Ocean heat content variability and change in an ensemble of ocean reanalyses
- 3.2k Downloads
Accurate knowledge of the location and magnitude of ocean heat content (OHC) variability and change is essential for understanding the processes that govern decadal variations in surface temperature, quantifying changes in the planetary energy budget, and developing constraints on the transient climate response to external forcings. We present an overview of the temporal and spatial characteristics of OHC variability and change as represented by an ensemble of dynamical and statistical ocean reanalyses (ORAs). Spatial maps of the 0–300 m layer show large regions of the Pacific and Indian Oceans where the interannual variability of the ensemble mean exceeds ensemble spread, indicating that OHC variations are well-constrained by the available observations over the period 1993–2009. At deeper levels, the ORAs are less well-constrained by observations with the largest differences across the ensemble mostly associated with areas of high eddy kinetic energy, such as the Southern Ocean and boundary current regions. Spatial patterns of OHC change for the period 1997–2009 show good agreement in the upper 300 m and are characterized by a strong dipole pattern in the Pacific Ocean. There is less agreement in the patterns of change at deeper levels, potentially linked to differences in the representation of ocean dynamics, such as water mass formation processes. However, the Atlantic and Southern Oceans are regions in which many ORAs show widespread warming below 700 m over the period 1997–2009. Annual time series of global and hemispheric OHC change for 0–700 m show the largest spread for the data sparse Southern Hemisphere and a number of ORAs seem to be subject to large initialization ‘shock’ over the first few years. In agreement with previous studies, a number of ORAs exhibit enhanced ocean heat uptake below 300 and 700 m during the mid-1990s or early 2000s. The ORA ensemble mean (±1 standard deviation) of rolling 5-year trends in full-depth OHC shows a relatively steady heat uptake of approximately 0.9 ± 0.8 W m−2 (expressed relative to Earth’s surface area) between 1995 and 2002, which reduces to about 0.2 ± 0.6 W m−2 between 2004 and 2006, in qualitative agreement with recent analysis of Earth’s energy imbalance. There is a marked reduction in the ensemble spread of OHC trends below 300 m as the Argo profiling float observations become available in the early 2000s. In general, we suggest that ORAs should be treated with caution when employed to understand past ocean warming trends—especially when considering the deeper ocean where there is little in the way of observational constraints. The current work emphasizes the need to better observe the deep ocean, both for providing observational constraints for future ocean state estimation efforts and also to develop improved models and data assimilation methods.
KeywordsHeat content Temperature Variability Climate change Global warming Energy budget Ocean reanalyses Ocean state estimation Ocean models Inter comparison Ocean Observations Data assimilation
Ocean reanalyses (ORAs) represent an important tool for understanding ocean variability and climate change (Lee et al. 2009) and underpin a number of forecast activities, such as operational oceanography and seasonal-to-decadal prediction (Rienecker et al. 2010). ORAs employ a variety of ocean general circulation models (OGCMs) and data assimilation schemes to synthesize a diverse network of available ocean observations in order to arrive at a dynamically consistent estimate of the historical ocean state. The nature of the underlying OGCM, data assimilation scheme and observations used varies, often according to the role for which the ORA is intended. For example, systems designed for near real-time forecasts of the ocean mesoscale tend to use higher resolution OGCMs and satellite altimeter data will feature strongly in the assimilation system. Conversely, systems that are used primarily for estimating the ocean state over the 20th Century will tend to use coarser resolution OGCMs and/or simpler assimilation schemes for reasons of computational expense. In addition, a number of products combine the available observations into spatially-complete gridded fields using purely statistical approaches, without the use of an OGCM.
Overview of the different ORAs used in this study
1°/3° no model
Optimal interpolation (T/S/SST)
1°/2° MOM4 coupled
Storto et al. (2011)
1°/3° MOM4 coupled
UK Met Office
1° no model
Optimal interpolation (T/S)
Ingleby and Huddleston (2007)
University of Hamburg
1 × 1°/3° MITgcm
Ferry et al. (2012)
UK Met Office
Blockley et al. (2014)
Rienecker et al. (2008)
1 × 1°/3° MOM3
Masuda et al. (2010)
0.3°–1° MRI.COM2 coupled
Fujii et al. (2009)
Fujii et al. (2015)
Toyoda et al. (2013)
1° no model
Objective analysis (T/S)
Levitus et al. (2012)
University of Maryland/Texas A&M University
0.4 × 1°/4° POP2.1
Carton and Giese (2008)
University of Reading
Haines et al. (2012)
OHC is a key variable for initialization of seasonal-to-decadal predictions (Balmaseda et al. 2010; Dunstone and Smith 2010) and the rate of ocean heat uptake under anthropogenic climate change plays an important part in determining future global surface temperature and sea level rise (Kuhlbrodt and Gregory 2012). Improved understanding the processes that control OHC variability and change may offer the potential to reduce uncertainties in future climate change projections through application of suitable observational constraints (e.g. Stott and Forest 2007). Thus ORAs have an important role to play in development of improved forecasts on a range of timescales and in refining our understanding of future global and regional climate change.
OHC variability and change is a particularly topical research area, given the strong scientific and wider media interest in the recent slowdown in surface temperature rise (e.g. Hawkins et al. 2014), often referred to as the global warming ‘hiatus’ (e.g. Trenberth and Fasullo 2013). Increased ocean heat uptake and the vertical re-arrangement of heat in the ocean have both been proposed as key mechanisms that have contributed to the ‘hiatus’. Heat re-arrangement has been linked primarily with the tropical Pacific (Meehl et al. 2011; England et al. 2014) but there is evidence that the higher latitudes may also have played a role (Chen and Tung 2014; Drijfhout et al. 2014; Roemmich et al. 2015).
ORAs provide an important resource for improving our understanding of the ocean’s role in modulating global surface temperature rise on interannual to decadal timescales. Analysis of climate model simulations has shown substantial vertical re-arrangement of ocean heat and highlighted the global ocean’s dominant role in Earth’s energy budget on annual-to-decadal timescales (Palmer et al. 2011; Palmer and McNeall 2014). Through combining the available ocean observations with OGCMs, ORAs may offer new insights into the processes of vertical heat re-arrangement and have also be used to derive estimates of Earth’s energy imbalance (Loeb et al. 2012; Trenberth et al. 2014; Smith et al. 2015).
Despite the recent development of the Argo network of profiling floats (Roemmich et al. 2009), historical observations of ocean temperature are sparse in time and space (Purkey and Johnson 2010; Desbruyères et al. 2014), often limited to a particular depth range, and may require correction for instrumental biases (Abraham et al. 2013). These issues mean that there are substantial—and difficult to quantify—uncertainties in our knowledge of ocean heat content change during the late twentieth and early twenty first centuries. One way to evaluate this uncertainty is using the spread of different reanalysis products that use ostensibly the same raw information, but different methodologies, to evaluate the same variable of interest. Previous studies have used similar ‘ensembles of opportunity’ to characterize the uncertainty in upper ocean heat content derived from statistical analyses (Lyman et al. 2010; Palmer et al. 2010; Abraham et al. 2013).
Estimates of the past ocean state are fundamentally limited by the availability of historical ocean profiles. Prior to the inception of the Argo array profiling floats in the early 2000s, reasonable ocean coverage is only afforded for the upper few hundred meters since the late 1960s (Roemmich et al. 2012; Lyman and Johnson 2014). As a result, many of the historical estimates of OHC variability and change have been limited to the upper 700 m or so (Lyman et al. 2010; Palmer et al. 2010; Abraham et al. 2013). In addition, the upper layers of the ocean are widely regarded to be the primary source of predictability for seasonal forecast systems, for example for initializing the tropical Pacific for ENSO forecasts (Xue et al. 2012). For these reasons, much of our analysis focuses ocean depth ≤700 m.
We present spatial patterns of depth-integrated temperature for the following: (i) the climatological time mean; (ii) the amplitude of the seasonal cycle; (iii) internannual variability; (iv) decadal and multi-decadal trends. We also compare zonal averages of (i) and (ii). The manuscript is organized as follows. In Sect. 2 we present an overview of the ORA datasets used. The pre-processing steps and analysis methods are presented in Sect. 3. The results are presented in Sects. 4–7 and cover the following aspects: the time-mean OHC and seasonal cycle; interannual variability; time series of OHC; and spatial trends of OHC. In Sect. 8 we present our closing discussion and summary.
A total of 19 ORAs are used in the intercomparison presented here. These span a range of time periods and have a diverse set of data assimilation methods and underlying ocean model configurations (Table 1). While the majority of products include a dynamic OGCM, three of the products are based on statistical analysis of the observations (ARMOR3D, EN3 and NODC) and do not include a dynamic model component. Note that the ARMOR3D product only covers the upper 2000 m. For the purposes of this intercomparison the deeper levels use the NODC climatology and therefore the variability and change signals below 2000 m for ARMOR3D are very small, by construction. We use a version of the EN3 analysis that is based on profiles with expendable bathythermograph (XBT) corrections applied following table 1 of Wijffels et al. (2008). The NODC data have mechanical baythermograph (MBT) and XBT bias corrections applied as described in Levitus et al. (2012). All data providers carried out the computation of vertically-integrated temperature for a number of vertical layers (see Sect. 3) and interpolated the data onto a regular 1 × 1 degree latitude-longitude grid. To ensure consistency among the grids used, all data were subsequently post-processed using the Climate Data Operator bilinear re-mapping tool.
Monthly climatologies for each product are computed for the period 1993–2009, with the exception of MOVECORE (see Sect. 2), for which we use the period 1993–2007. These climatologies are used in two aspects of the intercomparison. Firstly, the time-average for each grid box is computed to provide a comparison of the mean state of the period 1993–2009. Secondly, the amplitude of the climatological seasonal cycle is computed for each grid box by simply subtracting the minimum monthly value from the maximum monthly value. The remainder of the analysis presented here is carried out using annual mean values of 〈θ〉.
Maps of interannual variability are computed for the period 1993–2009 (except MOVECORE, which uses 1993–2007). The variability for each ORA is computed as the standard deviation of annual values for each grid box after removing a linear trend. Lastly, spatial maps of 〈θ〉 change are computed by fitting a linear trend to each grid box over the periods 1970–2009 and 1997–2009, with the latter period intended to characterize ocean heat content changes during the surface warming ‘hiatus’ period.
In our analysis of interannual variability (calculated for each product following the removal of ORA-specific climatologies), we follow the ‘signal-to-noise’ definition of Xue et al. (2012). Here, the ‘signal’ (S) is defined as the temporal standard deviation of the ensemble mean over a specified period. The ‘noise’ (N) is defined as the temporal average of the standard deviation of ensemble spread over the same period. This definition of S/N is thus a measure of the average spread across the ensemble relative to the signal of the ensemble mean. Areas where S/N < 1 can be interpreted as regions where differences in the underlying ocean models tend to dominate over the ability of the available observations and data assimilation schemes to constrain the ORA solutions. In the other spatial map comparisons presented here we often show the ensemble mean (M), the ensemble standard devidation (SD) and the ratio of the two (M/SD) to provide an indication of the spread and level of agreement among ORAs.
4 Time-mean and amplitude of the seasonal cycle
Many of the analyses show a widespread cool or warm bias, relative to the ensemble mean (e.g. SODA, GloSea). Disentangling the roles of differing ocean vertical mixing, ocean circulation and surface forcings in determining these biases is a challenging problem and requires further and detailed analysis. Errors in both momentum and heat fluxes have a role here (Lee et al. 2013; Valdivieso et al. 2015) and consideration of mixed layer depths may also offer additional insights (Toyoda et al. 2015). The representation of vertical mixing in ocean models remains an area that requires particular attention due to the different mixing schemes, related parameters and the often poorly quantified effects of numerical mixing (e.g. Buchard et al. 2008). Regionally, the departures from the ensemble mean can exceed ±500 Cm, or 0.7 C in terms of the depth-averaged temperature (Fig. 1). However, the zonal averages show that all analyses are generally within ±350 Cm (or 0.5 C) of the ensemble mean between 60S and 60N (Fig. 2). The increased spread north of 60N may be indicative of differing representation of the Arctic marginal ice zones and may be exacerbated by the reduction of ocean grid points at these latitudes.
5 Interannual variability
6 Time series of OHC change
Some of the full depth OHC time series show very large rates of change—exceeding 3 W m−2 globally during some periods—and such changes cannot be reconciled with satellite-based reconstructions of Earth’s energy imbalance (Allan et al. 2014) or the global sea level budget (Church et al. 2011). CFSR (Fig. 9a) shows very some large, almost oscillatory, changes in OHC below 700 m that may be linked to previously documented discontinuities in deep ocean temperature and salinity (Xue et al. 2011). For the MOVE products (Fig. 9b), the large rates of OHC change are a known feature related to the use of vertical empirical orthogonal functions in those products.
As one would expect, the 0–300 m layer shows a relatively narrow spread, particularly for the Northern Hemisphere (Fig. 10). It is interesting to note that the Northern and Southern Hemisphere appear to be anti-correlated in these upper layers and this could be related to changes in the AMOC and associated changes in inter-hemispheric heat transport. The magnitude of mean global OHC change for the 0–300 m layer peaks at about 0.5 W m−2. The 300–700 m layer shows similar rates and sign of OHC change in both hemispheres, but a much larger spread in the Southern Hemisphere. Peak rates of the mean global OHC change are about 0.4 W m−2 for this layer. Below 700 m the mean global rate of OHC change is dominated by the Southern Hemisphere with the peak rate around 1997/1998. OHC integrated over the full column (Fig. 10, 0–6000 m) shows a fairly steady mean global rate of 0.9 ± 0.8 W m−2 for the period 1995–2002 that drops to 0.2 ± 0.6 W m−2 for the period 2004–2006 (error bars indicate 1 standard deviation of ensemble mean), in qualitative agreement with the findings of Smith et al. (2015). However, there is no evidence in our ORA ensemble mean of the ‘spike’ in OHC trends reported by Smith et al. in the early 2000s.
In terms of the ORA ensemble standard deviation, the spread in systematically less in the Northern Hemisphere than the Southern Hemisphere. This is probably a combination of both the better observational coverage and the lesser ocean volume north of the equator. There is a tendency towards reduced spread for all series over time, but this is not monotonic. Although the data here are limited to 2009, we appear to see the impact of Argo observations becoming available in the sharp reduction in ORA ensemble spread for the Southern Hemisphere below 300 m after about 2004 (Fig. 10).
7 Spatial patterns of OHC change
We have presented a comparison of the representation of OHC variability and change as estimated by 16 state-of-the-art ORAs, focusing on five main aspects: (i) the time-mean state during 1993–2009; (ii) the amplitude of the seasonal cycle during 1993–2009; (iii) interannual variability during 1993–2009; (iv) global and hemispheric time series; (v) spatial patterns of change for 1970–2009 and 1997–2009.
The time-mean state and amplitude of the seasonal cycle for the 0–700 layer are generally well-constrained among the ensemble, but regional differences may warrant further investigation. We find that interannual variability for the 0–300 m can be resolved by the ORA ensemble over the majority of the Indo-Pacific and much of the Atlantic. However, the signal-to-noise ratio rapidly diminishes as one considers deeper layers.
Global time series of OHC show similar rates of change and ensemble spread to previous studies based only on statistical estimates. Comparison of the hemispheric OHC time series illustrates that the majority of the spread among analyses originates in the Southern Hemisphere. Several ORAs, including those that do not include an OGCM are in qualitative agreement about an increase in ocean heat uptake below 300 m from the mid 1990s or early 2000s. However, a number appear to suffer from large initialization shock and/or drifts in the deep ocean (where this is little observational constraint on the OGCM solution) so caution should be taken when using such products to estimate the global ocean heat inventory.
Spatial patterns of OHC change for the 0–300 m layer over the period 1970–2009 show large areas of agreement (ensemble mean trend > standard deviation of trends), based on the four analyses included here. While this area is reduced for the 300–700 m layer there remains some agreement on large-scale regions of warming in the Atlantic, Pacific and Southern Oceans. For the 700–6000 m layer only isolated regions of the Southern Ocean and central Atlantic show trends that are resolved by the ensemble spread.
The spatial trends of OHC for the 1997–2009 period generally show large areas of agreement for the 0–300 m, with the Pacific characterized by an east–west dipole. At deeper levels there is little or no agreement among the ensemble. However, the ensemble mean trends for the 700–6000 m layer show the largest warming trends in the North Atlantic sub-polar gyre, north Indian Ocean and Southern Ocean.
We would like to thank two anonymous reviewers for providing helpful comments on an earlier version of this manuscript. This work is a result of collaboration among the CLIVAR Global Synthesis and Observations Panel and the GODAE Ocean View communities. This work was partly funded by the Joint UK DECC/Defra Met Office Hadley Centre Climate Programme (GA01101) and represents a Met Office contribution to the Natural Environment Research Council DEEP-C project NE/K005480/1.
Compliance with ethical standards
Conflict of interest
The authors declare no conflicts of interest. All authors have given consent to the submission.
- Balmaseda M et al (2010) Initialization for seasonal and decadal forecasts. In: Hall J, Harrison DE Stammer D (eds) Proceedings of OceanObs’09: sustained ocean observations and information for society, vol 2, Venice, Italy, 21–25 September 2009, ESA Publication WPP-306. doi:10.5270/OceanObs09.cwp.02
- Balmaseda MA et al (2015) The Ocean Reanalyses Intercomparison Project (ORA-IP). J Oper Oceanogr 7:81–99Google Scholar
- Behringer DW (2007) The global ocean data assimilation system at NCEP. Preprints, 11th Symposium on integrated observing and assimilation systems for atmosphere, oceans, and land surface, San Antonio, TX, Amer. Meteor. Soc., 3.3. https://ams.confex.com/ams/87ANNUAL/webprogram/Paper119541.html
- Ferry N et al (2012) GLORYS2V1 global ocean reanalysis of the altimetric era (1993–2009) at meso scale. Mercat Ocean Q Newsl 44:28–39. http://www.mercator-ocean.fr/content/download/1625/12971/version/1/file/GLORYS2V1-newsletter44.pdf
- Fujii Y, Tsujino H, Toyoda T, Nakano H (2015) Enhancement of the southward return flow of the Atlantic Meridional Overturning Circulation by data assimilation and its influence in an assimilative ocean simulation forced by CORE-II atmospheric forcing. Clim Dyn. doi:10.1007/s00382-015-2780-1 Google Scholar
- Hanawa K, Talley LD (2001) Mode waters. In: Siedler G, Church J (eds) Ocean circulation and climate. International geophysics series. Academic Press, New York, pp 373–386Google Scholar
- Mogensen K, Balmaseda MA, Weaver AT (2012) The NEMOVAR ocean data assimilation system as implemented in the ECMWF ocean analysis for system 4. Tech memo 668. ECMWF, Reading, UKGoogle Scholar
- Palmer M et al (2010) Future observations for monitoring global ocean heat content. In: Hall J, Harrison DE, Stammer D (eds) Proceedings of OceanObs’09: sustained ocean observations and information for society, vol 2, Venice, Italy, 21–25 September 2009, ESA Publication WPP-306. doi:10.5270/OceanObs09.cwp.68
- Rienecker M et al (2008) The GEOS-5 data assimilation system—documentation of versions 5.0.1, 5.1.0, and 5.2.0. NASA/TM-2008-104606, vol 27Google Scholar
- Rienecker M et al (2010) Synthesis and assimilation systems—essential adjuncts to the global ocean observing system. In: Hall J, Harrison DE, Stammer D (eds) Proceedings of OceanObs’09: sustained ocean observations and information for society, vol 1, Venice, Italy, 21–25 September 2009, ESA Publication WPP-306. doi:10.5270/OceanObs09.pp.31
- Rhein M et al (2013) Observations: ocean. In: Climate change 2013: the physical science basis. In: Stocker TF, Qin D, Plattner G-K, Tignor M, Allen SK, Boschung J, Nauels A, Xia Y, Bex V, Midgley PM (eds) Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USAGoogle Scholar
- Valdivieso M et al (2015) An assessment of air-sea heat fluxes from ocean and coupled reanalyses. Clim Dyn (submitted)Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.