CORE-II Virtual Special IssueAn assessment of global and regional sea level for years 1993–2007 in a suite of interannual CORE-II simulations
Introduction
There are growing observation-based measures of large-scale patterns of sea level variations with the advent of the Argo floats (since the early 2000s) and satellite altimeters (since 1993). Such measures provide a valuable means to evaluate aspects of global model simulations, such as the global ocean-sea ice simulations run as part of the interannual Coordinated Ocean-sea ice Reference Experiments (Griffies et al., 2009b, Danabasoglu et al., 2014). In this paper, we present an assessment of such CORE-II simulations from 13 model configurations, with a focus on their ability to capture observation-based trends in ocean heat content as well as steric, thermosteric and halosteric sea level.
Our assessment focuses on the final 15 year period (1993–2007) of the CORE-II simulations to enable direct comparison of the simulations to both in situ and satellite based analyses. During this relatively short period, sea level variations have a large component due to natural variability (Zhang and Church, 2012, Meyssignac et al., 2012). This situation is compatible with the CORE-II simulations, as they are primarily designed for studies of interannual variability (Doney et al., 2007, Large and Yeager, 2012). Focusing our assessment on these years also ensures that the Large and Yeager (2009) atmospheric state, used as part of the CORE-II air-sea flux calculations, contains interannual satellite-based radiation, which is available only after 1983.
The practical basis for our study is a suite of global ocean-sea ice models forced with 60 years of the interannual CORE-II atmospheric state from Large and Yeager (2009), with this atmosphere state repeated five times for a total of 300 years. Details of the protocol can be found in Griffies et al. (2009b), which focused on the use of a repeating annual cycle; i.e., the Normal Year Forcing of the CORE-I project. Further details specific to the interannual CORE-II protocol are provided in the Atlantic study by Danabasoglu et al. (2014), with that study also providing many details of the models forming the suite of CORE-II simulations analyzed here.
Sea level change due to human-induced climate change has the potential to affect coastal regions over the remainder of the 21st century and for centuries thereafter. From among the many physical processes impacting sea level, it is the evolution of land ice sheets on Greenland and Antarctica that offers the greatest degree of uncertainty and broadest potential for significant impact. For example, the growth and decay of ice sheets have caused sea level change on the order of 100 m over the recent 450 thousand years with fluctuations of about 100 thousand years (Lambeck et al., 2002, Rohling et al., 2009). We ignore here such sea level changes associated with melting land ice (except to the extent that such water fluxes are contained in the CORE-II river runoff data based on Dai et al., 2009). There are complementary global ocean-sea ice studies that consider the ocean’s response to melt events, such as those from Gerdes et al., 2006, Stammer, 2008, Weijer et al., 2012, Lorbacher et al., 2012.
Ocean warming causes ocean volume to increase due to a decrease in density. As estimated by Church et al., 2011, Gregory et al., 2013, such changes in global mean thermosteric sea level determine about one-third to one-half of the observed global mean sea level rise during the late 20th and early 21st centuries, with changes in ocean mass contributing the remainder. Although limited largely to examinations of natural variability over the relatively short period of 1993–2007, our assessment is of some use to determine the suitability of global ocean-sea ice models for capturing longer term observed trends largely due to anthropogenic effects, such as those considered in Levitus et al., 2005, Boyer et al., 2005, Domingues et al., 2008, Ishii and Kimoto, 2009, Hosoda et al., 2009, Durack and Wijffels, 2010, Church et al., 2011, Gleckler et al., 2012, Levitus et al., 2012. In particular, we can assess the ability of forced global ocean-sea ice models to represent observed changes in patterns of ocean heat content and thermosteric sea level change (Lombard et al., 2009, Kuhlbrodt and Gregory, 2012). Furthermore, we note the importance of ocean warming on ice shelf melt (e.g., Yin et al., 2011), with this connection providing yet another reason that an assessment of how models simulate observed warming provides a useful measure of their skill for making projections.
The following two questions regarding the global mean sea level trends and associated spatial patterns frame our assessment of the CORE-II simulations.
- •
global mean thermosteric sea level: Do CORE-II global ocean-sea ice simulations reproduce the observed global mean sea level variations associated with thermosteric effects estimated from the observation-based analyses? To address this question, we focus on ocean temperature and heat content trends, and how these trends are associated with changes in thermosteric sea level.
- •
patterns of dynamic sea level: Do CORE-II ocean-sea ice simulations reproduce observation-based changes to dynamic sea level patterns? To address this question, we partition dynamic sea level trends into their halosteric and thermosteric patterns, as well as bottom pressure contributions.
Answers to these questions are not simple, nor do we presume our contribution leads to unequivocal results. Nonetheless, we aim to provide physical and mathematical insight in the process of assessing the physical integrity of the CORE-II simulations. An underlying hypothesis of CORE is that global ocean–sea ice models coupled with the same prescribed atmospheric state produce similar simulations (Griffies et al., 2009b, Danabasoglu et al., 2014). We consider this hypothesis in the context of our sea level analysis. We hope that our presentation assists in the ongoing scientific quest to understand observed sea level changes, and to characterize some of its causes as realized in global ocean-sea ice models.
We aim to physically motivate and mathematically detail a suite of methods for sea level studies, providing sufficient information to both understand and reproduce our analyses. In this way, we hope that this paper serves both as a benchmark for how the present suite of CORE-II simulations performs in the representation of sea level, and provides a reference from which the reader may understand this, and other, studies of simulated sea level even after the models used here become obsolete.
The remainder of this paper consists of the following sections. We initiate the main text in Section 2 by considering aspects of the sea level question as framed by the CORE-II simulations with global ocean-sea ice models. In particular, we refine the questions posed in Section 1.1 by exposing some of the limitations inherent in the CORE-II experimental design and the atmospheric state used to drive the models. Our analysis of the global mean sea level from the CORE-II simulations is then presented in Section 3. It is here that we focus on the first question posed above concerning how well the CORE-II simulations represent the global thermosteric rise in sea level as compared to observation-based estimates. We follow in Section 4 with a discussion of the ocean heating trends over the years 1993–2007, with comparison to estimated observation-based trends. In Section 5 we then present the regional patterns of sea level (second question raised above), partitioning sea level trends into thermosteric, halosteric, and bottom pressure trends. We complete the main text with a summary and discussion in Section 6.
We provide a selection of support material in the appendices. Some of this material is rudimentary, yet it is central to the theoretical and practical foundation of this paper. Appendix A focuses on the global mean sea level question as posed in ocean-sea ice climate models, which can be addressed through kinematic considerations. Appendix B presents dynamical notions of use to interpret patterns of sea level, in particular the partitioning of sea level tendencies into thermosteric, halosteric, and bottom pressure tendencies. Appendix C examines the ability of ocean models to conserve heat throughout the ocean fluid.
This paper contains a wealth of information in its many multi-panelled figures. However, we do not fully discuss each detail in the figures, as doing so requires a tremendous amount of discussion making a long paper even longer. We suggest that many readers may find it sufficient to focus on the CORE-II ensemble means that are provided for most of the figures, with our discussion often focusing on the ensemble mean.
Furthermore, our presentation is descriptive in nature, as framed within the physically based analysis methodology detailed in the appendices. There is, however, little insight offered for the underlying physical mechanisms that explain model–model or model–observational differences. For example, we do not try to associate a particular model behaviour with the choice of physical parameterization. Such work is beyond our scope, with the present analysis intent on helping to identify areas where process-based studies may be warranted to isolate mechanisms accounting for differences.
Some readers may be disappointed with our reticence to penetrate deeper into such mechanisms. We too are disappointed. However, we are limited in how much we can answer such questions based on available diagnostic output from the simulations. Nonetheless, this excuse, which is in fact ubiquitous in such comparison papers utilizing CORE or CMIP (Coupled Model Intercomparison Project) simulations, is unsatisfying. The logistics of coordinating a comparison become increasingly complex when aiming to compare detailed diagnostics, such as budget terms, in a consistent manner. Yet more should be done to mechanistically unravel model-model differences. We provide further comment in Section 6.6 regarding this point. We argue there that progress on this issue is possible, with one means requiring a physical process-based analysis of the heat, salt, and buoyancy budgets.
Section snippets
Sea level in CORE-II simulations
We frame here the sea level question for the CORE-II simulations. Of interest are salient ocean model fundamentals and limitations, and aspects of the CORE-II experimental design.
Steric impacts on global mean sea level
The CORE protocol (Griffies et al., 2009b, Danabasoglu et al., 2014) introduces a negligible change to the liquid ocean mass (non-Boussinesq) or volume (Boussinesq), and the salt remains nearly constant (except for relatively small exchanges associated with sea ice changes). For simulations with zero net water crossing the ocean surface and constant salt content, changes to the simulated global mean sea level arise predominantly through the global mean of thermosteric effects. That is, global
Temperature and heat content trends for 1993–2007
Global sea level change in the CORE-II simulations is directly correlated to the change in ocean heat content, with the global mean temperature shown in Fig. 3 directly related to the net heat flux entering the ocean through its boundaries (Eq. (40) in Appendix A.4). We thus find it useful to consider the heat fluxes and ocean heat content and temperature trends seen in the CORE-II simulations. Following the discussion in Sections 2.6 Global mean SST in the CORE-II simulations, 2.7 Restricting
Dynamic sea level during 1993–2007
In Fig. 15, we present the time mean of the dynamic sea level (Eq. (5)) over years 1993–2007 for the CORE-II simulations, as well as the dynamic sea level from the gridded satellite altimeter product from the AVISO project (Archiving, Validation, and Interpolation of Satellite Oceanographic) (Le Traon et al., 1998, Ducet et al., 2000). Recall from the definition in Eq. (5), the DSL has a zero global area mean. Fig. 16 shows the anomalies (model minus satellite), with model results mapped to the
Summary and discussion
Sea level emerges from mechanical and thermodynamic forcing on the ocean boundaries, and is affected by transport and mixing in the ocean interior. Thus, all physical processes impacting the ocean impact sea level, including physical oceanographic processes as well as geophysical processes associated with changes in the earth’s gravity and rotation. Sea level is a key field to accurately capture in simulations to assess the potential for climate impacts, particularly in coastal regions.
Acknowledgement
The WCRP/CLIVAR Working Group on Ocean Model Development (WGOMD) is responsible for organizing the Coordinated Ocean-sea ice Reference Experiments. The conceptual and technical details associated with global ocean-sea ice model comparisons have comprised the majority of the group’s deliberations since its inception in 1999. We thank Anna Pirani from the CLIVAR staff for her tireless and gracious support of WGOMD activities.
Much of the analysis in this paper made use of the free software package
References (145)
- et al.
Sea level variations at tropical Pacific islands since 1950
Global Planet. Change
(2012) - et al.
An ERA40-based atmospheric forcing for global ocean circulation models
Ocean Modell.
(2010) The axial angular momentum balance of a global ocean general circulation model
Dyn. Atmos. Oceans
(1997)- et al.
Basin patterns of global sea level changes for 2004–2007
J. Marine Syst.
(2010) A comparison of global ocean general circulation model solutions obtained with synchronous and accelerated integration methods
Ocean Modell.
(2004)- et al.
North Atlantic simulations in Coordinated Ocean-ice Reference Experiments phase II (CORE-II). Part I: Mean states
Ocean Modell.
(2014) - et al.
Lateral transport in the ocean interior
- et al.
Sensitivity of a global ocean model to increased run-off from Greenland
Ocean Modell.
(2006) - et al.
The theory of the seasonal variability in the ocean
Deep-Sea Res.
(1973) - et al.
Coordinated Ocean-ice Reference Experiments (COREs)
Ocean Modell.
(2009)