The freshwater balance of polar regions in transient simulations from 1500 to 2100 AD using a comprehensive coupled climate model
- First Online:
- Cite this article as:
- Lehner, F., Raible, C.C., Hofer, D. et al. Clim Dyn (2012) 39: 347. doi:10.1007/s00382-011-1199-6
- 386 Downloads
The ocean and sea ice in both polar regions are important reservoirs of freshwater within the climate system. While the response of these reservoirs to future climate change has been studied intensively, the sensitivity of the polar freshwater balance to natural forcing variations during preindustrial times has received less attention. Using an ensemble of transient simulations from 1500 to 2100 AD we put present-day and future states of the polar freshwater balance in the context of low frequency variability of the past five centuries. This is done by focusing on different multi-decadal periods of characteristic external forcing. In the Arctic, freshwater is shifted from the ocean to sea ice during the Maunder Minimum while the total amount of freshwater within the Arctic domain remains unchanged. In contrast, the subsequent Dalton Minimum does not leave an imprint on the slow-reacting reservoirs of the ocean and sea ice, but triggers a drop in the import of freshwater through the atmosphere. During the twentieth and twenty-first century the build-up of freshwater in the Arctic Ocean leads to a strengthening of the liquid export. The Arctic freshwater balance is shifted towards being a large source of freshwater to the North Atlantic ocean. The Antarctic freshwater cycle, on the other hand, appears to be insensitive to preindustrial variations in external forcing. In line with the rising temperature during the industrial era the freshwater budget becomes increasingly unbalanced and strengthens the high latitude’s Southern Ocean as a source of liquid freshwater to lower latitude oceans.
KeywordsFreshwater balanceHydrological cyclePolar climateLittle Ice AgeExternal forcing
2 Model and methods
The model used for the simulations is the Community Climate System Model version 3.0.1 (CCSM3), which was developed by the National Center for Atmospheric Research (NCAR) (Collins et al. 2006). The model includes four components, namely the atmosphere, ocean, sea ice, and land surface, all linked through a coupler which exchanges fluxes (without flux corrections) and state information. We selected the low-resolution version: 3.75° by 3.75° in the atmosphere (26 vertical levels) and on land and a sea ice and ocean resolution of about 3.6° in longitude and 0.6° to 2.8° in latitude with finer resolution in the tropics and around Greenland, allowing for an open passage through the Canadian Arctic Archipelago. The ocean has 25 levels and a rigid lid. The land model does not include a dynamic ice sheet model but is able to lose and grow snow and ice up to 1,000 kg m−2. Additional mass above this threshold is inserted into runoff. We discuss the implications of a missing ice sheet model for our study in Sect. 5.
More detailed descriptions of the model components are available at the NCAR CCSM webpage (http://www.cesm.ucar.edu/models/ccsm3.0). The version used here and its simulation of present-day climate is comprehensively described in Yeager et al. (2006).
2.2 Experimental setup
For this and other studies (Yoshimori et al. 2010; Hofer et al. 2011) a series of model simulations was conducted. In this paper we use the following two setups: a 1500 AD equilibrium simulation with perpetual 1500 AD conditions (hereafter CTRL1500) and an ensemble of four transient simulations from 1500 to 2100 AD (TRa1 to TRa4, as in Hofer et al. 2011).
Our analysis focuses on six 31-year periods with different forcing levels: the quasi-equilibrium period just after branching from CTRL1500 which corresponds approximately to the beginning of the Little Ice Age (Mann et al. 2009), when the solar forcing generally is lower than today (averaged over 1500–1530 AD, hereafter Initial Conditions); the period of the pronounced solar forcing minimum of the Maunder Minimum (1685-1715 AD); the Dalton Minimum (1800-1830 AD), when a solar forcing minimum coincides with two large eruptions of tropical volcanoes; a period when GHGs start to increase (1840-1870 AD, Preindustrial); a period of present-day climate (1960-1990 AD, Present-Day) and finally, a period at the end of the twenty-first century when the average radiative forcing is projected to be approximately 5 W m−2 larger than 1990 AD (2068-2098 AD, Future Projections). For the twenty-first century we apply the IPCC SRES A2 scenario (IPCC SRES 2000).
The CCSM3 has been thoroughly tested for various resolutions (Hack et al. 2005; Yeager et al. 2006; Kiehl et al. 2006; Bryan et al. 2006; Holland et al. 2006a, b). Generally, the low-resolution version differs more strongly from observations than the middle or high-resolution version. This is largely due to the fact that the middle and high-resolution versions both run with the higher resolved 1° ocean/sea ice component. This has implications for properties such as salinity, heat flux, sea ice extent and volume as will be discussed here.
In the Antarctic region the model generally tends to be too fresh as well (Fig. 3). Still, it is able to simulate the more saline areas of the Weddell Sea and Ross Sea but it underestimates the amplitude and spatial extent. Furthermore, the model produces a patch of relatively fresh water at the northern end of the Antarctic Peninsula which is not observed in reality. The pattern of the sea ice concentration around Antarctica generally agrees with observations, while the total sea ice extent and area are overestimated by about 23 and 22%, respectively, during 1990–1999 AD (not shown). Regionally, the sea ice is overestimated in the Atlantic sector and underestimated in the Ross Sea sector.
Ensemble mean freshwater budget of the Arctic Ocean
Serreze et al. (2006)
Holland et al. (2007)
3,300 ± 680
2,411 ± 360
2,265 ± 109
−1,300 ± 710
−868 ± 275
−687 ± 32
3,200 ± 110
3,162 ± 776
3,482 ± 187
Ocean transp. liquid
−3,450 ± 693
−2,899 ± 569
– without CAA
−250 ± 615
−1,388 ± 2,332
−395 ± 579
Ocean transp. solid
−2,460 ± 340
−1,967 ± 421
– without CAA
−2,300 ± 340
−1,841 ± 626
−1,967 ± 421
−710 ± 1,255
194 ± 331
177,930 ± 1,652
Sea ice storage
29,347 ± 2,557
3 Freshwater balance of the Arctic domain
3.1 Present-day climate
We mentioned the overestimated freshwater storage in the ocean and sea ice. The domain’s freshwater budget is balanced in part by export of sea ice and ocean freshwater. Thereby, the Labrador and Irminger Sea, and to a smaller extent the western Bering Sea, are the major areas of export for both sea ice and ocean freshwater (Fig. 7). The eastern part of the Bering Sea including the Bering Strait imports about as much ocean freshwater as is exported through the Irminger Sea, thereby providing the second largest source of freshwater to the Arctic domain. Finally, rivers crossing 60°N (e.g., Ob, Yenisey, Lena, see Fig. 7) close the land budget.
3.2 Preindustrial changes
The present-day freshwater budget differs significantly from those during the two preindustrial time periods (Initial Conditions and Maunder Minimum) when NH as well as Arctic temperatures were lower than during Present-Day (see Fig. 2). VQ and Q are significantly smaller during Initial Conditions and the Maunder Minimum, resulting in a decrease of Plnd, Pocn and Pice, i.e., total P over the domain. As a consequence of the reduction in Plnd, R is reduced as well. Changes in E during Initial Conditions and the Maunder Minimum are robust only over land where Elnd is significantly decreased. The sea ice volume is significantly larger in Initial Conditions and the Maunder Minimum than during Present-Day, resembling the colder conditions of the preindustrial era.
The sea ice volume maximum is reached during the Maunder Minimum when the annual sea ice growth exceeds the melt rate by approximately 300 km3 year−1. Not surprisingly, the sea ice export reaches a maximum as well during that time. The ocean liquid freshwater export (FW transport), on the other hand, is significantly reduced during Initial Conditions and the Maunder Minimum. However, the ocean as a whole is saltier, i.e., stores less freshwater than during Present-Day. This is largely because during the Maunder Minimum sea ice grows, resulting in a negative freshwater flux to the ocean surface. In other words, brine rejection from the sea ice growth makes the ocean saltier and overcompensates for the reduced FW transport. In fact, the sum of all fluxes to and from the ocean (through the surface and ocean-internal) during the Maunder Minimum is less than half of that during Present-Day (300 vs. 800 km3 year−1). Due to large internal variability, however, this difference is significant only in two ensemble members.
The detection of significant and robust changes within the preindustrial era (Initial Conditions-Preindustrial, 1500–1870 AD) proves to be difficult as the interannual variability of several budget terms is high; although in observations of the Arctic freshwater cycle interannual variability is even larger than in GCMs (Rawlins et al. 2010). Prior to Preindustrial the freshwater balance of the Arctic domain as a whole is stable with the only notable change being internal redistribution of freshwater masses. Table 2 compares selected quantities of the polar freshwater budgets for different preindustrial time periods: FW storage, sea ice storage, and Q are storage terms inside the domain, while FW transport, sea ice export and VQ are transport terms at the domain’s boundary. With the comparison of Initial Conditions, the Maunder Minimum, and the Dalton Minimum we try to determine whether these three periods of similarly weak external forcing differ significantly among each other.
In general, changes between Initial Conditions and the Maunder Minimum are more robust in the storage terms than in the transport terms. This can be seen by the significant reduction of FW storage in all and of Q in three of the ensemble members, respectively. In concert with that, the sea ice storage significantly increases in three members. This is consistent with the absolute numbers in Fig. 5 where sea ice grows at the expense of FW, while all transport terms at the domain’s boundary change only insignificantly. Comparing Initial Conditions to the Dalton Minimum, the decrease in FW storage is the only significant and robust change. Comparing the Maunder Minimum to the Dalton Minimum, the reduction of the sea ice storage is the only robust change. All other changes are either not significant or, if they are, do not show a consistent sign of change among the ensemble members. This allows for a first conclusion that the difference between the three periods Initial Conditions, Maunder Minimum and Dalton Minimum is characterized by minor redistributions of freshwater between the reservoirs of the ocean and sea ice. At the same time, the total amount of freshwater in the domain does not change notably.
The atmospheric terms Q, VQ, total P−E, and R (which is largely controlled by Plnd−Elnd) do not show a significant response in the Maunder Minimum, but gradually increase along with the radiative forcing from the Maunder Minimum onwards (Fig. 8c, e, f). During the Dalton Minimum, however, these terms experience an abrupt dip. This dip sets them back to values similar as during Initial Conditions and the Maunder Minimum, even though the radiative forcing from solar and GHGs was approximately 5 and 10% weaker during Initial Conditions and the Maunder Minimum than during the Dalton Minimum, respectively. This raises the question of the necessity of the volcanic forcing to explain the atmospheric state of the Arctic freshwater balance during the Dalton Minimum. On a shorter time scale, the atmospheric response to volcanoes has been investigated using a composite of the years following the 21 largest eruptions from 1500 to 2000 AD (not shown). There, no significant large-scale response on high latitude total P−E and VQ could be detected. However, P−E and VQ are reduced just at 60°N in the first 2 years after an eruption. This contributes to the simulated reduction in the fields of P−E and VQ during the Dalton Minimum, nevertheless, the solar forcing is the main driver. Note that while the direct radiative effects from volcanic aerosols are correctly represented in the CCSM3, it has its deficits in simulating the dynamical effects of volcanic eruptions. This is namely the absence of the NH mid- to high-latitude winter-warming pattern that would result from a positive phase of the North Atlantic Oscillation (Stenchikov et al. 2006). Interestingly, Schneider et al. (2009) found a winter-warming in idealized experiments with the medium resolution version of CCSM3 while the high (Stenchikov et al. 2006) and low-resolution version do not reproduce this pattern.
As the solar minimum of the Dalton Minimum ceases and GHGs start to increase the combined radiative forcing exceeds the preceding range since 1500 AD and so does the Arctic domain’s SAT (Fig. 2). As a consequence the atmospheric terms VQ, Q, and total P−E recover to pre-Dalton Minimum values and then continue to increase in line with the strengthening radiative forcing. At the same time trends in FW storage become positive and are better constrained, i.e., the ensemble members’ spread becomes smaller. This indicates a robust response to the change in forcing. Trends in sea ice storage, however, remain around zero until the end of the nineteenth century when they finally become negative in all ensemble members. This means that the increase in FW storage just after the Dalton Minimum cannot be attributed to a redistribution of the freshwater from sea ice to the ocean. It rather seems that the synchronous increases of total P−E and R after their forcing-induced minimum during the Dalton Minimum are responsible for the freshening ocean. This increase of total P−E and R is consistent with the significant and robust increase of VQ and Q from the Dalton Minimum to Preindustrial (Table 2).
This leads to the conclusion that the mean state of the Arctic freshwater balance during Initial Conditions, the Maunder Minimum, and the Dalton Minimum is fairly stable, the exception being small but significant differences in the freshwater reservoirs which seem to reflect the multi-decadal to centennial variations in the external forcing. During the sustained cold conditions of the Maunder Minimum a saltier ocean, more sea ice and less water vapor (compared to the onset of the Little Ice Age at 1500 AD) were the marked anomalies in the freshwater balance. During the Dalton Minimum a short-lived response of the atmospheric component of the freshwater cycle occurred. However, the small signal-to-noise ratio indicates the important role of the internal variability of the freshwater cycle.
3.3 Future projections
The changes in the freshwater budget projected for the Future Projections are substantial and very robust, as for every investigated budget term all four ensemble members show a significant difference in mean compared to our reference Present-Day. VQ is roughly 27% larger than during Present-Day, caused by the thermally induced increase of the moisture holding capacity. At the same time the volume transport remains largely unchanged, which is reflected also in the cyclone activity in the domain: Both the number and mean intensity of the cyclones stay constant in winter and even decrease slightly in spring, summer, and autumn for the Future Projections (not shown). This confirms that the strengthened atmospheric water import is due to the amount of water transported by cyclones rather than by an increase of their frequency or intensity. Cyclones are detected as minima in the 12-hourly 1,000 hPa geopotential height field and need to exceed a gradient of 20 gpm 1,000 km−1 across a surrounding area of 1,000 km × 1,000 km (a detailed description of the cyclone detection and tracking method is given in Raible et al. 2007, 2010). The additional precipitable water in the domain is the reason for an increase of total P. As annual mean sea ice extent is reduced by 12% compared to Present-Day (not shown) the open ocean area grows and receives much more P directly from the atmosphere (+43%). Runoff from lower latitudes and Plnd−Elnd increase as well, leading to a substantial increase in R to the ocean of roughly 26% (manifested in the positive anomalies close to the river’s discharge areas in Fig. 7a, bottom panel). Further, the annual mean melt-freeze ratio of sea ice rises from 1.05 during Present-Day to 1.15 during the Future Projections, representing an enhanced release of sea ice-stored freshwater to the ocean (due to the large sea ice volume at present-day no sea ice-free summer conditions are reached in these simulations). This leads to a spatial shift in the freshwater flux to the ocean surface, causing strong local anomalies in the GIN Seas (Fig. 7a, bottom panel). Together with R, sea ice is the main driver of changes in the ocean freshwater budget.
Overall, the freshwater flux from Pocn−Eocn and R to the Arctic domain’s ocean increases by 39% from Present-Day to the Future Projections. Consequently, the export of freshwater from the Arctic Ocean is strongly increased as well during the twenty-first century. Sea ice export, on the other hand, is decreased in all sectors and disappears completely in the eastern Bering Sea and in the Labrador sector (Fig. 7). The latter causes the FW storage in the GIN Seas to decrease as less sea ice is transported there. This mechanism has been identified to counteract temperature-driven stratification and thereby maintaining deep-water formation in the GIN Seas (Holland et al. 2006b). The vast rest of the Arctic domain’s ocean is freshening during the twenty-first century. The largest changes occur in the Beaufort Gyre, which strengthens and increases its freshwater storage in the future (Fig. 4). Correlation and trend analysis suggest that, due to its proximity to the Beaufort Gyre, the FW transport through the CAA is coupled directly to the growing FW storage in the Gyre (not shown). Together with increased surface freshwater forcing in the CAA (Fig. 7a, bottom panel) and a growing sea surface height gradient across the CAA (as described by Jahn et al. (2010) for interannual transport variability) this results in an increase of the FW transport in the Labrador sector over the twenty-first century. This increase is about three times stronger than the increase of FW transport in the Irminger Sea sector (Fig. 7). The freshwater import through the Bering Strait decreases in line with the weakening of the meridional pressure gradient across the Bering Strait due to the freshening Arctic Ocean (Arzel et al. 2008).
The findings for the mean changes from Present-Day to the Future Projections are generally in agreement with numerous model studies focusing on the Arctic Ocean only (Holland et al. 2006a, b; Koenigk et al. 2007; Arzel et al. 2008). In some respect, however, different models still perform differently. Koenigk et al. (2007) project the Arctic Ocean total freshwater export to stay fairly constant until the twenty-second century, i.e., the decrease in sea ice export is matched by the increase in the liquid export. We find, when considering either the Arctic Ocean only or the complete Arctic domain north of 60°N, that the ocean increases its total export of freshwater significantly already during the twenty-first century. Arzel et al. (2008) discovered a threshold behavior of the Fram Strait throughflow as a consequence of ocean-atmosphere heat exchange in an A1B simulation which was not reproduced in our simulations. Here, the Fram Strait volume flux stays constant. The increase in FW transport, which we find in the Fram Strait, is mainly due to fresher waters transported and not due to changes in the volume flux (not shown). The proposed mechanism of Arzel et al. (2008) requires a strong reduction of sea ice in the Barents Sea at the beginning of the twenty-first century and a freshening of the GIN Seas in the late twenty-first century. While the first of these features occurs to a lesser extent (not shown), the second is not simulated at all in the CCSM3 (see Fig. 7).
4 Freshwater balance of the Antarctic domain
4.1 Present-day climate
The Antarctic domain is characterized by a completely different geographic setting compared to the Arctic. 63% of the domain’s area as well as the domain’s lateral boundaries at 60°S are covered by ocean, which—together with a larger atmospheric mass transport (not shown)—results in more water vapor advected (VQ) to the Antarctic domain than to the Arctic domain (Fig. 6). Due to the dominant circumpolar circulation this import of water vapor is uniformly distributed along 60°S with a maximum in the sector north to the Ross Sea (Fig. 7). This corresponds well to observations (Tietäväinen and Vihma 2008). Total Q is smaller than over the Arctic domain owing to the lower temperatures, especially south of 70°S, i.e., over the Antarctic ice sheet (not shown).
For the larger part, P falls on the sea ice and ocean surface while the Antarctic ice sheet receives only about 19% of total P during the reference period Present-Day. Due to the lower temperatures the portion of P that is evaporated to the atmosphere is smaller than in the Arctic domain. P over the ice sheet falls mainly as snow (not shown), which means that R is made up solely by the melt flux from the terrestrial snow layers.
During Present-Day the sea ice freshwater balance is positive as illustrated by the sea ice melt-freeze ratio of 0.96 (800 km3 year−1). However, as sea ice is exported at a rate of about 6,100 km3 year−1 (more than a factor three of the Arctic domain’s export rate during Present-Day) the sea ice volume shrinks nonetheless during Present-Day. In agreement with Schmitt et al. (2004), the vast majority of the sea ice export takes place north of the Weddell Sea (Fig. 7). As the FW transport in the Antarctic domain has to be calculated from the residuals of the FW storage changes and the ocean surface fluxes, we cannot split it up into sectors. However, investigating a few years of sub-daily output the model has proven to be capable of simulating the Drake Passage effect (e.g., Talley 2008), i.e., the major freshwater export due to Ekman transport in the Drake Passage and the smaller return flow in the Atlantic sector.
4.2 Preindustrial changes
Looking at the preindustrial time periods Initial Conditions and the Maunder Minimum, the mean differences compared to Present-Day are generally less robust than in the Arctic domain. This could already be anticipated from the smaller preindustrial temperature anomalies in the Antarctic domain (Fig. 2). Nevertheless, we find significant reductions in VQ and Q during Initial Conditions and the Maunder Minimum (Fig. 6). Thereby the changes in VQ are uniformly distributed across the sectors, i.e., no large changes in certain sectors dominate the total change in VQ (Fig. 7). This reduced VQ results in less total P and P–E, in turn influencing the FW transport, which is significantly smaller during the Maunder Minimum in three of the ensemble members. R and the export of sea ice remain unchanged. The melt and freeze rates of sea ice are significantly different compared to Present-Day, but not the ratio of the two. Therefore, sea ice production during Initial Conditions and the Maunder Minimum is only about 100 km3 year−1 larger than during Present-Day. As a consequence the difference in sea ice volume between the preindustrial time periods and Present-Day is small, but significant in three ensemble members nevertheless (Fig. 6). When comparing with the first half of the twentieth century instead of Present-Day the difference quickly becomes insignificant. This confirms results by Sedláček and Mysak (2009), who found no significant long-term trend in simulated Southern Hemisphere sea ice volume between 1500 and about 1950 AD (see also our Fig. 8i). This means that the ocean and sea ice freshwater budget during Initial Conditions, the Maunder Minimum, and Present-Day are primarily controlled by the atmospheric input and stay approximately in balance. Note, however, that the unstable behavior of the FW storage in the CTRL1500 (Fig. 8h) points to a potential inaccuracy of the FW transport calculated from monthly means.
As in the Arctic domain, most variations within the preindustrial era in the Antarctic freshwater balance are not robust. In fact, only the reduction of the FW storage from Initial Conditions to the Maunder Minimum is significant in all four ensemble members (Table 2). At the same time, two members show a significant increase in sea ice storage. As no notable change in the total amount of freshwater in the domain is detectable, i.e., the fluxes at the boundary stay constant, this corresponds to a shift of freshwater from the ocean to the sea ice. However, the signal-to-noise ratio does not allow for an attribution to the external forcing. The same holds true for the comparison of Preindustrial with Initial Conditions, the Maunder Minimum, and the Dalton Minimum: no robust signal can be detected for any quantity of the budget as either significance is not reached or the ensemble members show significant, though opposed trends. This is not surprising as the Antarctic SAT during preindustrial times does not resemble the forcing variations such as the Maunder Minimum or the Dalton Minimum.
Similar to the high northern latitudes, the volcanic eruptions during the Dalton Minimum do not leave a significant imprint in the P–E and VQ fields south of 60°S. This result is confirmed by the mean response of P–E and VQ to the 21 largest eruptions from 1500–2000 AD in our simulations, as well as a case study of the atmospheric circulation response to Pinatubo (Robock et al. 2007). The response is even weaker than in the high northern latitudes as the larger ocean area of the SH dampens the radiative perturbation from the erupted aerosols (Robock 2000) and as the stronger polar vortex of the SH is thought to be more resistant to perturbations (Stenchikov et al. 2002).
4.3 Future projections
For the Future Projections the changes are, similar to the Arctic domain, very robust as nearly all budget and flux terms in all ensemble members experience a significant shift in mean. The domain is fed by 20% more water vapor compared to Present-Day which mainly arises from increased moisture content (Liu and Curry 2010). The largest increase in VQ occurs in the Bellinghausen Sea (Fig. 7) where the frequency of occurrence of cyclones decreases, but on average cyclones carry more moisture into the domain (not shown, see also Lambert and Fyfe 2006; Lynch et al. 2006). As in the Arctic, the majority of this additional freshwater is precipitated onto the open ocean surface, which is enlarged due to the reduction of the sea ice area (−11%, not shown). The sea ice storage is reduced by 32%, indicating that sea ice is substantially thinned. This increases the freshwater flux to the ocean surface in large regions of the Southern Ocean south of 55°S (Fig. 7b, bottom panel). While sea ice export weakens in nearly all longitudinal sectors (Fig. 7), the export by FW transport strengthens. The latter dominates these changes as export substantially exceeds the import during the twenty-first century.
5 Summary and conclusion
We use an ensemble of four transient simulations with the low-resolution version of the CCSM3, spanning the time period from 1500 to 2100 AD, to investigate past and future changes of the polar freshwater balance. The simulations produce a realistic general NH temperature evolution over the past five centuries with an overestimated warming in the twentieth century due to missing aerosols (Hofer et al. 2011). Temperatures on the poles, however, are rather underestimated. The simulated present-day freshwater cycle of the Arctic Ocean agrees reasonably well with observations (Serreze et al. 2006), except that freshwater stored in sea ice and ocean is overestimated significantly.
We show that many of the transport and storage components in the present-day polar freshwater cycle are significantly different from preindustrial states and in particular from the projected future state. While this conclusion is valid for both polar domains the relative magnitude of the differences is larger in the Arctic where temperature anomalies in the past and future are larger as well.
Regarding the Arctic, the simulated freshwater cycle appears to be closely coupled to temperature on decadal time scales. We showed this for the twentieth and twenty-first century in consistency with other model studies (Goosse and Holland 2005; Holland et al. 2007) as well as observational records (White et al. 2007). This relationship was found to hold beyond the temporal scope of the studies mentioned before as during the cold phase of the Little Ice Age simulated freshwater transports are smaller than during present-day and the domain contains less freshwater in total. As temperature rises during the twentieth and twenty-first century more water passes through the hydrological cycle and the reservoir of FW grows at the expense of the reservoir of sea ice and due to increased P−E and R. Synchronously, the FW transport increases and represents a strengthening freshwater forcing to the North Atlantic. These features are consistent among most coupled GCMs (Holland et al. 2007), however, the magnitude of change depends on the model and its resolution. This is particularly evident for the reservoirs of sea ice and FW. Compared to the high-resolution CCSM3 the low-resolution has a weaker ocean heat transport into the Arctic and a weaker FW transport out of the Arctic. Resulting from that, sea ice is overestimated in this study, leading in turn to a lagged peak of freshwater release in the future. At the same time, the weak FW transport yields a reservoir of FW that grows faster than in the high-resolution CCSM3. Based on this, the magnitude of preindustrial changes might be model-dependent as well, which needs to be considered when interpreting these results.
Within the preindustrial era of 1500–1870 AD the Arctic domain’s total amount of freshwater is largely controlled by VQ while for the present and future the changes in FW transport and sea ice export contribute crucially as well.
Sedláček and Mysak (2009) showed that during preindustrial times wind-stress plays an equally important role as direct radiative effects regarding changes in sea ice volume and ocean properties in the Arctic. At the same time, the sea ice volume was found to be basically insensitive to changes in radiative forcing as they occurred during the Maunder Minimum and the Dalton Minimum. We are unable to depict this hypothesis as precisely as the sensitivity study of Sedláček and Mysak (2009). At least we see a weak but significant response of sea ice storage and FW storage to changes in the external forcing occurring from 1500 AD to the Maunder Minimum. The Dalton Minimum, on the other hand, leaves no imprint on the sea ice and ocean freshwater storage, as its duration is too short to significantly affect these slow-reacting reservoirs. The atmospheric quantities of total P−E, VQ, and Q, on the other hand, display a clear dip during the Dalton Minimum, forced primarily by the sun.
As our domain 60−90°N includes Greenland, in contrast to many other studies focusing on the Arctic Ocean only, we need to account for potential melt fluxes from the Greenland ice sheet when assessing the present-day and future state of this domain’s freshwater balance. The processes of a melting or disintegrating ice sheet are not simulated by the land component of the CCSM3 due to the absence of a dynamic ice sheet model. However, recent estimates of both present-day (267 ± 38 km3 year−1; Rignot et al. 2008) and future (year 2100 AD, 220 km3 year−1, Ren et al. 2011; >400 km3 year−1, Graversen et al. 2011) Greenland mass balance are a magnitude smaller than the changes projected for other freshwater fluxes such as P, R, or the melting of sea ice. The projected acceleration of the Greenland mass loss by about 100−200 km3 year−1 over the twenty-first century even lies within the ensemble uncertainty for, e.g., changes in R or sea ice melting. Therefore, the absence of an ice sheet model does not render the Arctic freshwater balance of the CCSM3 unrealistic. Similar conclusions can be drawn for the Antarctic domain, where recent literature suggests the twenty-first century contribution of freshwater from the ice sheet (<500 km3 year−1, Pfeffer et al. 2008; Katsman et al. 2011) to be substantially smaller than changes in other components of the ocean’s surface freshwater balance, e.g., P − E (+1,400 km3 year−1). However, the estimates and projections from current literature for the Antarctic mass balance are far less robust than the ones for Greenland. Taking into account all possible scenarios in Pfeffer et al. (2008) and Katsman et al. (2011) it remains uncertain, whether the Antarctic ice sheet will retain a close-to-zero mass balance or become a significant source of freshwater within the twenty-first century.
The Antarctic domain’s response to changes in external forcing is expected to be dampened compared to the Arctic. This is due to the more zonal circumpolar current in both atmosphere and ocean, hampering the heat exchange with lower latitudes, as well as the larger ocean area, acting to reduce amplitudes in surface warming due to changes in the radiative balance (IPCC 2007). Additionally, the ice and snow albedo feedback is smaller in the Antarctic domain (Masson-Delmotte et al. 2006). This was confirmed in our study as anomalies in the freshwater cycle are smaller than in the Arctic.
During the preindustrial period no significant and robust changes in the individual terms of the Antarctic freshwater balance could be detected from our ensemble, suggesting that minima in external forcing (Maunder Minimum, Dalton Minimum) do not leave an imprint on the SH high latitudes’ water cycle. As FW transport and sea ice export do not change much during the first 450 years, the atmospheric import VQ is the controlling mechanism regarding the domain’s total freshwater content. The future projections then show a substantial acceleration of the hydrological cycle, affecting all fluxes, with a linear dependence on the SAT projections. Thereby, the strongly increasing FW transport strengthens the Antarctic domain as a freshwater source to the lower latitude oceans. Liu and Curry (2010) stress the potentially too weak response of the twentieth century Southern Ocean climate to anthropogenic forcing in models due to an overestimated sea surface temperature internal variability. Taking into account this and a possible lack of decadal SAT variability in the southern high latitudes of the model, the simulated amplitudes in the budget and transport terms of the Antarctic domain have to be interpreted with caution as they might have been larger during preindustrial times than simulated in this study. Therefore, other millennial simulations with new and higher resolved models are in need, which – together with proxies – will add to the picture of past variations of the polar freshwater reservoirs.
We gratefully acknowledge Andreas Born for valuable discussion, as well as two anonymous reviewers for constructive comments. We are grateful to the ECMWF, the NASA GISS, the NOAA ESRL, the ESG, Igor Polyakov, and Marika Holland for providing access to data and to the NCAR for providing the code of the CCSM3 model. This study is supported by the National Centre of Competence in Research (NCCR) Climate funded by the Swiss National Science Foundation, and the European Commission Past4Future project. The simulations were performed on an IBM Power 4, a CRAY XT3, and a CRAY XT5 at the Swiss National Supercomputing Centre (CSCS) in Manno.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.