1. Introduction
Climate studies suggest that the Arctic will experience the most severe changes under global warming conditions (Hartmann et al. 2013; Stjern et al. 2019). These changes are communicated mostly in terms of temperature increase, and are additionally fueled by the positive ice–albedo feedback. However, changes will involve substantial impacts on all climate parameters, including permafrost thawing, the hydrological cycle, and, last but not least, sea level change.
Most observation-based studies of Arctic SSH variability focus on time scales shorter than multidecadal, as there is not much high-quality data available on time scales longer than about 70 years (Holgate et al. 2013; PSMSL 2022). Since the larger component of sea surface height (SSH) variability in the Arctic is halosteric SSH (Antonov et al. 2002; Xiao et al. 2020), studies of freshwater content fluctuations on centennial time scales could fill the gap, but they are also in short supply. One of the only ways to explore multicentennial SSH variability in the Arctic, and the mechanisms that contribute to it, is to look at simulations of coupled climate models and see what variability and mechanisms they suggest could be possible.
Previous studies of Arctic SSH variability and change have found that sea level pressure changes, such as those associated with the Arctic Oscillation, could drive basin-scale SSH variability on longer than interannual time scales through the associated low-frequency wind forcing (e.g., Proshutinsky and Johnson 1997; Proshutinsky et al. 2002). Low-frequency variability from wind forcing has also been examined by Köhl and Serra (2014) based on the idea by Joyce and Proshutinsky (2007) to describe the transport through Davis Strait by Godfrey’s Island Rule (Godfrey 1989) applied to Greenland. Godfrey’s Island Rule is a calculation of the Sverdrup transport, which depends on the wind stress curl, for the case of an island present in the basin. The skill of applying the island rule suggests wind stress curl as the most relevant driver of the variability.
Observational studies have regarded the change of accumulation of freshwater in the Beaufort Gyre in the transition from an anticyclonic atmospheric circulation regime to a cyclonic circulation regime as the key factor regulating the freshwater content of the Arctic (e.g., Proshutinsky et al. 2002). Joyce and Proshutinsky (2007) found that the modulation of freshwater export into the Atlantic through Davis Strait explains the majority of the interannual variations in response to the transition from an anticyclonic circulation regime to a cyclonic circulation regime around Greenland. In addition to wind forcing, sea ice thickness and transport variability can generate large salinity anomalies in sporadic events (Haak et al. 2003), which can translate into low-frequency freshwater content variability. Further factors influencing the Arctic freshwater anomalies have been summarized by Haine et al. (2015) and include changes of precipitation, runoff, and changes of the sea ice volume.
While these explanations consider only internal variability of climate system changes, external forcing associated with solar forcing variability and volcanic eruptions are also important drivers of decadal climate variability (Shindell et al. 2003; Stenchikov et al. 2009; Segschneider et al. 2013; Zanchettin et al. 2012, 2013) and multidecadal-to-centennial fluctuations in sea ice (Zhong et al. 2011), and are necessary to explain the largest preindustrial climate changes (Jungclaus et al. 2010). These are known to affect not only temperature but also precipitation (Gu and Adler 2011). In addition to this, small variability in external forcing can also act as a pacemaker for internal climate variability (e.g., Otterå et al. 2010). In this respect Johnson et al. (2018) see the recent increase in Arctic freshwater storage mostly driven by natural variability, while Jahn and Laiho (2020) have argued that it is related to climate change as predicted by climate models (e.g., Koenigk et al. 2007).
Among previous publications examining, at least in part, Arctic SSH in models, Pardaens et al. (2011) published a model study of factors influencing projected changes in regional SSH over the twenty-first century. In general, the authors find that halosteric changes dominate SSH changes in the Arctic and that strong compensation between thermosteric and halosteric changes characterizes a large section of the subpolar North Atlantic. The magnitude of SSH and relevant steric changes in the North Atlantic appear to be linked to the amount of Atlantic meridional overturning circulation (AMOC) weakening. Landerer et al. (2007), in an earlier version of the model in this study, also found increased Arctic sea level rise in future projections due to increased precipitation.
In an earlier study, Carson et al. (2015) analyzed multicentennial control-run model output and identified centennial-scale natural variability in sea level as a substantial contribution to sea level variability and showed that this holds especially in high latitudes, whereas subtropical regions appear more dominated by decadal variability. We build on this earlier study that provided the motivation to understand the drivers of centennial-scale Arctic SSH variability in a specific model, as a case study. The goal of the present paper is to investigate the physical causes underlying those centennial-time-scale sea level changes simulated to naturally occur in the Arctic, and particularly to find the dominant processes controlling low-frequency freshwater variability.
2. Methods and data
The past1000 model experiment was produced with the Max Planck Institute for Meteorology Earth System Model in a configuration called MPI-ESM-P, similar to the low-resolution configuration (MPI-ESM-LR; Jungclaus et al. 2013). It is a CMIP5 experiment simulating the past 1000-yr historical period between 850 and 1850 (Braconnot et al. 2011). Given that there is a grid pole in southern Greenland, the resolution for this low-resolution model setup (the GR15 grid version of the model described in Jungclaus et al. 2013) is best over the Arctic, yielding grid cells with about 15-km spacing near Greenland extending to about 80 km by the continental shelves near the Bering Strait. This model experiment is similar to the previous Millennium experiment (Jungclaus et al. 2010), with the main difference being the higher resolution in the newer model. The forced run and its data are referred to hereafter as “past1000,” and the control run as “Ctl-P” from MPI-ESM-P’s piControl (preindustrial control) experiment.
The MPI-ESM-P model should be able to represent aspects of the Arctic climate system well enough for our results. For example, salinity biases are not expected to impact the variability of ocean freshwater transport although salinity biases exist within the Arctic, as the salinity contrast between inflowing North Atlantic water and outflowing Arctic water is large (Jungclaus et al. 2013). Biases could also affect EOF amplitudes, but the large contrast between the fresh surface and higher salinity at depth should minimize this effect. Sea ice is in good agreement with observations (Jungclaus et al. 2013). However, realistic responses to forcing in the past1000 run are difficult to evaluate (Bothe et al. 2013). Still, some low-frequency variability patterns, such as the North Atlantic Oscillation (NAO) in the atmosphere, are well produced by this model (Domeisen et al. 2015; Koul et al. 2019). Atmospheric variability plays an important role in our results.
Reproduction of Fig. 9 from Schmidt et al. (2011), showing the forcing components to be used in CMIP5 past1000 experiments. The climate forcings in the past1000 run are dominated by large infrequent volcanic events, but there is multidecadal to centennial-scale variability in the solar and greenhouse gas forcings as well. The time series are smoothed by a 21-yr Hamming window. Note the large volcanic forcing starting shortly before 1250, and continuing afterward for nearly 100 years, as this will be important later.
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
Arctic integrated values are calculated by area-weighted mean within the Arctic, with boundaries defined by the Bering Strait, the straits along the edge of the Labrador Sea (so it is not included), and two lines in the North Atlantic: one mostly along 80°N between Greenland and Svalbard, and one between Svalbard and Norway, both along an unbroken line of grid boxes.
As stated before, all data were averaged into annual values, but are anomalies relative to the total record’s mean value—e.g., all figures showing SSH components or freshwater and its components are plotted around a zero mean, except for one figure of sea ice transport, which is discussed near the end. Data were detrended right before final statistical analysis and graphing. The paper’s aim is only to look at variability, not mean behavior or trends in the data. Last, data are usually smoothed with either 111-yr boxcar windows or 11-yr Hanning windows to isolate decadal and centennial-scale variability better. Some time series are noisier than others, and smoothed time series are so noted in the figures. The 111-yr window was chosen over a 101-yr window to slightly improve the suppression of multidecadal variability much shorter than 100 years, though the results turned out to be essentially the same either way. For centennial-scale variability, we simply take standard deviations of low-pass-filtered data so obtained.
In our previous work (Carson et al. 2015), we showed that median values of centennial-scale variability in control runs between 21 models in a CMIP5 ensemble point to heightened variability in the Arctic. A reexamination of the data used in that study shows that close to 50% of the 21 models investigated contain as much, or even more, regional Arctic SSH variability on centennial time scales as the MPI-ESM-P control run, which provided the motivation for this study. The number of models in the CMIP5 ensemble in Carson et al. (2015), with as much or more Arctic SSH variability as the current model analyzed here, is dependent on precisely how one defines and computes low-frequency variability, but it occurs in 9 out of 21 models in the study when based on the variability of 100-yr trends using the method described in Carson et al. (2015). This is to say that the high centennial timescale Arctic variability in the MPI-ESM-P model is common, although the details of the model responses leading to enhanced Arctic SSH variability could be different to varying degrees between CMIP5 models’ past1000 experiments.
In the forced past1000 run for MPI-ESM-P, there is a stronger response of Arctic SSH variability, and freshwater content variability, due to the additional forcing (volcanic + solar) in the past1000 experiment compared to the control run, about 37% more (Fig. 2). This in turn yields an enhanced steric sea surface height variability in the Arctic, though local thermosteric SSH is of greater importance in high-latitude regions in the Southern Hemisphere, which we do not explore in this paper.
Centennial-scale variability calculated from the standard deviation of 111-yr running means of sea surface height for (top) the MPI-ESM-P past1000 CMIP5 experiment, with a nominal resolution of 1.5°, and (bottom) for the MPI-ESM-P’s piControl experiment. The piControl run yielded 1156 years of output. In addition to the strongly varying western boundary current extension regions in the North Pacific and Atlantic, the subpolar and polar regions have the highest decadal variability.
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
3. Results
a. Past1000 centennial-scale freshwater content variability
In the next sections, we show that, from the MPI-ESM-P model output, SSH variability is dominated by halosteric SSH variability, or equivalently freshwater content (FWC) changes. Generally speaking (for model and observational data), Arctic halosteric SSH varies more than thermosteric SSH due to the larger range of halosteric contraction versus the smaller range of thermosteric expansion for the salinity and temperature values commonly found there (Antonov et al. 2002; Xiao et al. 2020). First looking at the coherent large spatial patterns of SSH (ZOS) variability, we found via empirical orthogonal function (EOF) analysis (Fig. 3) that low-frequency variability—multidecadal and longer—is dominant for the first two EOFs, and describes a substantial portion of the total Arctic variability near the Canadian Archipelago and poleward of Svalbard and Franz Josef Land. The EOFs of control-run SSH (Figs. 3c,d,g,h, on the right), also exhibit some low-frequency behavior, but it is weaker than in the forced run, a bit more localized, and the variability occurs more in multidecadal time scales shorter than a century. Nevertheless, we note that there is some degree of low-frequency variability inherent in the system when driven by the same external forcing every year. Comparing the EOFs of the two runs reveals that the EOF1of past1000 (Fig. 3a) resembles EOF2 of Ctl-P (Fig. 3g), and EOF1 of Ctl-P (Fig. 3c) resembles EOF2 of past1000 (Fig. 3e). These similarities suggest that the external forcing in particular enhances the internal mode 2 variability (Fig. 3g), therefore becoming EOF1 in the forced run (Fig. 3a), which is characterized by a more spatially uniform variability. Note that the EOFs are only carried out over the Arctic as we have defined it here, and can be seen by the extent of filled gridbox values (most visible in Fig. 3a). All Arctic integrated values later described are area-averaged over this same region.
(a)–(h) EOFs of decadal SSH over Arctic regions, calculated using 11-yr running means, with the grid boxes scaled by area before computation, with (left) the first two EOFs from the forced run and (right) those from the control run (CTL-P) Insets in (a) and (d) show the first eight eigenvalues of the first eight EOFs/PCs. The percentage of variance the EOF/PC explains is shown in the upper-right corners of the EOFs. Centennial-scale variability is evident over much of the Arctic region in both (a),(b) EOF/PC 1 and (e),(f) EOF/PC 2, describing ∼30% and ∼16% of the variability in the forced run, respectively. [The small white “†” sign in panel (a) is a single location referenced in a subsequent figure.]
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
Arctic integrated SSH changes track halosteric SSH changes very closely in both runs, with freshwater content–halosteric Arctic integrated time series correlations: forced run = 0.97 (90%CL: [0.93, 0.98]), control run = 0.99 (90%CL: [0.998, 0.987]), which demonstrate the strong connection between the two properties (Fig. 4, top panels). The range is about ±2–3 cm in the control run, but large excursions occur much less frequently in the control run compared to the forced run (top-right panel). The forced run contains approximately 38% more low-frequency variability than the control run (as calculated using standard deviations of 11-yr running means).
Time series of total SSH and halosteric SSH (in cm) of (top) the area-averaged mean SSH for the entire Arctic region, and (bottom) the SSH at a specific grid box near Svalbard, marked with a small white “†” sign in Fig. 3. The corresponding freshwater content time series (in m) is shown in red. Columns show (left) the forced-run (past1000) experiment data and (right) the piControl run. The “†” sign for the control run represents data from the same grid box as in the forced run. All time series are of 11-yr running-mean data.
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
For individual regions, such as in a geographical position poleward of Svalbard (marked by a cross sign in the map in Fig. 3a), the halosteric signal can be much larger than the total SSH (Fig. 4, bottom panels), which directly points to some compensation by thermosteric and nonsteric SSH. Some compensation from thermosteric SSH can be expected, as warmer salty water (e.g., Atlantic-sourced water) or colder fresher water (precipitation, runoff, ice melt, or Pacific-sourced water) are often advected into this region. Under these conditions, the SSH signal can range nearly up to ±10 cm, and the halosteric signal up to ±20 cm. Generally, thermosteric SSH changes tend to be much smaller than halosteric changes in both the Arctic integrated values (Fig. 5b) and in individual locations, as stated earlier, and was also found by Xiao et al. (2020) in a higher-resolution model.
(a) Arctic integrated nonsteric SSH (green) plotted along with global-mean steric SSH, which is multiplied by −1 (red) (in cm). (b) Arctic integrated halosteric (blue), thermosteric (magenta), and nonsteric (green) contributions to SSH (in cm). The nonsteric component is the Arctic SSH residual after subtracting the total steric (halosteric plus thermosteric) sea level from the “zos” variable. (c) The same components, but only for the grid box marked with a “†” symbol in Fig. 3, and the same location as in the bottom panels of Fig. 4. All data shown are 11-yr running means.
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
The thermosteric and nonsteric (i.e., changes in mass) components of SSH are covered next. The nonsteric portion of SSH in the forced run of the model, that is, the residual from SSH after subtracting the thermosteric and halosteric components from it, is also generally smaller than the halosteric, varying between −1 and +1 cm in the Arctic integrated values with a couple of exceptions (Fig. 5b, green time series). Because global nonsteric SSH results from the difference between SSH and the steric component, from which we had removed its global mean, the global nonsteric SSH has to be zero as this is true for SSH. The steric global mean has substantial variability through the imprint of thermosteric SSH changes due mostly to volcanism (cf. red line in Fig. 1 with Fig. 5a). The Arctic integrated nonsteric sea level has some minor correlation with the global steric mean, about −0.35, coinciding with the larger changes in the global mean. The nonsteric SSH is a larger signal than the Arctic integrated thermosteric SSH (magenta time series), so it is the second largest SSH component after the halosteric SSH. The Flux-Anomaly-Forced Model Intercomparison Project (FAFMIP; Gregory et al. 2016) is a model intercomparison designed to isolate model differences in responses to various anomaly forcings and flux parameter implementations. Analysis of these experiments by Couldrey et al. (2021) suggest that a positive global mean steric signal in response to global warming is associated with a negative nonsteric signal in the interior of the Arctic, and a larger positive nonsteric amplitude over the shallow shelves around the sides of the basin. A more detailed investigation of one of the FAFMIP experiments by Zhang et al. (2022) shows that the redistribution of mass within the Arctic accounts for only a little more than one-third of the mass gain over the shelves, implying import of mass (i.e., nonsteric SSH) into the Arctic related to global warming and therefore a positive correlation between the global steric mean and the Arctic integrated nonsteric SSH, different from the present results. Our results, presented in a later section, include volcanic cooling via the associated reduction of freshwater fluxes as one of the causes of large freshwater changes in this model. The volcanic cooling in this model and the doubling CO2-related warming the FAFMIP experiments try to represent could cause differences in the SSH fingerprints that, apart from just the difference in sign, could yield regionally different contributions to the SSH response, and differences in model resolution might play a part as well.
The Arctic nonsteric signal does not directly covary with the halosteric or total SSH in a consistent way, with the large anomalous stand in the second half of the 1200s of opposing sign, and the decrease before 1600 coinciding with the halosteric decrease over the same period (cf. Fig. 4a and Fig. 5b). Although the variability of the freshwater fluxes related to volcanic forcing is very relevant for Arctic sea level through changes in the salinity of the Arctic waters, its associated mass flux signal is different from what is suggested by Peralta-Ferriz and Morison (2010), and is not important because any mass anomaly would quickly propagate via barotropic waves to the rest of the World Ocean (Volkov and Landerer, 2013). As the focus in the following analysis is on freshwater content variability, freshwater sources are explored in more detail, while nonsteric and thermosteric SSH are not explicitly examined further.
As illustrated in Fig. 3a, a large amount of the forced SSH variability is nearly basinwide. The two EOFs (Figs. 3, all panels) describe substantial but different portions of the total SSH variability providing a picture of large low-frequency freshwater changes locally within the Arctic, as also seen in Fig. 4 (bottom panels). Some of the local changes average together to yield a smaller Arctic integrated SSH (Fig. 4, top panels) and are suspected to be due to circulation of various water masses (surface freshwater, North Atlantic source water, etc.) within the Arctic, though a detailed analysis of this phenomenon is not pursued here.
Since much of the Arctic is experiencing anomalously high or low SSH stands concurrently on centennial time scales, due mostly to large halosteric SSH changes, the sources of freshwater changes, averaged over the Arctic Ocean, were budgeted and examined. The model diagnostics provide only the surface freshwater flux from the atmosphere into the ocean (through the WFO variable—water flux into the ocean); the contribution from the freezing and melting, including the contribution of the freshwater flux from the atmosphere via the sea ice into the ocean, had to be accounted for by the time change of the sea ice volume with the added consideration of the precipitation on ice and the lateral transport (export) of sea ice. In balancing the ice volume within the Arctic, it turns out that the ice transport TRANSIX and TRANSIY variables balance the unknown freezing and melting of sea ice, and precipitation onto sea ice, for the most part. This is because the variability of freshwater storage in sea ice (reported in the SIM sea ice mass variable) is a reasonably small residual on centennial time scales, which was confirmed. This allows the ice transport out of the Arctic to serve as a proxy for the unknown freezing and melting of sea ice, which yields good correlation in the integrated net freshwater flux to halosteric SSH, as shown below. To close the balance, the liquid transport of freshwater also has to be considered.
Removing the respective mean values to highlight the variability, the time series reveal the individual contributions to mean halosteric SSH in the Arctic as a function of time after integrating the various annual freshwater flux values to obtain freshwater volume changes (Fig. 6). There is no single source of freshwater which dominates as a driver for the majority of centennial-scale variability of Arctic SSH for this model. Instead, all three sources contribute to the total FWC with similar magnitudes, albeit with different phases depending on the time period in question. As an example, early in the record (ca. years 850–1250), the freshwater flux from WFO (yellow line) and the ice component (green line) contribute most to the total. In contrast, the contribution from the oceanic transport (red line) continues to decrease the FWC via the increase of FW transport out of the Arctic from year 1050 through to almost 1150, while the other two inputs increase FWC anomalies, and is thus more out of phase with the other components and FWC (the sum) than prior to 1050. However, during the last 350 years (approximately from 1500 to the end of the record), the oceanic sources of freshwater are more in phase throughout this period with the total FWC, and the freshwater flux from WFO and ice components exhibit long delays causing them to be out of phase with the overall FWC after about 1720. When adding the three components of freshwater flux together (WFO, ocean transport, and ice transport) and integrating, the resultant FWC estimate from the fluxes (dashed black line) correlates with halosteric SSH at 0.97 (90% CL: [0.94, 0.98]). The peaks of the FWC estimate (dashed black line) mismatch at times with halosteric SSH (blue line), which is likely due to the freshwater storage in sea ice having larger variability on shorter time scales, in which the net ice transport does not fully reflect shorter-term freezing and melting.
Sources of freshwater content changes in the Arctic Ocean, based on annual means, and then cumulatively integrated and divided by the area of the Arctic Ocean to provide units of centimeters of freshwater. The means were removed in order to highlight the variability of the sources in the balance, which are plotted along with the Arctic integrated halosteric SSH changes in cm, which is scaled by a factor of 36, close to the Arctic mean ratio of freshwater content to halosteric SSH. Positive values represent anomalously high freshwater entering the Arctic (WFO) or anomalously low freshwater leaving the Arctic (net ocean and ice transports). Different sources of freshwater contribute to the majority of halosteric changes at different times. The time series of halosteric SSH is an 11-yr running mean. The net freshwater plotted here is the integration of the freshwater flux total, as calculated in Eq. (4).
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
b. Forced run versus control run data
To investigate whether freshwater content variability is similar between the control run and the forced run, we broke down the FWC into the same three source components for the piControl run of the past1000 CMIP5 experiment, and additionally for the piControl run for MPI-ESM-LR, which acted as the base control run for historical and future projections in CMIP5. The characteristics of the two control run results did not differ from each other substantially, but some variation in low-frequency variability was evident (not shown), which is likely just from undersampling of the low-frequency variability in this model. So, we focus our comparison of the forced past1000 results mostly on the MPI-ESM-P piControl data in the following analyses.
The variability in the freshwater source time series is compared by removing these means and integrating the source components in time (Fig. 7). One can see that the oceanic freshwater transport (Fig. 7b) is pretty similar between forced and control runs in terms of the size of the variability, but the external freshwater signal (contained in the CMIP5 variable WFO) is significantly larger in the forced run (Fig. 7a), with the ice transport signal also somewhat larger (Fig. 7c). Relative to control runs, the WFO is the largest contributor to the enhanced FWC and halosteric SSH variability (Fig. 7f), with ice transport contributing to a lesser degree.
(a)–(c) The freshwater content changes (in cm) due to the three main sources of freshwater flux into the Arctic, where the variability of the forced run (blue) is compared to the Ctl-P run (red). The standard deviation of each property is shown in parentheses in the legends, with the ratio of the standard deviation of the past1000 run to the Ctl-P run listed as the “ratio” at the bottom. (d) Precipitation is another indication of the low-frequency variability in the atmosphere, and tracks with past1000 WFO in (a). Note that (e) the sea ice-cover fraction is smoothed by a 111-yr running mean, and (f) the halosteric SSH is smoothed by an 11-yr running mean for greater visibility of the low-frequency variability. The other four panels show annual anomalies relative to their means.
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
The external freshwater input could be driven by either increased precipitation over the Arctic (which includes runoff), changes in the ice cover in the Arctic (controlling how much precipitation lands directly in the ocean), or both. The analysis of the previous coarser-resolution version, from the Millennium project (Jungclaus et al. 2010), showed global mean reduction in air temperature of more than 0.5°C and up to 0.1 mm day−1 precipitation reduction over a period lasting for around 4 years following eruption (Zanchettin et al. 2014). The associated pattern was shown to follow the projections under global warming with the opposite sign—basically a dry-gets-wetter, wet-gets-drier rule (Iles et al. 2013). Figures 7d and 7e suggest that it is both precipitation and ice-cover changes, as the precipitation variability poleward of 60°N is much larger in the forced run, but so is the ice-cover variability.
The ice transport variability is slightly more in phase with the WFO variability in the forced run than in the control run (comparing Figs. 7a–c). In the forced run, the correlation between WFO and the ice transport time series is 0.47, whereas for Ctl-P it is −0.2. Due to the high autocorrelation of WFO in both forced and Ctl-P runs, these correlations are strictly speaking not statistically significant from each other but, as a measure of coherence, it points to a substantial difference in how much more they can contribute jointly to total freshwater variability in the forced run than they do in the control run. The drivers for the increase in WFO and ice transport variability individually as well as this enhanced coherence between them in the forced run is examined in the next section.
A detail worthy of mention regarding the curves in Fig. 7: the freshwater changes in Figs. 7a–c were calculated by cumulative integration of annual anomalies. This analysis method suppresses interannual variability. For our results here, this method is fine as we can focus on the lower-frequency multidecadal and centennial variability of these properties. To diagnose shorter-frequency interannual variability, other analysis methods would be required.
c. External drivers of Arctic freshwater variability
Previously Zhong et al. (2011) suggested that a long-lasting effect of volcanic eruptions on subpolar North Atlantic SSH exists, despite the short residence time of the associated aerosol release into the atmosphere, primarily through the expansion of Arctic sea ice exported into the North Atlantic, which suppresses convective warming of subpolar surface waters. The associated positive feedback causes the long-lasting effect and thereby centennial-scale impact. However, it is difficult to completely isolate the drivers of climate response (solar versus volcanic), and there can be differences in responses between various climate models. For example, Shindell et al. (2003) argued that, for the Medieval Warm Period to the Little Ice Age, solar forcing alone already generates much larger long-term responses in comparison to unforced simulations, in the GISS (Goddard Institute for Space Studies) global climate model.
Correlations of the freshwater component time series with atmospheric properties can suggest possible mechanisms which drive these low-frequency FWC changes, and drives them differently between forced and control runs (Fig. 8). The correlations are evaluated between the 111-yr running-mean time series in order to focus particularly on the causes of the centennial-scale variability. Results show that WFO seems to be related to an Arctic Oscillation–like (AO-like) pattern, although it is broken in the Pacific Sector (Fig. 8a), which drives periods of enhanced and reduced precipitation. The control-run WFO data also show an atmospheric connection to an AO-like pattern (Fig. 8b). However, the WFO in the forced run is strongly correlated to a nearly hemispheric surface air temperature (SAT) signal (Fig. 8c), which can occur during very large volcanic eruptions (Zanchettin et al. 2013), whereas the control-run WFO is not nearly as tightly controlled by external SAT forcing (Fig. 8d). Stenchikov et al. (2009) showed that volcanic cooling tends to enhance the positive phase of the AO and strengthens the AMOC, which is also found by Slawinska and Robock (2018). The impact of the opposite signal of a reduced AMOC can be seen in Fig. 8c, which suggests that a pronounced warming hole in the subpolar North Atlantic appears in the context of overall positive temperature anomalies. The warming hole is characteristic for an AMOC decline (Drijfhout et al. 2012). However, Zhong et al. (2011) also found weakening of the AMOC under volcanic forcing conditions. These other results strengthen the notion that, in this model, volcanic forcing is an important driver on multidecadal time scales.
(a) Point-correlation map of the sea level pressure (SLP) field with Arctic integrated time series of external freshwater from the atmosphere and runoff (WFO) for the forced run. (b) As in (a), but from the MPI-ESM-P control run data. (c) Point-correlation map of the surface air temperature (SAT) field with the forced-run WFO time series. (d) As in (c), but from the Ctl-P run. (e) The SAT field correlated with the ice transport time series out of the Arctic. (f) As in (e), but from the Ctl-P run.
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
The ice transport time series through the Arctic boundary from the model output ice transport variables is shown in blue, with positive into the Arctic. The red line is the ice transport as modeled by a single averaged ice transport velocity value for the whole boundary multiplied by the effective ice thickness adjacent to the boundary for each inner (Arctic-side) boundary-bordering grid box, with the grid box width also being applied as a factor. The ice transport variability is almost wholly due to the variability of effective ice thickness leaving the Arctic, and not so dependent on wind-driven changes in outflow speed. Time series are 111-yr running means.
Citation: Journal of Climate 35, 13; 10.1175/JCLI-D-21-0484.1
These two results in particular seem to indicate that the driving feature for the forced-run ice transport is mostly just varying air temperature plus precipitation, being somewhat correlated with WFO, and not as dependent on ocean surface velocities or prevailing winds beyond the mean value, on these centennial time scales.
4. Discussion and concluding remarks
Centennial-scale variability is evident in both the unforced control runs and forced past1000 experiment in the MPI-ESM-P climate model. The forcing of the variability generally is driven by the three main sources of freshwater Arctic variability: external freshwater forcing (precipitation into the ocean plus runoff), oceanic freshwater transport, and sea ice transport. The main difference in variability between the forced run and control run is due to external input, which is larger in the forced run, leading to larger centennial-scale variability in the forced run. This enhanced external freshwater source appears to be largely driven by the presence of large volcanic forcing in the past1000 experiment.
Although the centennial-scale variability of Arctic freshwater content, and relatedly Arctic sea level, is not large enough to provide a particularly strong influence on future climate projections in the region, it is large enough to be aware of and to take into account. Indeed, estimates of this variability will improve with continued observation over longer time intervals. For instance, the centennial-scale SSH variability described here can be as high as 4 cm in standard deviation (Fig. 2, top panel), showing peak to peak variability of 10 cm (Fig. 4, lower left), whereas the RCP4.5 scenario Arctic sea level change is more than 30 cm in amplitude over much of the region, with changes over 50 cm throughout the Barents Sea region in the CMIP5 ensemble described in the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (Figs. 13 and 20b in Church et al. 2013). We find that the larger amplitude in forced-run variability compared to the control-run variability is caused by precipitation and freshwater storage in ice rather than oceanic freshwater exchange with the Atlantic. This might be indicative of the fact that Arctic freshwater changes can be larger than expected if any large climate forcing—larger than those maintaining the mean internal climate variability—trigger additional responses of the climate system components that can affect freshwater transport.
For the current example, when large volcanic eruptions cause widespread cooling, the response seems to include a cooling-induced reduction in Arctic precipitation and the associated decrease in freshwater content and SSH. Noticeable large-scale changes in the climate system on multidecadal time scales due to volcanism were also found in the North Atlantic by Mann et al. (2021), who argue that the Atlantic multidecadal oscillation is solely due to volcanic forcing, also using CMIP5 past1000 experiments. Contrary to their conclusions, we find small, but nonzero Arctic low-frequency variability also in the Ctl-P control, i.e., unforced, run. Similar low-frequency variability can also be found in the other configurations of the MPI-ESM-LR model (the “piControl run” from that model’s modern and future projections, here “Ctl-LR”) used in Carson et al. (2015) and the AR5 report, though it is smaller than in the Ctl-P run.
The sources of centennial freshwater variability are interwoven and complex. Ocean transport values in this study are only one portion of the centennial-scale variability in both forced and control runs, with the formation and export of ice plus precipitation changes being of similar magnitude (Fig. 7). The interplay between these freshwater sources could be further explored in other models, perhaps better in the CMIP6 where there is a new generation of coupling schema between these processes.
In a future warming scenario, one expectation from models is an increase in freshwater into the Arctic (Jahn and Laiho 2020), which may also be observable already (Woodgate and Peralta-Ferriz 2021). Even so, the extent to which this is true in the real climate at these centennial time scales remains to be seen. Further exploration of substantial Arctic centennial-scale sea level variability could examine the robustness of results across a variety of different model parameterizations and sensitivities, with the new set of CMIP6 model experiments providing higher resolutions globally.
Acknowledgments.
Funded in part through the RACE-Synthese project, supported by the German BMBF (Grant 03F0825B), and the DFG funded SPP 1889: Regional Sea Level Change and Society (Grant STA 410/32-2). Contribution to the Center für Erdsystemforschung und Nachhaltigkeit (CEN) of Universität Hamburg.
Data availability statement
Model data used in this study can be found in most CMIP5 archives, including the WDCC archive at DKRZ: https://cera-www.dkrz.de/. Scripts and instructions for reproducing the paper’s figures from the raw model data accessible in web repositories are available in a package at https://github.com/evanescentstar/Arctic_century_SSH.
REFERENCES
Antonov, J. I., S. Levitus, and T. P. Boyer, 2002: Steric sea level variations during 1957–1994: Importance of salinity. J. Geophys. Res., 107, 8013, https://doi.org/10.1029/2001JC000964.
Bothe, O., J. H. Jungclaus, and D. Zanchettin, 2013: Consistency of the multi-model CMIP5/PMIP3-past1000 ensemble. Climate Past, 9, 2471–2487, https://doi.org/10.5194/cp-9-2471-2013.
Braconnot, P., S. P. Harrison, B. Otto-Bliesner, A. Abe-Ouchi, J. Jungclaus, and J.-Y. Peterschmitt, 2011: The paleoclimate modeling intercomparison project contribution to CMIP5. CLIVAR Exchanges, No. 56, International CLIVAR Project Office, Southampton, United Kingdom, 15–19.
Carson, M., A. Köhl, and D. Stammer, 2015: The impact of regional multidecadal and century-scale internal climate variability on sea level trends in CMIP5 models. J. Climate, 28, 853–861, https://doi.org/10.1175/JCLI-D-14-00359.1.
Church, J. A., and Coauthors, 2013: Sea level change. Climate Change 2013: The Physical Science Basis. T. F. Stocker, Eds., Cambridge University Press, 1137–1216.
Couldrey, M. P., and Coauthors, 2021: What causes the spread of model projections of ocean dynamic sea-level change in response to greenhouse gas forcing? Climate Dyn., 56, 155–187, https://doi.org/10.1007/s00382-020-05471-4.
Domeisen, D. I. V., A. H. Butler, K. Fröhlich, M. Bittner, W. A. Mülle, and J. Baehr, 2015: Seasonal predictability over Europe arising from El Niño and stratospheric variability in the MPI-ESM seasonal prediction system. J. Climate, 28, 256–27, https://doi.org/10.1175/JCLI-D-14-00207.1.
Drijfhout, S., G. J. Van Oldenborgh, and A. Cimatoribus, 2012: Is a decline of AMOC causing the warming hole above the North Atlantic in observed and modeled warming patterns? J. Climate, 25, 8373–8379, https://doi.org/10.1175/JCLI-D-12-00490.1.
Godfrey, J. S., 1989: A Sverdrup model of depth-integrated flow for the world ocean allowing for island circulations. Deep-Sea Res., 45, 89–112, https://doi.org/10.1080/03091928908208894.
Gregory, J. M., and Coauthors, 2016: The flux-anomaly-forced model intercomparison project (FAFMIP) contribution to CMIP6: Investigation of sea-level and ocean climate change in response to CO2 forcing. Geosci. Model Dev., 9, 3993–4017, https://doi.org/10.5194/gmd-9-3993-2016.
Gu, G., and R. F. Adler, 2011: Precipitation and temperature variations on the interannual time scale: Assessing the impact of ENSO and volcanic eruptions. J. Climate, 24, 2258–2270, https://doi.org/10.1175/2010JCLI3727.1.
Haak, H., J. Jungclaus, U. Mikolajewicz, and M. Latif, 2003: Formation and propagation of great salinity anomalies. Geophys. Res. Lett., 30, 9, https://doi.org/10.1029/2003GL017065.
Haine, T. W., and Coauthors, 2015: Arctic freshwater export: Status, mechanisms, and prospects. Global Planet. Change, 125, 13–35, https://doi.org/10.1016/j.gloplacha.2014.11.013.
Hartmann, D. L., and Coauthors, 2013: Observations: Atmosphere and surface. Climate Change 2013: The Physical Science Basis. T. F. Stocker, Eds., Cambridge University Press, 159–254.
Holgate, S. J., and Coauthors, 2013: New data systems and products at the permanent service for mean sea level. J. Coastal Res., 29, 493–504, https://doi.org/10.2112/JCOASTRES-D-12-00175.1.
Iles, C. E., G. C. Hegerl, A. P. Schurer, and X. Zhang, 2013: The effect of volcanic eruptions on global precipitation. J. Geophys. Res. Atmos., 118, 8770–8786, https://doi.org/10.1002/jgrd.50678.
Jahn, A., and R. Laiho, 2020: Forced changes in the Arctic freshwater budget emerge in the early 21st century. Geophys. Res. Lett., 47, e2020GL088854, https://doi.org/10.1029/2020GL088854.
Johnson, H. L., S. B. Cornish, Y. Kostov, E. Beer, and C. Lique, 2018: Arctic Ocean freshwater content and its decadal memory of sea-level pressure. Geophys. Res. Lett., 45, 4991–5001, https://doi.org/10.1029/2017GL076870.
Joyce, T. M., and A. Proshutinsky, 2007: Greenland’s Island rule and the Arctic Ocean circulation. J. Mar. Res., 65, 639–653, https://doi.org/10.1357/002224007783649439.
Jungclaus, J. H., and Coauthors, 2010: Climate and carbon-cycle variability over the last millennium. Climate Past, 6, 723–737, https://doi.org/10.5194/cp-6-723-2010.
Jungclaus, J. H., and Coauthors, 2013: Characteristics of the ocean simulations in the Max Planck Institute Ocean Model (MPIOM) the ocean component of the MPI-Earth system model. J. Adv. Model. Earth Syst., 5, 422–446, https://doi.org/10.1002/jame.20023.
Koenigk, T., U. Mikolajewicz, H. Haak, and J. Jungclaus, 2007: Arctic freshwater export in the 20th and 21st centuries. J. Geophys. Res., 112, G04S41, https://doi.org/10.1029/2006JG000274.
Köhl, A., and N. Serra, 2014: Causes of decadal changes of the freshwater content in the Arctic Ocean. J. Climate, 27, 3461–3475, https://doi.org/10.1175/JCLI-D-13-00389.1.
Koul, V., C. Schrum, A. Düsterhus, and J. Baehr, 2019: Atlantic inflow to the North Sea modulated by the subpolar gyre in a historical simulation with MPI-ESM. J. Geophys. Res. Oceans, 124, 1807–1826, https://doi.org/10.1029/2018JC014738.
Landerer, F. W., J. H. Jungclaus, and J. Marotzke, 2007: Regional dynamic and steric sea level change in response to the IPCC-A1B scenario. J. Phys. Oceanogr., 37, 296–312, https://doi.org/10.1175/JPO3013.1.
Mann, M. E., B. A. Steinman, D. J. Brouillette, and S. K. Miller, 2021: Multidecadal climate oscillations during the past millennium driven by volcanic forcing. Science, 371, 1014–1019, https://doi.org/10.1126/science.abc5810.
Otterå, O. H., M. Bentsen, H. Drange, and L. Suo, 2010: External forcing as a metronome for Atlantic multidecadal variability. Nat. Geosci., 3, 688–694, https://doi.org/10.1038/ngeo955.
Pardaens, A. K., J. M. Gregory, and J. A. Lowe, 2011: A model study of factors influencing projected changes in regional sea level over the twenty-first century. Climate Dyn., 36, 2015–2033, https://doi.org/10.1007/s00382-009-0738-x.
Peralta-Ferriz, C., and J. Morison, 2010: Understanding the annual cycle of the Arctic Ocean bottom pressure. Geophys. Res. Lett., 37, L10603, https://doi.org/10.1029/2010GL042827.
PSMSL, 2022: Tide gauge data. Permanent Service for Mean Sea Level, accessed 7 February 2022, http://www.psmsl.org/data/obtaining/.
Proshutinsky, A. Y., and M. A. Johnson, 1997: Two circulation regimes of the wind-driven Arctic Ocean. J. Geophys. Res., 102, 12 493–12 514, https://doi.org/10.1029/97JC00738.
Proshutinsky, A., R. H. Bourke, and F. A. McLaughlin, 2002: The role of the Beaufort Gyre in Arctic climate variability: Seasonal to decadal climate scales. Geophys. Res. Lett., 29, 2100, https://doi.org/10.1029/2002GL015847.
Schmidt, G. A., and Coauthors, 2011: Climate forcing reconstructions for use in PMIP simulations of the last millennium (v1.0). Geosci. Model Dev., 4, 33–45, https://doi.org/10.5194/gmd-4-33-2011.
Segschneider, J., and Coauthors, 2013: Impact of an extremely large magnitude volcanic eruption on the global climate and carbon cycle estimated from ensemble Earth system model simulations. Biogeosciences, 10, 669–687, https://doi.org/10.5194/bg-10-669-2013.
Shindell, D. T., G. A. Schmidt, R. L. Miller, and M. E. Mann, 2003: Volcanic and solar forcing of climate change during the preindustrial era. J. Climate, 16, 4094–4107, https://doi.org/10.1175/1520-0442(2003)016<4094:VASFOC>2.0.CO;2.
Slawinska, J., and A. Robock, 2018: Impact of volcanic eruptions on decadal to centennial fluctuations of Arctic Sea ice extent during the last millennium and on initiation of the little Ice Age. J. Climate, 31, 2145–2167, https://doi.org/10.1175/JCLI-D-16-0498.1.
Stenchikov, G., T. L. Delworth, V. Ramaswamy, R. J. Stouffer, A. Wittenberg, and F. Zeng, 2009: Volcanic signals in oceans. J. Geophys. Res., 114, D16104, https://doi.org/10.1029/2008JD011673.
Stjern, C. W., and Coauthors, 2019: Arctic amplification response to individual climate drivers. J. Geophys. Res. Atmos., 124, 6698–6717, https://doi.org/10.1029/2018JD029726.
Volkov, D. L., and F. W. Landerer, 2013: Nonseasonal fluctuations of the Arctic Ocean mass observed by the GRACE satellites. J. Geophys. Res. Oceans, 118, 6451–6460, https://doi.org/10.1002/2013JC009341.
Woodgate, R. A., and C. Peralta-Ferriz, 2021: Warming and freshening of the Pacific inflow to the Arctic from 1990-2019 implying dramatic shoaling in Pacific Winter Water ventilation of the Arctic water column. Geophys. Res. Lett., 48, e2021GL092528, https://doi.org/10.1029/2021GL092528.
Xiao, K., M. Chen, Q. Wang, X. Wang, and W. Zhang, 2020: Low-frequency sea level variability and impact of recent sea ice decline on the sea level trend in the Arctic Ocean from a high-resolution simulation. Ocean Dyn., 70, 787–802, https://doi.org/10.1007/s10236-020-01373-5.
Zanchettin, D., C. Timmreck, H.-F. Graf, A. Rubino, S. Lorenz, K. Lohmann, K. Krueger, and J. H. Jungclaus, 2012: Bi-decadal variability excited in the coupled ocean–atmosphere system by strong tropical volcanic eruptions. Climate Dyn., 39, 419–444, https://doi.org/10.1007/s00382-011-1167-1.
Zanchettin, D., O. Bothe, H. F. Graf, S. J. Lorenz, J. Luterbacher, C. Timmreck, and J. H. Jungclaus, 2013: Background conditions influence the decadal climate response to strong volcanic eruptions. J. Geophys. Res. Atmos., 118, 4090–4106, https://doi.org/10.1002/jgrd.50229.
Zanchettin, D., and Coauthors, 2014: Inter-hemispheric asymmetry in the sea-ice response to volcanic forcing simulated by MPI-ESM (COSMOS-Mill). Earth Syst. Dyn., 5, 223–242, https://doi.org/10.5194/esd-5-223-2014.
Zhang, X., S. Ojha, A. Köhl, H. Haak, J. H. Jungclaus, and D. Stammer, 2022: Sea level changes mechanisms in the MPI-ESM under FAFMIP forcing conditions. Climate Dyn., https://doi.org/10.1007/s00382-022-06231-2, in press.
Zhong, Y., G. H. Miller, B. L. Otto-Bliesner, M. M. Holland, D. A. Bailey, D. P. Schneider, and A. Geirsdottir, 2011: Centennial-scale climate change from decadally-paced explosive volcanism: A coupled sea ice-ocean mechanism. Climate Dyn., 37, 2373–2387, https://doi.org/10.1007/s00382-010-0967-z.
