Jamieson [1] [2] understood that the Earth is not rigid and that the load of an ice cap had to deform the bedrock beneath, causing down warping and uplift in response to the glacial advance and recession. He saw the Scottish and Fennoscandian uplift as evidence of this effect; i.e. “glacial isostasy”. The full evidence and description of glacial isostasy were given by De Geer [3] , however, and with this paper a new epoch begun in the study of Fennoscandian uplift [4] .

When it was understood that the Earth had passed cold period known as Ice Ages with expansions of continental ice caps in the Alps [5] , in North America [6] and in Fennoscandia [7] , the phenomena of glacial eustasy [8] and glacial isostasy [3] followed logically. Glacial eustasy implies that water is transferred to the ice caps and global sea level by consequence falls [9] , later to rise when the ice caps melt in postglacial time. Glacial isostasy implies that the load of the ice caps deforms the bedrock leading the crustal subsidence, later to transform into crustal uplift in response to the vanishing load during postglacial melting [3] [4] .

The lowering of global sea level also meant that the Earth’s rate of rotation increased significantly [10] and that the geoid surface was deformed affecting the crustal dynamics and local sea level [11] .

The climatic alternations between Ice Ages and Interglacials [12] are linked to a spectrum of related changes of global and local terrestrial processes as illustrated in Figure 1 [13] . In this paper, however, we will confine the discussion to the process of glacial isostasy, and the question whether this is a primarily regional phenomenon or if it has global dimensions.

Bloom [14] was the first to suggest that the process of glacial isostasy might also affect the rest of the globe via the distribution of the loading effects into the asthenosphere. Walcott [15] called attention to the fact that the postglacial glacial isostatic compensation after the vanishing of the huge ice caps in Fennoscandia and North America (and elsewhere, too) would affect the crustal compensation regionally, if it took place via a low- viscosity channel flow, whilst it would generate global compensational crustal motions if the viscosity had a linear profile (i.e. no channel flow). The discrimination between these two concepts will be the focus of the pre- sent paper.

Figure 2 illustrates the two alternative concepts of the spatial extension of glacial isostasy (from [16] ), here termed Model A and Model B, respectively.

Model A (Figure 2(a)) is based on a linear viscosity model [17] -[20] , and would imply that the glacial isostatic compensation had direct global dimensions, and hence would lack a peripheral bulge (at leas a major one) surrounding the glacially depressed region. This theory was further developed into a global standard correction model [21] [22] , and has even evolved into a general correction factor for present day sea level records [23] .

Model B (Figure 2(b)) implies that the glacial loading and de-loading is compensated by lateral flow in a low viscosity channel and that the glacially down warped area was surrounded by a compensational forebulge. This is the classical theory of glacial isostasy in Fennoscandian [3] [24] - [29] . All data available from Fennoscandia are in favor of a low-viscosity channel flow (Figure 2(b)). The mass in the cone of absolute postglacial uplift is

the same as the mass in the surrounding peripheral subsidence trough, indicating a horizontal mass flow from the collapsing forebulge (subsidence trough) to the rising uplift cone as further discussed below (Section 3).

The determining factor for the discrimination between the two models (Figure 2) is the mode of crustal deformation in the near field [3] [27] and the mode of sea level changes in the far field [26] [30] .

During the Quaternary Ice Ages large ice caps covered the Fennoscandian region [3] [4] [7] [16] . De Geer [3] was the first to show that there was clear relationship between the extension of the Fennoscandian ice cap of the Ice Age and the geometry of crustal deformation (i.e. postglacial uplift) of the Fennoscandian Shield with a maximum central uplift in the order of 200 m (Figure 3). Subsequent studies have, of course, sharpened the picture. Today [4] [27] , we know that the absolute glacial isostatic uplift of the Fennoscandian Shield had the form of a cone with a maximum central uplift of 800 m (Figure 4), and being surrounded by a subsidence trough (i.e. the collapsing forebulge of the Ice Age glacial isostatic down warping of Fennoscandia). We also know that the rate of uplift right after the time of deglaciation amounted to as much as 30 - 40 cm/yr [13] [27] .

The global isostatic loading models [19] - [23] predict high Mid-Holocene sea levels in the Pacific and Indian Ocean. This does not concur with observational facts, either in the Indian Ocean or in the Pacific [16] [30] . The new sea level curve of the Maldives [38] , exhibits a long term base-curve not above present sea level and a number of rapid oscillations caused by dynamic forces. In the Pacific, observed short and rapid fluctuations in sea level [39] [40] do not concur with the loading model but represent high-frequency dynamic sea surface changes. Grossman et al. [41] reconstructed the spatial distribution of Mid to Late Holocene sea level changes in the Pacific. Their reconstruction does not concur with the prediction from the global loading models, but with geoid deformation and/or changes in sea surface topography [30] .

Therefore, one should be very careful in the application of model reconstruction and prediction based on proposed global loading mechanisms. This is, of course, especially true in an area like the Mediterranean dominated by tectonics and orogenic processes [16] .

Present changes in sea level are primarily measured by coastal morphology, tide gauges, and satellite altimetry, but also by considering changes in the Earth’s rate of rotation (LOD) and global gravity (GRACE).

In a few places, we know the long-term crustal component [34] [36] [42] [43] , and are able to separate the absolute sea level component from the relative sea level recorded by tide gauges or other means. This is, for example the case with:

Stockholm: uplift 4.9, minus tide gauge 3.8 = eustasy 1.1 mm/yr [34] [42]

Korsör (the stable axis of tilting for 8000 years): uplift ±0.0, minus tide gauge 0.9 = eustasy 0.9 mm/yr [34] [36]

Cuxhaven: subsidence 1.4, minus tide gauge 2.5 = eustasy 1.1 mm/yr [36] [42]

Amsterdam: subsidence 0.4, minus tide gauge 1.6 = eustasy 1.2 mm/yr [34] [42] [67]

Brest: crustal component ~0.0, tide gauge 1.0 = eustasy ~1.0 mm/yr [34] [42]

Venice: subsidence 2.3, minus tide gauge 2.3 = eustasy ±0.0 mm/yr [42] [51]

Connecticut: subsidence 1.0, sea level rise 2.2 = eustasy 1.2 mm/yr [30] [43]

Figure 10 gives a histogram of the tide gage records used by University of Colorado [44] in their global sea level assessment [42] . The mean of 182 sites (excluding a few out-layers) scattered all over the globe is 1.6 mm/ yr [30] [42] . Because of long-term subsidence of many river mouth sites and site-specific compaction problems [42] , this value may, in fact, represent a slightly too high value. The key sites here discussed provide values of about 0.0 mm/yr, and the Kattegatt and North Sea records give firm values around 1.0 ± 0.1 mm/yr [36] [42] .

This data set is in deep conflict with the high rates proposed by the IPCC [45] [46] and satellite altimetry [47]

[48] . The differences in rates can only be understood in terms of errors and mistakes [30] [42] . The true mean global eustatic component is likely to be found in the zone ranging from +2.0 mm/yr to ±0.0 mm/yr, and most probably in the lower half of this zone; i.e. within 1.0 - 0.0 mm/yr [42] . The error was found to be in the satellite altimetry values for reasons of incorrect “corrections” [30] .

Yes, something must be wrong (Figure 10), and we know what it is; viz. the corrections applied to the satellite altimetry records. After all my critics of these “corrections” [30] [36] [42] [49] -[55] , neither specified nor backed up by established facts, the University of Colorado now for the first time admits that the record is “GIA corrected” [44] [48] ; applied after 2011 and amounting only to 0.3 mm/yr, however [70] .

In order to establish some harmony in the sea level records of Figure 10, we need to remove the global glacial isostatic (GIA) correction applied to the satellite records. In respect to the evidence of a regional (not global) glacial isostatic deformation of Fennoscandia and surrounding areas of northwest Europe (Figure 7), we must remove the whole idea of a global adjustment (Figure 2(a)).

Next question is which value we should subtract. This is not so easy, as this “correction” was not specified in the handling of satellite data. Peltier and Tushinghan [56] used a global GIA long-term sea level residual correction of 2.4 mm/yr, and Lambeck [22] a value of 1.8 mm/yr. Cazenave et al. [57] used a GIA correction of 2.0 mm/yr for their correction of the GRACE data of 2003-2008 with a pre-corrected trend of −0.12 ± 0.06 mm/yr. Hence, the different corrections range from 1.8 to 2.4 mm/yr.

A simple way of removing the “GIA” factor from the satellite altimetry records would be to subtract 1.8 to 2.4 mm/yr from the 3.3 mm/yr by University of Colorado [48] and from 2.9 mm/yr by NOAA [47] . In the first case, we obtain values ranging between 1.5 and 0.9 mm/yr. In the second case, we obtain values ranging between 1.1 and 0.5 mm/yr. This brings the satellite records down into the zone of likely global mean sea level changes of ±0.0 to +2.0 mm/yr [42] , marked in dark orange in Figure 10.

The original satellite record did not exhibit any trend; just a variability around a zero value from 1992 to 2000 [50] [58] . This represents a measured sequence before corrections started to be applied; first a 2.3 mm/yr jump in 2003 and then an additional 0.8 mm/yr jump in 2008 [30] . If one would be able to identify the original, pre-correction, sequence in the present curves of NOAA [47] and University of Colorado [48] , one might obtain a better holding on what value one actually need to subtract in order fully to remove the erroneous GIA corrections.

In Figure 11, I have tried to identify the pre-correction trend 1992-2000 in the satellite altimetry record of NOAA [47] , and extend this trend over the entire period of recording up to 2015. I have also added the trend of GRACE [57] of −0.12 mm/yr for the period 2003-2008. It runs virtually parallel to the extended pre-correction

line. Then I tilted the whole graph into a horizontal position with respect to the extended pre-correction graph. I claim that this provides a good representation of the satellite altimetry record after removal of all erroneous “corrections”. Now the remaining sea level trend is 0.45 mm/yr. This is in very good agreement with observational facts from tide gauges and coastal morphology from all over the globe [30] [36] [38] [42] [50] [51] . Furthermore, it is in full agreement with a previous conclusion [42] that the true global eustatic component most likely is to be found within the zone of ±0.0 to +1.0 mm/yr.

In Figure 12, the same thing has been done with the satellite altimetry record of University of Colorado [48] . When the GIA corrections are removed, the remaining curve gives a rise of 0.65 mm/yr, which implies good agreements with the Figure 10 data from tide gauges and global key sites [42] .

The mode of glacial isostatic deformation of the Fennoscandian shield [4] [13] [26] -[28] [33] indicates―beyond doubts―that the deformation took place via horizontal mass-flow in a low-viscosity channel. This implies a glacial isostatic deformation, which is fully compensated on the regional scale (i.e. model B of Figure 2). There remains no reason to advocate forces penetrating the globe and giving rise to global glacial adjustments of coasts and seafloors all around the globe (model A in Figure 2).

The relative sea level records in Sweden (i.e. the spectrum of dated synchronous shorelines recorded over

hundreds of kilometers and transferred in local shore level displacement curves) were successfully split up in their components of absolute glacial isostatic crustal movements and absolute eustatic sea level changes [26] [35] . The isostatic component gave fundamental information on rheological parameters [4] [27] [28] [33] and neotectonics [13] [59] [68] . The eustatic component allowed for global comparisons [26] [35] leading to the new concepts of eustasy [60] [61] , the geoid theory [11] , theory of differential rotation [10] [62] [63] , and well- founded views on the perspectives of future sea level changes [30] [42] [50] [67] .

The rheological parameters recoded as well as the global sea level data indicate that is high time to abandon the hypothesis of a global internal response to glacial loading (Figure 2(a)).

The above-mentioned relations between relative sea level observations and resulting scientific outcome are illustrated in Figure 13 from [13] .

Recent investigation of the rheological character of the upper mantle record the presence of a low-viscosity zone [64] [65] is in full agreement with the Figure 7 model (i.e. model B of Figure 2). So, even rheological data from outside northwestern Europe are in agreement with the presence of a low-viscosity channel contradicting a linear viscosity profile as required for global transfer of glacial isostatic loading (model A of Figure 2). This supports an abandoning of the globally isostatic loading model, in favour of a regional glacial isostatic model.

The second test of the models [16] refers to the far-field sea level data (section 4, above). Observational facts in favour of a global isostatic loading model are lacking. It seems significant that Houston and Dean [66] comparing GIA predictions [23] and actual tide gauge records at 147 far-field sites found “remarkably little correlation”.

The satellite altimetry records [47] [48] are claimed [70] to be “a proxy for ocean water volume changes”, but behind the curves are unspecified “corrections” hidden, applied by NOAA [47] and CU [48] in order to obtain the product they personally assumed to be the correct “proxy of ocean water volume changes”. There is a major problem, however: their satellite altimetry records differ by 100% to 800% from observed tide gauge measurements (Figure 10).

With the removal of GIA corrections (the basic long-term residual factor of 2.3 mm/yr as well as later additional corrections) from the satellite altimetry data (Figure 11 and Figure 12), we finally obtain agreements among global tide gauge data, costal morphology data and satellite altimetry data; all agreeing on a mean global eustatic sea level factor somewhere within the zone ±0.0 to +1.0 mm/yr. This is illustrated in Figure 14. The only data set which hangs far above the others is the IPCC predictions. Those data, however, refer to assumptions and model out-puts, and are, by no means, anchored in observational facts.

The final and general conclusion of this paper is firm and says: it is high time to abandon the idea of global isostatic adjustment, and to stop all kinds of GIA corrections of records of sea level changes (i.e. satellite altimetry, GRACE, tide gauges, etc.).

At the IUGG meeting in 1975 in Grenoble, France, the organization committee of the Geodynamics Project

asked me to arrange a meeting and excursion on the Fennoscandian uplift and related topics in 1977. The proceedings of this meeting became the benchmark book on “Earth Rheology, Isostasy and Eustasy” (Mörner, ed., Wiley & Sons, 1980). The work continued at the unit of Paleogeophysics & Geodynamics at Stockholm University, visited by numerous scientists and the organizer of major international excursions on uplift, neotectonics and Paleoseismicity (1999, 2008). I am indebted to two excellent reviews.