Global integrals of 0 to 700 m UOHC (Figure 3.2a) estimated from ocean temperature measurements all show a gain from 1971 to 2010 (Palmer et al., 2007; Smith and Murphy, 2007; e.g., Domingues et al., 2008; Ishii and Kimoto, 2009; Levitus et al., 2012) . These estimates usually start around 1950, although as noted in Section 3.2.1 and discussed in the Appendix, historical data coverage is sparse, so global integrals are increasingly uncertain for earlier years, especially prior to 1970. There is some convergence towards agreement in instrument bias correction algorithms since AR4 (Section 3.2.1), but other sources of uncertainty include the different assumptions regarding mapping and integrating UOHCs in sparsely sampled regions, differences in quality control of temperature data, and differences among baseline climatologies used for estimating changes in heat content (Lyman et al., 2010). Although there are still apparent interannual variations about the upward trend of global UOHC since 1970, different global estimates have variations at different times and for different periods, suggesting that sub-decadal variability in the time rate of change is still quite uncertain in the historical record. Most of the estimates in Figure 3.2a do exhibit decreases for a few years immediately following major volcanic eruptions in 1963, 1982 and 1991 (Domingues et al., 2008).
Again, all of the global integrals of UOHC in Figure 3.2a have increased between 1971 and 2010. Linear trends fit to the UOHC estimates for the relatively well-sampled 40-year period from 1971 to 2010 estimate the heating rate required to account for this warming: 118 [82 to 154] TW (1 TW = 1012 watts) for Levitus et al. (2012), 98 [67 to 130] TW for Ishii and Kimoto (2009), 137 [120 to 154] TW for Domingues et al. (2008), 108 [80 to 136] TW for Palmer et al. (2007), and 74 [43 to 105] TW for Smith and Murphy (2007). Uncertainties are calculated as 90% confidence intervals for an ordinary least squares fit, taking into account the reduction in the degrees of freedom implied by the temporal correlation of the residuals. Although these rates of energy gain do not all agree within their statistical uncertainties, all are positive, and all are statistically different from zero. Generally the smaller trends are for estimates that assume zero anomalies in areas of sparse data, as expected for that choice, which will tend to reduce trends and variability. Hence the assessment of the Earth’s energy uptake (Box 3.1) employs a global UOHC estimate (Domingues et al., 2008) chosen because it fills in sparsely sampled areas and estimates uncertainties using a statistical analysis of ocean variability patterns.
Globally integrated ocean heat content in three of the five 0 to 700 m estimates appear to be increasing more slowly from 2003 to 2010 than over the previous decade (Figure 3.2a). Although this apparent change is concurrent with a slowing of the increase global mean surface temperature, as discussed in Box 9.2, this is also a time period when the ocean observing system transitioned from predominantly XBT to predominantly Argo temperature measurements (Johnson and Wijffels, 2011). Shifts in observing systems can sometimes introduce spurious signals, so this apparent recent change should be viewed with caution.