Suzuki et al. (2005)

Title:: Projection of future sea level and its variability in a high-resolution climate model: Ocean processes and Greenland and Antarctic ice-melt contributions
Corresponding author:: Tatsuo Suzuki
Citation:: Suzuki, T., Hasumi, H., Sakamoto, T. T., Nishimura, T., Abe‐Ouchi, A., Segawa, T., et al. (2005). Projection of future sea level and its variability in a high‐resolution climate model: Ocean processes and Greenland and Antarctic ice‐melt contributions. Geophysical Research Letters, 32(19), n/a-n/a. https://doi.org/10.1029/2005gl023677
URL:: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2005GL023677

Abstract

Using a high-resolution climate model, we projected future sea level and its variability based on two scenarios for 21st century greenhouse gas emission. The globally averaged sea level rise attributable to the steric contribution was 23 and 30 cm for the two scenarios. The results of the high-resolution model and a medium-resolution version of the same model for global and local sea level change agreed well. However, the high-resolution model represented more detailed ocean structure changes under global warming. The changes affected not only the spatial distribution of sea level rise, but also the changes in local sea level variability associated with ocean eddies. The enhanced eddy activity was responsible for extreme sea level events.

Introduction

Sea level change is an important aspect of future climate change for human societies and the environment. Estimates of the rate of globally averaged sea level change during the 20th century are in the range of 1 to 2 mm yr−1 [Church et al., 2001]. This globally averaged rise in sea level is mainly the result of the thermal expansion of seawater and land-ice melt. Future projections of sea level change have been calculated using various coupled atmosphere–ocean general circulation models (CGCMs [e.g., Gregory and Lowe, 2000; Gregory et al., 2001; Meehl et al., 2005]). Gregory et al. [2001] compared the results of the projections that followed the IS92a scenario for greenhouse gas (GHG) emissions. The rates of globally averaged sea level rise predicted by the models for the 21st century were in the range of 2.0–3.7 mm yr−1, but regional sea level change was not spatially uniform and some regions experienced more than twice the global average rate of rise. However, different models, while sharing some features, predicted different distributions. The geographical distribution of sea level change is principally determined by changes in density structure and wind stress forcing, both of which affect ocean circulation [Church et al., 2001]. Reproducing ocean structures is important for estimating the distribution of future sea level changes. However, the details of ocean structures, such as western boundary currents and fronts with pronounced horizontal gradients of water properties, have not been reproduced by existing coarse-resolution CGCMs. In this study, we compared sea level projections made by CGCMs with differing resolutions, focusing particularly on detailed features of the spatial distribution of sea level changes as predicted by a high-resolution model.

We also present global and local changes in 21st century sea level variability. Extreme changes in sea levels could severely affect human activities, and storm surges and coastal wave height have been identified in previous studies as sources of extreme sea levels [Church et al., 2001]. Unfortunately, the resolution of CGCMs has not been adequate to resolve these events. Extreme sea levels associated with ocean eddies, however, are represented by high-resolution models. Warm eddies increase the flooding risk in coastal areas. For example, Okinawa Island flooded on 22 July 2001 without the passage of an atmospheric low; rather, a warm eddy was responsible for increasing the sea level by more than 15 cm [Tokeshi and Yanagi, 2003]. Therefore, it would be advantageous to project changes in sea level variability using a high-resolution climate model that included eddies.

Models and Experiments

The CGCM used in this study was the Model for Interdisciplinary Research on Climate, version 3.2 (MIROC3.2), developed at the Center for Climate System Research, University of Tokyo, National Institute for Environmental Studies and Frontier Research Center for Global Change [K-1 Model Developers, 2004]. The ocean component explicitly represented sea surface elevation. The higher-resolution version of MIROC3.2 (MIROC3.2_hi) consisted of a T106 global atmospheric spectral model with 56 vertical levels, an eddy-permitting global ocean model in which horizontal resolution was zonally 0.28° and meridionally 0.19° with 48 vertical levels, and other components (i.e., land, river, and sea ice). In the medium-resolution version of MIROC3.2 (MIROC3.2_med), a T42 global atmospheric spectral model with 20 vertical levels was coupled with a medium-resolution OGCM in which horizontal resolution was zonally 1.4° and meridionally 0.56° around the equator, with 44 vertical levels. The same physical parameterizations were applied to these models, but resolution-dependent parameters were adjusted per model.

Both models were integrated by prescribing external conditions, including solar and volcanic forcing, GHG concentrations, aerosol emissions, and land use. MIROC3.2_hi was spun up for 109 years under the fixed external conditions of the year 1900 after coupling the atmosphere and the ocean. MIROC3.2_med was spun up for 560 years under the fixed conditions of the year 1850. For control runs, we continued the integration under the same conditions for 100 years in MIROC3.2_hi and 400 years in MIROC3.2_med. Twentieth century experiments (20C3M) were performed after the spin-up. These runs were forced by the historical external conditions from 1900 to 2000 in MIROC3.2_hi and from 1850 to 2000 in MIROC3.2_med. Future projections for the 21st century were initiated by the final states of the 20C3M runs and performed by prescribing the external conditions according to the Intergovernmental Panel on Climate Change (IPCC) special report for emission scenarios. CO2 concentrations at the end of the 21st century were 720 and 550 ppm under the A1B scenario and the B1 scenario, respectively [Intergovernmental Panel on Climate Change, 2000]. For MIROC3.2_med, three 20th century runs were performed using different initial conditions (1st, 101st, and 201st years for the control run). We used the ensemble mean of the three members in this study. The model integration was restricted by computing resource limitations, especially for the high-resolution model, and the control runs showed some trends. Therefore, we subtracted these trends from the results.

Globally Averaged Sea Level Rise

The steric contribution (thermal expansion and haline contraction) to sea level rise was estimated from the model ocean temperature and salinity. As the Boussinesq approximation was adopted in the ocean model, the globally averaged sea level rise attributable to steric factors was diagnosed indirectly from density changes as the equivalent volume change under mass conservation:

urn:x-wiley:00948276:media:grl20251:grl20251-math-0001

where ΔH is the globally averaged sea level rise, S is the surface area of the ocean, Z is the ocean depth, ρ is the in situ density, and Δρ is its difference from the reference state. Because the haline contribution was small, the globally averaged change in sea level was caused mainly by changes in ocean heat content. To validate the model, the upper ocean heat content was compared with observations [Levitus et al., 2005]. Decadal variations were not in phase, but long-term trends and the changes over the last decade for both models were similar to observations (data not shown). The 21st century steric contribution was projected to be about 30 cm for the A1B scenario and 23 cm for the B1 scenario in MIROC3.2_hi (Figure 1a). This result was similar to that for MIROC3.2_med (Figure 1b) and within the range of estimations of previous CGCMs [Gregory et al., 2001]. Both models had linear trends in the control runs: about 0.15 mm yr−1 in MIROC3.2_hi and about 0.14 mm yr−1 in MIROC3.2_med. These linear trends were subtracted from the projections.

Figure 1: A time series of globally averaged sea level change in (a) MIROC3.2_hi and (b) MIROC3.2_med (ensemble mean). Solid lines indicate the steric contribution. Broken lines represent the Greenland ice sheet and dotted lines indicate the Antarctica ice sheet.

The contributions of ice-sheet melt were estimated using the methods of Wild et al. [2003]. The contributions of the Greenland and Antarctic ice-sheet melts exhibited opposite tendencies in both models, as in previous estimations [Church et al., 2001]. However, the amplitude in MIROC3.2_hi was larger than that in MIROC3.2_med (Figure 1). The difference between the two models appeared to be a result of the differences in projected temperature rises in the Greenland and snowfall increases in the Antarctica. Both of these differences in arctic temperature and snowfalls are strongly related to the difference in the climate sensitivity of the MIROC3.2_hi and MIROC3.2_med, in which the A1B run induced global warming of about 4.0°C in MIROC3.2_hi and 3.4°C in MIROC3.2_med at the end of the 21st century, respectively. This different sensitivity may partly due to the difference in the control SST and sea ice distribution in the two models.

The globally averaged sea level rise in MIROC3.2_hi is similar to that in MIROC3.2_med in spite of the different sensitivity. It is because the total heat flux into the ocean for the 21st century is similar in the both models, though the net heat flux into ocean in MIROC3.2_hi is larger than that in MIROC3.2_med during the early 21st century. The upper ocean in MIROC3.2_hi also warms up more than that in MIROC3.2_med. The reasons for these differences are currently under investigation.

Regional Sea Level Change

The sea level patterns corresponding to major ocean gyres were well represented by both models (data not shown). In particular, narrow western boundary currents, such as the Kuroshio, were reproduced realistically in MIROC3.2_hi [Sakamoto et al., 2005]. The globally averaged sea level rise estimated in the previous section was added to the local sea levels calculated by each model to obtain sea level projections. Mass balance was ensured in a globally averaged sense by this procedure [Greatbatch, 1994]. Some regions experienced substantially higher sea level rise in both models than the 21st century global average (Figure 2).

Figure 2: The changes in mean sea level between 1980 and 2000 (20C3M) and between 2080 and 2100 (A1B) in (a) MIROC3.2_hi and (b) MIROC3.2_med (ensemble mean).

To distinguish sea level changes caused by global warming from background variability, such as decadal variations, we estimated area-weighted spatial standard deviations of the local sea level change with respect to the control climate. Gregory et al. [2001] assumed that sea level changes associated with global warming and background variability were not spatially correlated. Under this assumption, increased spatial standard deviation would be attributable to global warming if changes in background variability were relatively small. The spatial standard deviation increased with time for both models and reached 7–13 cm by the end of the 21st century (Figure 3), indicating that the projected sea level changes were sufficiently significant relative to background variability. There was a conspicuous signal from the last decade of the 20th century in MIROC3.2_hi.

Figure 3: The spatial standard deviation of the decadal mean field of local sea level change with respect to the control climate. Solid lines indicate MIROC3.2_hi and broken lines represent MIROC3.2_med (ensemble mean).

The regions with large sea level changes were more restricted to specific areas and the magnitudes of change were more pronounced in MIROC3.2_hi than in MIROC3.2_med (Figure 2). These results were consistent with the fact that spatial variability in MIROC3.2_hi was larger than in MIROC3.2_med at the end of the 21st century (Figure 3). The sea level variability associated with eddies was also shown in MIROC3.2_hi (Figure 4a), and this spatial distribution was consistent with satellite obsersvations. These changes were closely related to regional sea level changes and were as large as several centimeters for some regions during the 21st century (Figure 4b).

Figure 4: (a) The root-mean-square (rms) of the sea level anomaly from the 3-month running mean for the control run in MIROC3.2_hi. (b) Changes in the rms between 1980 and 2000 (20C3M) and between 2080 and 2100 (A1B) in MIROC3.2_hi.

Both models exhibited a region of large sea level rises in the North Pacific. These sea level changes were also shown in the Hadley Centre coupled atmosphere-ocean general circulation model (HadCM3), which has a horizontal resolution similar to that of MIROC3.2_med [Gregory and Lowe, 2000]. This feature has not been represented in previous coarser-resolution models [Gregory et al., 2001]. With higher resolution, as in MIROC3.2_med and HadCM3, fronts at the western boundary currents and their extensions were more sharply reproduced, so sea level changes associated with their shifting or intensification were better captured. Such features became further differentiated at higher resolution (Figure 2).

There was a reduced sea level rise north of the Kuroshio Current at approximately 150°E and an enhanced sea level rise to the south in MIROC3.2_hi. This sea level change was associated with the acceleration of the Kuroshio caused by changes in wind stress and the consequential spin-up of the Kuroshio recirculation [Sakamoto et al., 2005]. In contrast, the Kuroshio in MIROC3.2_med overshot to the north in comparison with that in MIROC3.2_hi. Therefore, the region of large sea level rises in MIROC3.2_med extended northward relative to that in MIROC3.2_hi. MIROC3.2_hi also exhibited a region of reduced sea level rises in the North Pacific subpolar gyre. We believe that this feature was related to the intensification of the Aleutian Low, which is also considered to be the cause of the Kuroshio acceleration [Sakamoto et al., 2005]. This acceleration was associated with the enhanced eddy activity in the Kuroshio and the Kuroshio extension under global warming (Figure 4b). These features were not represented in MIROC3.2_med, partly because the subpolar gyre was not well represented due to the overshooting of the Kuroshio. Similar acceleration of a western boundary current and enhanced eddy activity were also detected east of Australia.

Another important difference between the two models was found in the western tropical Pacific. In MIROC3.2_hi, there was a reduced sea level rise east of Mindanao Island that spread to the eastern tropical Pacific (Figure 2a). This feature was caused by intensification of the wind-induced Ekman upwelling under global warming. This wind-induced Ekman upwelling in this region was not well resolved in MIROC3.2_med [Suzuki et al., 2005]. The reduced sea level rise was associated with the acceleration of the North Equatorial Current (NEC) and the North Equatorial Counter Current (NECC). The acceleration of these currents and the Subtropical Counter Current increased the meridional gradient of zonal velocity, which was associated with an enhanced eddy (Figure 4b). The region stretched zonally to 120°W. These responses of the zonal flows to global warming will be investigated in future studies.

Both models showed a narrow band of enhanced sea level rise in the Southern Ocean. Under global warming in our models, the westerlies shifted southward and strengthened. These changes in the wind field contributed to the southward shift and the intensification of the circumpolar fronts, which are linked to sea level change. These changes were also connected with a couple of narrow bands of enhanced and reduced eddy activity that stretched east from Argentina to the south of Australia in MIROC3.2_hi.

A dipole pattern of sea level change in the North Atlantic Ocean, i.e., an enhanced rise north of the Gulf Stream extension and a reduced rise to the south, was recognized in both models. Bryan [1996] suggested that this pattern was consistent with weakening of the upper branch of the Atlantic Meridional Overturning Circulation (AMOC). The AMOC was weakened from 14 Sv (1 Sv = 106 m3 s−1) to 9 Sv in MIROC3.2_hi and from 19.5 Sv to 12.5 Sv in MIROC3.2_med during the 21st century.

Discussion and Conclusions

The dynamic effect of sea level pressure was not included in either of the ocean components. The change in spatial standard deviation estimated from model sea level pressure during the 21st century was less than 2 cm. These changes, while not negligible, were small in comparison to the spatial variability caused by ocean structure changes (Figure 3). Therefore, we did not indicate the contribution of sea level pressure in this study.

The strengthening of eddy activity was recognized in a globally averaged sense. The global average of the root-mean-square (rms) increased from 4.8 to 5.1 cm in the A1B scenario and from 4.8 to 5.0 cm in the B1 scenario during the 21st century. These changes were small compared to levels of globally averaged sea level rise. However, enhanced eddy activity was confined to specific areas, and those areas overlapped with the areas of enhanced sea level rise around some coastal regions and islands, suggesting that the frequency of extreme sea levels may increase in those regions during the 21st century.

We have described future sea level changes as projected by MIROC3.2_hi according to the 21st century scenarios for GHG emissions and compared them with the results of MIROC3.2_med. The globally averaged sea level rise during the 21st century predicted by the two models was similar. The distribution of sea level changes in MIROC3.2_hi also resembled that in MIROC3.2_med on a large scale. However, MIROC3.2_hi presented more detailed ocean structure changes under global warming. The changes in the ocean structure affected not only the spatial distribution of sea level rise, but also changes in local sea level variability. Therefore, it is critical to consider changes in sea level variability when assessing the possible effects on human activities.