Observations of planetary heating since the 1980s from multiple independent datasets

Time series of global mean surface temperature are widely used to measure the rate of climate change that results from Earth’s energy imbalance. However, studies based on climate model simulations suggest that on annual-to-decadal timescales global ocean heat content is a more reliable indicator. Here we examine the observational evidence for this, drawing together multiple datasets that span the past ∼30 years. This observational analysis strongly supports the model-based finding that global ocean heat content and sea level are more reliable than surface temperature for monitoring Earth’s energy accumulation on these timescales. Global ocean temperature anomalies in the 0–100 m and 100–250 m layers are negatively correlated (r = −0.36), primarily explained by the influence of the Tropical Pacific, and a clearer heating signal is revealed by integrating over deeper ocean layers. The striking agreement between multiple independent datasets represents unequivocal evidence of ongoing planetary heating.


Introduction
Greenhouse gas emissions have caused a persistent radiative imbalance at the top of the atmosphere (TOA), referred to as Earth's energy imbalance (EEI), resulting in ongoing planetary heating that is driving the various facets of observed climate change (von Schuckmann et al 2016). Time series of global mean surface temperature (GMST) are widely used to quantify the rate of anthropogenic climate change and to define warming targets for policy discussions. However, GMST is strongly influenced by internal variability and its decadal trends do not always reflect the underlying long-term warming, for example during the recent 'hiatus' period where temperature rise stalled despite strong evidence of sustained planetary heating associated with greenhouse gas forcing (e.g. England et al 2014, Medhaug et al 2017; figure 1). More than 90% of the multi-decadal EEI is manifested in global ocean heat content (GOHC) gain (von Schuckmann et al 2016). Climate model evidence suggests that on decadal timescales GOHC provides a more robust measure of EEI than GMST does (Palmer et al 2011, Palmer andMcNeall 2014). But what is the observational evidence for this?Studies that examined a recent decade of improved ocean observations (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) showed that year-to-year variations in GOHC agree well with independent measurements of TOA radiances, promoting confidence in our ability to observe variations in EEI at sub-decadal timescales  and, in contrast to the variable surface temperature record, the global sub-surface ocean (300-2000m) warmed steadily over this period (Wijffels et al 2016, Cheng et al 2017a. Here we examine EEI over a longer time horizon (∼30 years) and make use of new reconstructions of accumulated TOA flux and global mean sea level (GMSL) alongside GMST and GOHC estimates to explore the agreement between these independent observational datasets.  . We address this issue by taking the mean of six GOHC products to reduce the noise in order to better estimate the signal of GOHC change and its variations.
Estimates of GMST are better constrained by the available observations than those of GOHC, but we follow a similar approach in deriving a time series of observed GMST (1986GMST ( -2018

Results
The time series of planetary heat content anomaly inferred from time-integrated TOA flux reveals a clear and largely monotonic increase since the mid-1980s ( figure 1(a)). In contrast, GMST exhibits considerable interannual and decadal variability. This variability is largely absent from GOHC and GMSL (which is closely linked to GOHC through thermal expansion); these 'full-ocean' variables more closely replicate the nearmonotonic planetary heating inferred from TOA radiative flux measurements. It is clear that interannual trends in GMST are dominated by near-surface variability and are not representative of changes in planetary heat content.
As GOHC is integrated to deeper levels, the surface noise is diminished. This can be seen in figure 1(b), which compares the TOA-implied heating with the MOSORA GOHC anomalies over various depth ranges, with each time series normalised by its own standard deviation to emphasise the change signals captured in each layer. Upper 100 m GOHC contains significant interannual variability, with features similar to those seen in the GMST time series, reflecting the physical link between GMST and the heat content of the upper ocean mixed layer. However, even full-column GOHC estimates show some variability overlying the trend, some of which is likely artificial and may be attributed to residual noise associated with limited ocean sampling and changes in observing practices over time (Abraham et al 2013, Smith et al 2015, Allison et al 2019. The results in figure 1(b) suggest that on this ∼30-year timescale, integrating GOHC to 300 m depth removes much of the near-surface noise and captures the character of the planetary heating signal. Integrating to deeper limits yields normalised signals that are similar to that of the upper 300 m. However, comparison of non-normalised time series (not shown) reveals that the deeper layers are important for capturing the magnitude of the long-term heating trend. In MOSORA, the 0-100 m layer captures 14% of the linear trend in the full-depth GOHC over 1986-2018, while the 0-300 m and 0-700 m layers capture 38% and 58% of the full-depth trend, respectively. The depth structure of global mean ocean temperature variability (figure 2) reveals layers of anticorrelated anomalies above and below 100 m Gilson 2011, Wijffels et al 2016) demonstrating that surface temperature variations are not representative of changes in deeper ocean heat content. These anticorrelated layers can be traced to vertical heat rearrangement in the Tropical Pacific associated with the El Niño Southern Oscillation (ENSO) on interannual timescales. Variations in the strength of the Pacific trade winds alter the subduction and convergence of heat in the equatorial thermocline and upwelling of cool water into the surface layer, changing the thermocline's east-west tilt (Roemmich and Gilson 2011). It can be seen in figure 2 that for a positive ENSO index (El Niño), the upper layer is anomalously warm and the lower layer anomalously cool, with a reversal of this pattern during negative events (La Niña). Global ocean temperature anomalies in the 0-100 m and 100-250 m layers are negatively correlated (r= −0.36). This anticorrelation between ocean layers is dominated by the Tropical Pacific; when this region (120°-280°E, 10°S-10°N) is excluded from the global mean (not shown) the correlation becomes positive (r=+0.37), illustrating the impact of regional Tropical Pacific temperature variations on the global mean. In addition to the interannual variability within the upper few hundred metres, figure 2 also reveals clear decadal variability in sub-surface global mean ocean temperature that extends to 2000 m depth. The periods of sub-surface cooling (∼1970-1995) and warming (∼1995-2018) show close correspondence to observed epochs of positive and negative trends in the Interdecadal Pacific Oscillation (IPO) respectively, indicating that sub-surface ocean heat rearrangement also plays a role in global mean surface temperature variability on decadal timescales (England et al 2014, Meehl et al 2016. These modes of variability in the Pacific have been identified as important drivers of unforced variability in global mean surface temperature, but variations in the Atlantic Ocean and external forcings may also play a role (Dai et al 2015, Smith et al 2016.

Discussion
This observational analysis strongly supports previous findings based on climate model simulations, illustrating the de-coupling between EEI and GMST on decadal and shorter timescales. This de-coupling occurs primarily due to dynamic ocean heat rearrangement processes associated with climate variability in the Pacific. GOHC is largely independent of these internal rearrangements and remains strongly indicative of EEI on all timescales, exhibiting a much steadier rise than GMST. GMST is a fundamental quantity to monitor; it has a long and reliable historical record and it plays a central role in determining many important climate impacts. However, the implication here is that GOHC presents a more reliable basis for drawing insights on the evolving magnitude of EEI on decadal and shorter time periods. Our ability to track EEI for climate monitoring relies on a suite of complementary observation sources, including GOHC from sustained and improved ocean observations (e.g. the international Argo program, amongst others) as well as TOA measurements, and may also be enhanced through schemes that incorporate observations of GMSL in a physically consistent way (Meyssignac et al 2019). The striking agreement between the independent observational datasets of time-integrated net TOA flux, GOHC and GMSL (as well as multi-decadal trends in GMST) represents unequivocal evidence of ongoing planetary heating.