Two recent papers, Zanna et al. (2019) (hereafter ZKGIH19) and Gebbie & Huybers (2019) (hereafter GH19), independently reconstructed ocean heat content (OHC) changes prior to the instrumentally-based records (which start ~1950). The goals (and methodologies) of the two papers were quite different – ZKGIH19 investigated regional patterns of ocean warming and thermal sea level rise, while GH19 analyzed the long-term memory of the deep ocean – but they both touch on the same key questions of climate forcing and response.
The two studies independently show that subsurface temperature change is well described on century-long timescales by surface imprints that are transported by the modern ocean circulation. Both studies highlight the following points: changes in ocean circulation have very little impact on global OHC changes, and that the deep ocean adjusts slowly to surface temperature changes, showing a slow emergence of anthropogenic trends at depth in the recent decade.
The two methodologies are distinct: ZKGIH19 uses a Green’s function approach to derive the imprint on the deep ocean of changes in surface fluxes from the ECCO project (a reanalysis model that uses modern ocean data to reconstruct a state estimate of the ocean circulation). This is then combined with the history of observed SST (HadISST2) to infer the deep ocean changes.
GH19 use a functionally similar inversion technique for ocean tracers (outlined in Gebbie and Huybers (2011)) to create a forward model that produces an estimate of deep ocean changes given the surface forcing. They then constrain this estimate using the difference between the HMS Challenger data (1872–1876) and modern WOCE data to produce an optimised estimate of ocean temperature structure going back to 15 CE. Unfortunately, there isn’t an uncertainty quantification that goes with this.
For 1955–2017, both estimates in the upper and deep ocean are similar to estimates made by infilling the available 3D time-dependent ocean temperature observations (GH19: 0.35 W/m2, ZKGIH19: 0.33 ± 0.07 W/m2 (one sigma)). The warming between 1921-1946 is also consistent between both estimates (ocean heat uptake GH19: 175 ZJ, ZKGIH19: 145±62 ZJ (1 sigma)) and is comparable to the 1990-2015 time interval (GH19 = 135 ZJ, ZKGIH19: 153 ± 44 ZJ). Even when the time interval is pushed back to 1880, the two numerical estimates of ocean heat uptake are self-consistent.
Below 2 kilometers depth and before 1880, the two estimates show larger differences. For the time period, 1871-2015, the GH19 total estimate (570 ZJ) is bigger than ZKGIH19 (436 ± 91 ZJ), but not significantly so. (Note that GH19 ends in 2015, ZKGIH19 in 2017.)
The differences probably have 2 major causes: 1) the lingering influence of surface climate from before 1870, and 2) discrepancies in the estimated surface warming in the late 19th century. Below 2 kilometers depth, the heat uptake has differing signs for the interval, 1871-2015 in the two analyses (GH19: -25 ZJ, ZKGIH19: approx. 14 ZJ ± 6 ZJ), due to a cooling up from 1971 until 1950 in the deep ocean in the reconstruction in GH19 compared to ZKGIH19. Surface warming is stronger in GH19 in the late 19th century due to ingestion of data from the HMS Challenger observations and Ocean2k paleo reconstructions (though errors are included in the reconstructions), while ZKGIH19 only used sea surface temperatures starting from 1871. The differences could also be due to the different transport models, GH19 used WOCE data in parts of the 1990s, while ZKGIH19 used ECCO over 1992-2003 (which nonetheless incorporated WOCE data as part of it’s input).
ZKGIH19 shows that regional changes in ocean circulation have an imprint on patterns of OHC and thermosteric sea level in the Atlantic Ocean over 1955-2016. By comparing with observations, they argue that up to half of the observed ocean warming and thermosteric sea level trends between 20ºS and 50ºN are due to time-dependent ocean horizontal and vertical redistribution. They showed that there are large variations in patterns of warming between the early and late periods. GH19 mostly focus on the deep Pacific and showed large OHC cooling trends over the past century, (but did not show global latitudinal distributions). They argue that the deep Pacific cooling is a signature of long-term adjustments of the ocean after the Little Ice Age. ZKGIH19 shows basin-scale estimates but those include the Southern Ocean and cannot be directly compared to GH19 basin estimates which are north of 45ºN. A more apples-to-apples comparison will take a bit more effort to produce.
(Additional text from gavin)
Given the average transit time for the deep Pacific (1000’s of years), it is expected that the deep Pacific won’t be in equilibrium with surface climate changes over shorter time scales. GH19 are not the first to quantify this deep dis-equilibrium (previous work had looked at the lingering impact of Tambora in 1815 for instance), but this might be the estimate most consistent with the (sparse) early observations. The caveats are (as usual) there are still imperfections in the ocean models being used and the systematic biases in old observations are always being looked at. The differences between the two studies are thus understandable from their study design.
These long-term estimates are however an interesting new metric to compare to the models. Just for fun, I plot (below) the total ocean heat uptake from the historical GISS CMIP5 model ensembles (normalized to 1871). These ensembles have two different ocean models (version R and version H), and two different treatments of atmospheric composition (non-interactive and interactive) and start from quasi-equilibrium in 1850. There are still residual drifts in the deep ocean which have been subtracted out using the control runs. The ocean model definitely makes a difference, and the GISS-E2-R runs had excessive mixing down of heat into the ocean (in the CMIP6 version this is reduced). Both sets of simulations have more cumulative heating than either the ZKGIH19 or GH19 estimates, though whether that is typical of the CMIP5 ensemble, or the new CMIP6 runs, is still to be determined. Note that an excessive uptake of ocean heat would be associated with a lower transient climate response (TCR), but the match in the upper ocean for more recent periods suggests it is likely that there may be issues with early 20th C forcings. It will also be interesting to see whether the longer millennial simulations (starting in 850 CE) might have a different pattern… Watch this space!
- L. Zanna, S. Khatiwala, J.M. Gregory, J. Ison, and P. Heimbach, "Global reconstruction of historical ocean heat storage and transport", Proceedings of the National Academy of Sciences, vol. 116, pp. 1126-1131, 2019. http://dx.doi.org/10.1073/pnas.1808838115
- G. Gebbie, and P. Huybers, "The Little Ice Age and 20th-century deep Pacific cooling", Science, vol. 363, pp. 70-74, 2019. http://dx.doi.org/10.1126/science.aar8413
- G. Gebbie, and P. Huybers, "How is the ocean filled?", Geophysical Research Letters, vol. 38, pp. n/a-n/a, 2011. http://dx.doi.org/10.1029/2011GL046769
- G. Stenchikov, T.L. Delworth, V. Ramaswamy, R.J. Stouffer, A. Wittenberg, and F. Zeng, "Volcanic signals in oceans", Journal of Geophysical Research, vol. 114, 2009. http://dx.doi.org/10.1029/2008JD011673