqortoq [62], Tasiusaq [66]) became ice free at ~15 ka, after which RSL fell rapidly until it reached its present level at ~10 ka and prior to 4 ka fell to a lower lowstand compared to western Greenland of −6 to −8 m. Ammassalik, southeast Greenland [67], shows the characteristic “J-shaped” form in its RSL history observed in western Greenland, although displays a rapid fall in RSL in the early Holocene more similar to southern Greenland records [64, 67], which suggests fast removal of a large volume of ice during initial ice margin retreat in these areas [62, 64, 67]. Such RSL spatial variability reflects the complex interplay among the GIA response to retreat and advance of the GrIS, glacio-isostatic subsidence from collapse of the Laurentide ice sheet (LIS) proglacial forebulge, and eustatic sea-level rise.

North America

Atlantic Coast

Near- and intermediate-field sites along the North American Atlantic coast exhibit spatially variable RSL histories due to deglaciation of the LIS and movement towards isostatic equilibrium [68, 69]. RSL histories of once-glaciated, near-field regions from Canada to Long Island have been driven by GIA and eustatic sea-level changes. Glacio-isostatic uplift caused RSL fall in Newfoundland and Labrador (from a maximum of 120 m at 16 ka) since the LGM [70]. Nova Scotia experienced a RSL lowstand of −65 m at 13.5 ka and then rose at decreasing rates through the Holocene [70]. The record from Maine [71] shows retreat of the LIS around 15 ka with a marine limit >50 m (Fig. 3c). RSL fell rapidly at a rate of −47 ± 9 m/ka to a lowstand of −60 m at 12.5 ka due to glacio-isostatic uplift [71]. RSL then rose quickly to −20 m as early as 10.5 ka before reaching a period of slowly rising RSL from −17 to −22 m that lasted until 7.5 ka. A switch from glacio-isostatic uplift to subsidence occurred at this time and RSL rose at a relatively constant rate of 2.0 ± 0.3 m/ka from 6 ka to present (Fig. 5c).

Located near the former LIS margin, mid-Atlantic regions from New Jersey to Virginia show monotonic rise in RSL with some of the highest rates of deglacial RSL rise due to the collapse of the forebulge and the associated glacio-isostatic subsidence [68]. Horton et al. [43] accounted for local factors impacting RSL (tidal range change and sediment compaction) to reduce scatter in the Holocene record from New Jersey (Fig. 3d). Using the model of Hill et al. [72], they estimated paleotidal range from 10 ka to present in New Jersey and predicted large Holocene paleotidal changes from 9 to 8 ka. Sediment compaction explained meter-scale variations in RSL where intercalated index points were corrected using the thickness of sediment overburden. New Jersey RSL rose rapidly in the early Holocene at an average rate of 4.4 ± 0.6 m/ka from 9 to 7 ka, slowing in the mid-Holocene to a rate of 2.4 ± 0.2 from 7 to 3 ka. The rate of RSL rise slowed further in the late Holocene to a rate of 1.3 ± 0.1 m/ka from 3 ka to present (Fig. 5d).

South of North Carolina into the U.S. Gulf coast, the effects of GIA diminish [68]. A RSL record derived from compaction-free basal peats demonstrated that Louisiana experienced RSL rise from 8.2 ka to present (Fig. 3f). Following the rapid drainage of proglacial Lake Agassiz, combined with eustatic sea-level rise, high rates of RSL are observed in the early Holocene from 9 to 7 ka (7.2 ± 0.4 m/ka; Fig. 5d) [69, 73, 74]. RSL rise was relatively constant throughout the mid- to late Holocene with a rate of 0.8 ± 0.0 m/ka from 6 ka to present. The data does not support RSL in this region being above present at any time during the Holocene [68, 74].

Pacific Coast

The Pacific coast of North America includes near- and intermediate-field regions [49]). Deglacial RSL histories of the Pacific coast [7585] are supplemented by studies of the earthquake and tsunami history of the Cascadia subduction zone along the coasts of British Columbia, Washington, Oregon, and northern California [8693].

Regions formerly covered by the Cordilleran ice sheet (southern Alaska to Puget Sound, Washington) are strongly influenced by local ice loading, which results in variable RSL histories [85]. In northwest Georgia Strait, RSL fell from a marine limit as much as 154 m at 14 ka to a lowstand of >−25 m at 11 ka after deglaciation due to glacio-isostatic uplift. The rate of RSL fall was as extreme as −69 ± 9 m/ka (from 10 to 9 ka) (Figs. 3b, 5c). A small RSL highstand <5 m was reached at 5.5 ka, followed by a slow rate of RSL fall of −0.4 ± 1.2 m/ka during the late Holocene from 4 ka to present. The magnitude of the mid-Holocene highstand varies among regions formerly covered by the Cordilleran ice sheet with highstands considerably lower (~1.3 m) in central Alaska than in much of southern Alaska and British Columbia (up to 7.2 m), likely due to the influence of tectonics [94].

Sites south of the Cordilleran ice margin (southern Washington, Oregon, and central California, USA) demonstrate a continuous rise of RSL throughout the Holocene due to collapse of the forebulge and the associated glacio-isostatic subsidence. In central California, RSL rose rapidly in the early Holocene (10 ± 1 m/ka between 11 and 8 ka) before slowing to 1.4 ± 0.1 m/ka in the mid- to late Holocene (7 ka to present) (Figs. 3e, 5d). This rate is of similar magnitude to those observed at intermediate-field sites on the U.S. Atlantic coast [68] and the UK and Fennoscandia [95, 96] (Fig. 5), although a portion of this RSL rise may be attributed to Cascadia subduction zone deformation.

Caribbean and South America

The Caribbean and South American Atlantic coast represent intermediate- and far-field regions. Early studies of the Caribbean island of Barbados focused on estimating the deglacial eustatic signal [97], although subsequent investigations indicate the Caribbean and South America are influenced by GIA processes, such as glacio-isostatic subsidence, rotational effects, ocean siphoning, and continental levering [3, 5, 36, 98, 99]. The tectonic setting of the Caribbean and South America may also impact RSL records [98]. For example, Austermann et al. [6] found strong perturbation of RSL model predictions due to a high-viscosity slab associated with subduction of the South American Plate beneath the Caribbean Plate, which suppresses local deformation and decreases subsidence.

Records from Caribbean islands such as St. Croix [99] (Fig. 3g) show a monotonic rise in RSL with relatively high rates of 3.9 ± 1.2 m/ka between 8 and 6 ka (Fig. 5h), which Milne et al. [5] attribute to eustatic contributions. The rate of RSL rise slowed in the mid- to late Holocene to 0.8 ± 0.2 m/ka between 6 ka and present. The absence of a highstand in St. Croix is attributed to glacio-isostatic subsidence from LIS forebulge collapse [5, 99].

Moving south to Suriname and Guyana [100] (Fig. 3h), a small (<1 m) highstand emerges at ~7.5 ka. Faster rates of RSL rise are found in the early Holocene (5.2 ± 0.8 m/ka from 9 to 7 ka) compared to the mid-Holocene (1.0 ± 0.5 m/ka from 7 to 5 ka; 0.1 ± 0.4 m/ka from 5 to 2 ka) before RSL fall to present (−0.5 ± 0.4 m/ka from 2 ka to present) (Fig. 5h). Because Suriname and Guyana are located beyond the southern limit of LIS forebulge collapse, hydro-isostatic loading of the shelf becomes the dominant process and causes uplift of the continent to produce a highstand in the mid- to late Holocene [5].

Rio Grande do Norte State in northeast Brazil shows a similar form in RSL to Suriname and Guyana, and although the timing of the initiation of the highstand is comparable, the magnitude of the highstand at Rio Grande do Norte State is greater at 2.9 ± 0.6 m [101, 102]. From 5 ka to present, RSL fell at a rate of −0.3 ± 0.4 m/ka (Figs. 3i, 5h). Angulo et al. [103] indicated the magnitude of the highstand varied from 2 to 5 m above present along the Brazilian coastline.

Further south, Rostami et al. [98] showed a highstand of 6 to 7 m at ~7 to 8 ka at sites along the Patagonian coast. Rostami et al. [98] suggested that the influence of the Patagonian ice sheet on the RSL data is negligible. The Beagle Channel shows a similar magnitude 6-m highstand, but this highstand occurs at ~6 ka [104, 105]. The presence of a highstand of the observed magnitude in southern Argentina may be related to local hydro-isostatic loading of the broad continental shelf or proximity to the subduction zone on the Chilean trench [5, 98].

Antarctic Peninsula

RSL data from Antarctica can constrain its ice sheet history [106109] in this near-field region where large glacio-isostatic gradients occur [110]. The studies that exist from the Antarctic continent show RSL falling from an early Holocene marine limit of ~20 m in the Antarctic Peninsula, ~40 m in East Antarctica, and ~30 m in Ross Ice Shelf [107, 109].

The most recent RSL studies are from the South Shetland Islands of the Antarctic Peninsula [107, 108, 111]. Watcham et al. [112] used radiocarbon dating of isolation basins and beach ridges from the Fildes Peninsula to reconstruct RSL. To overcome the large uncertainties that exist in the 14C reservoir age [113], Simms et al. [114, 115] used optically stimulated luminescence dating of beach ridges in Maxwell Bay to constrain former sea level. These studies show RSL in the South Shetland Islands remained stable between 15 and 16 m above present from 9.0 to 5.5 ka, although there is scatter and large uncertainties in the mid-Holocene data (Fig. 3j). RSL fell from the mid-Holocene to the present level at −3.2 ± 1.0 m/ka from 3 ka to present (Fig. 5a).

Europe

Trends of RSL from near- and intermediate-field sites in Europe are characterized by glacio-isostatic uplift (e.g., Scotland and Sweden) and subsidence (e.g., the Netherlands and Germany) in response to melting of the British and Fennoscandian ice sheets since the LGM [1, 96, 116120]. Despite its relatively small size, the British ice sheet created large variations in RSL trends from the north to south of the UK [95].

Sea-level index points from isolation basins of Arisaig, northwest Scotland [121], where LGM ice thickness was ~900 m, show RSL fall from 40 to 36 m at ~16 ka to ~3 m above present at 11 ka (Fig. 4a). Fast rates of RSL fall (−4.6 ± 0.8 m/ka) occurred between 14 and 12 ka (Fig. 5b), after which RSL rose to a mid-Holocene highstand at 6.6 ka (6.5 ± 0.5 m). RSL fell from this highstand with rates not exceeding −1.6 ± 1.1 m/ka during the late Holocene in response to a reduction in glacio-isostatic uplift and eustatic input [95].

RSL trends from the southern coast of Sweden reveal evidence for rapid sea-level rise following the retreat of the Scandinavian Ice Sheet [122]. Sea-level index points from isolation basins reveal relatively high rates of RSL rise during the early to mid-Holocene with rates of 4.2 ± 0.6 m/ka from 8.5 to 6.5 ka (Fig. 5b) to a highstand of 8.9 ± 0.5 m (Fig. 4b). From 6.5 ka, the gradual fall in RSL reflects the slow continued glacio-isostatic uplift coupled with minimal eustatic input [122] and is reflected by rates of RSL fall of −1.4 ± 0.4 m/ka from 5 ka to present.

Hijma and Cohen [123] produced a Holocene RSL record from Rotterdam, Netherlands using basal peats (Fig. 4c). RSL rose rapidly in the early Holocene at a rate of 5.3 ± 0.9 m/ka from 8 to 6 ka (Fig. 5b). This rapid rise included a 2.1 ± 0.9-m sea-level jump occurring within 200 years commencing at 8.5 ka. The observed trends were attributed to meltwater release from proglacial Lakes Agassiz and Ojibway through the Hudson Strait into the North Atlantic [123]. During the mid to late Holocene, there was a gradual decline in the rate of RSL rise to 1.2 ± 0.4 m/ka.

Mediterranean

The microtidal regime and presence of numerous remains of coastal archeological structures have made the intermediate- to far-field Mediterranean region a focus for sea-level studies [34, 124, 125]. Regional compilations of RSL data have been produced for both the eastern [126128] and western [129131] Mediterranean coast. Due to its complex tectonic setting [132], RSL histories in the eastern Mediterranean were influenced by the Holocene activity of major faults [126, 133, 134]. In this region, RSL rise slowed significantly in the last 7.5 ka, with consequent initiation of deltaic progradations of the largest Mediterranean rivers [135] such as the Nile delta [136, 137].

RSL records from tectonically stable sectors of the western Mediterranean document a continuous rise from −45.9 ± 0.8 m at 11.5 ka due primarily to glacio- and hydro-isostatic factors [138]. In southern France, RSL rose at 11.5 ± 0.9 m/ka in the early Holocene between 11.5 and 9.5 ka with rates decreasing significantly at ~7.5 ka when RSL reached −7.4 ± 0.3 m. RSL in the mid-Holocene (7.5 to 4 ka) rose at 1.7 ± 0.1 m/ka, slowing further in the late Holocene to a rate of 0.4 ± 0.1 m/ka (Figs. 4d, 5i).

The Southern Tunisia RSL record is characterized by a well-documented mid-Holocene highstand beginning at 7.5 ka (Fig. 4e) [139]. RSL reached a maximum highstand of 1.1 ± 0.2 m at ~5.8 ka, with RSL falling to present slowly at rate of −0.19 ± 0.07 m/ka (Fig. 5i) from 6.5 ka to present. According to Mauz et al. [140], the Tunisian highstand is compatible with the melting history of the Antarctic ice sheet.

RSL in the Nile, Egypt during the early Holocene rises relatively rapidly from ~−23 m at 9 ka to ~−6 m at 6 ka at a rate of 5.6 ± 0.6 m/ka (Figs. 4f, 5i). These rates slowed at ~6 ka with RSL remaining stable in the late Holocene with a rate of rise of 1.0 ± 0.3 m/ka between 6 and 2 ka. Recently, Marriner et al. [141] demonstrated that subsidence was a dominant process, especially in the last ~4.0 ka, and significantly impacted the late Holocene sea-level history of the Nile delta.

Africa

In the far-field location of mainland southeastern Africa (e.g., South Africa, Zanzibar, Mozambique), uncertainties exist regarding the timing and amplitude of a mid-Holocene highstand [42142144] caused primarily by continental levering [39]. Part of these uncertainties results from the large distances between field sites (e.g., South Africa to Mozambique) and the varied RSL reconstruction methods used [145]. At sites on the mainland of southeast Africa, RSL reached ~2.5 to 3.5 m above present by 6 ka [144]. Stable islands off the coast of southern Africa (Mauritius, Reunion and Mayotte) record no highstand (with RSL stabilizing at present level by 3 ka) because these small atoll islands subside with the ocean floor from hydro-isostatic loading [44146, 147]. A recent study of late Holocene RSL in the Seychelles [148], a location demonstrated to closely measure eustatic sea level [39], showed RSL rising <2 m in the last 2 ka.

In South Africa, Compton [44] studied Langebaan Lagoon, a salt-marsh lagoon on the southwest coast to produce a 9000-year Holocene RSL record, which is in general agreement with a previous study of beach rock by Ramsay [149]. In Langebaan Lagoon, RSL rose in the early Holocene at a rate of 4.2 ± 0.8 m/ka between 9 and 6 ka (Figs. 4g, 5g). It reached its present level by ~7 ka, rising to a maximum of 3.1 ± 0.5 m at 6.3 ka. RSL slightly fell at a rate of −1.5 ± 1.9 m/ka between 6 and 5 ka, remaining relatively stable until 2 ka, at which time it fell at a rate of −0.8 ± 1.1 m/ka to its present value.

Asia

Investigations of Holocene RSL in the far-field region of Asia are concentrated in southeast Asia [150153], although studies have expanded in recent years to include South Korea [154], the Philippines [155], China [145 148], the Malay-Thai peninsula [160 164 ], and Japan [165 ]. These records are characterized by a mid-Holocene highstand attributed to hydro-isostatic processes [166, 167 ], which varies in magnitude and timing of initiation. RSL may also be influenced by active subduction zone tectonics from collision of the Eurasian, Indian, Philippine, and Pacific plates [44, 168] or sediment loading from large river deltas [166].

Zong [166] examined China’s Yangtze Delta and Taihu area RSL histories and separated regions by geological setting to examine influences of global and local factors. Within large river deltas, RSL did not reach higher than present due to local subsidence and sediment consolidation. Conversely, a highstand of 1 to 2 m was recorded at Fuijan and East Guangdong where tectonic uplift was observed along the active plate margin of the Eurasian and Philippine plates. In stable coastal sites (e.g., West Guangdong; Fig. 4i), the highstand reaches slightly higher (<1.4 m) than present-day RSL. In West Guangdong, RSL rose from −19.6 ± 1.8 m at 9.6 ka to its present level at 6.5 ka at rates as high as 7.7 ± 2.2 m/ka. RSL remained stable from this time to present with a slight fall between 6.5 ka to present at a rate of −0.2 ± 0.2 m/ka (Figs. 4h, 5e). A recent high-resolution (~ ± 0.1 m) study of corals by Yu et al. [156], however, indicates mid-Holocene RSL has fluctuated on multi-decadal to century scales likely related to climatic changes.

Holocene studies from the Malay-Thai peninsula suggest a RSL highstand of up to 5 m between 6 and 4 ka, which may have exhibited an oscillatory subsequent fall [153, 160, 164, 169173 ]. The highstand varies in magnitude by 3 to 4 m due to spatial variability from hydro-isostasy [167 ]. Bird et al. [160] present sea-level data from the stable location of Singapore, which varies in form from South China (Fig. 4i). High rates of RSL rise (4.4 ± 0.5 m/ka) were observed in the early Holocene between 9 and 6 ka (Fig. 5e) until RSL reached its greatest altitude of 3.3 ± 0.3 m at 5.8 ka. RSL then fell in the mid Holocene from 6 ka to present at a rate of −0.4 ± 0.2 m/ka.

Oceania

Their far-field location and relative tectonic stability have made the coastal margins of Australia an important setting for studies of deglacial sea-level changes [174178]. This region is characterized by a mid-Holocene highstand. Both smooth and oscillating sea-level trends have been recorded since the highstand [179, 180]. This discrepancy in part reflects differences in the sea-level indicators used and their interpretation [178 ].

Woodroffe [181] investigated Holocene sea-level changes from northern Queensland (Fig. 4j). During the Holocene marine transgression, the rate of RSL rise was rapid between 11 and 8 ka (13 ± 5 m/ka; Fig. 5f). RSL reached above present level between 8.0 and 6.2 ka when the average rate of change decreased to 3.0 ± 2.4 m/ka. A mid-Holocene highstand at 5.0 ka reached 2.8 ± 0.4 m above present level and remained relatively stable until at least 2.3 ka. After this time, RSL gradually fell to its present level at a rate of −0.2 ± 2.5 m/ka between 2 ka and present. Woodroffe [181] did not find any evidence for an oscillatory trend in RSL.

In southeastern Australia, Sloss et al. [182] reviewed the sea-level history for the New South Wales coast. They show RSL rose from −15 m between 9.5 and 8 ka at a rate of 8.3 ± 1.6 m/ka, reaching its present level at ~7 ka and a highstand of 1.4 ± 0.1 m at ~5.5 ka. While negative and positive fluctuations are observed during the highstand, they are superimposed over a broader RSL fall of −0.3 ± 0.2 m/ka during this period [182].

In northern New Zealand, Gibb [183] presented RSL data near relatively inactive tectonic regions close to Auckland. Here, RSL rose 25 m from 9 ka to its present level at ~7 ka and has remained within 1 to −0.5 m since [184]. Gehrels et al. [185] suggested that RSL changes from southeastern New Zealand and Tasmania have been close (±0.5 m) to modern level for the last 6000 years.

Conclusions

RSL data reveal spatial and temporal changes caused by the varying dominance of eustatic and (glacio- and hydro-) isostatic factors over the Holocene. Near-field regions (e.g., Antarctica, Greenland, Canada, Sweden, and Scotland) are characterized by a complex, spatially variable RSL fall from a maximum marine limit, which reflect the relative influences of eustatic sea-level rise and glacio-isostatic uplift. In intermediate-field regions (e.g., mid-Atlantic and Pacific coasts of the USA, Netherlands, Southern France, St. Croix) isostatic and eustatic effects collectively produced RSL rise. Far-field locations show a characteristic mid-Holocene highstand in RSL, which differs in timing and magnitude across sites from South America, South Africa, Asia, and Oceania.