by Judith Curry
Central to arguments related to the hiatus and the ‘missing heat’ is the assertion that unusual amounts of heat are being stored in the deep ocean, and that this heat will eventually reappear at the surface. Exactly how good is the ocean heat content data on which this argument is based?
At the time of the AR4 (2007), our view of ocean heat content is reflected by Figure 5.1 from the AR4:
Figure 5.1. Time series of global annual ocean heat content (1022 J) for the 0 to 700 m layer. The black curve is updated from Levitus et al. (2005a), with the shading representing the 90% confidence interval. The red and green curves are updates of the analyses by Ishii et al. (2006) and Willis et al. (2004, over 0 to 750 m) respectively, with the error bars denoting the 90% confidence interval. The black and red curves denote the deviation from the 1961 to 1990 average and the shorter green curve denotes the deviation from the average of the black curve for the period 1993 to 2003.
The equivalent figure from the AR5 is shown below:
Figure 3.2: a) Observation-based estimates of annual global mean upper (0–700 m) ocean heat content in ZJ (1 ZJ =1021 Joules) updated from (see legend): (Levitus et al., 2012), (Ishii and Kimoto, 2009), (Domingues et al., 2008), (Palmer et al., 2007), and (Smith and Murphy, 2007). Uncertainties are shaded, and plotted as published (at the one standard error level, except one standard deviation for Levitus, with no uncertainties provided for Smith). Estimates are shifted to align for 2006–2010, five years that are well measured by Argo, and then plotted relative to the resulting mean of all curves for 1971, the starting year for trend calculations.
The larger uncertainty in 1990 is not surprising, but I am surprised by the markedly different trends for the past decade, since 2003. Some climatologies show no trend since 2003, whereas others show continued warming.
Note, in comparing the figures from AR4 with AR5, multiply the AR4 OHC values by 10 to have the same units. Also the two figures use a different baseline period for the anomalies. There are some striking differences between these two figures, even if you only focus on the curves from Levitus and Ishii (which are common to both figures). The most striking thing is the much greater uncertainty represented by a larger number of data sets in the AR5 collection. Also the different data sets are estimating the uncertainties in different ways, as reflected by the much different width of the shaded uncertainties. The second thing that struck me about the comparison is that the substantial ‘bump’ that peaked circa 1980 in the AR4 data sets has disappeared in all of the AR5 data sets. It seems that there is a great deal of uncertainty in calibration of the XBT measurements during the 1970’s and 1980’s.
A new paper in press at J. Climate by Lyman and Johnson [link] provides a good overview of the uncertainties in the OHC data, and examines the impact of different choices on assumptions made about missing data.
FIG. 4. Time series of annual average global integrals of upper ocean heat content anomaly (1021 J, or ZJ) for (a) 0–100 m, (b) 0–300 m, (c) 0–700 m, and (d) 0–1800 m. Time series are shown using ZIF estimates relative to both ClimArgo (dashed grey lines) and Clim1950 (dashed black lines). Time series are also shown using REP estimate (black solid lines), which are not affected by shifts in the mean climatology (B11). Thin vertical lines denote when the coverage (Fig. 3) reaches 50% for (a) 0–100 m, (b) 100– 300 m, (c) 300–700 m, and (d) 900–1800 m.
The authors regard the REP values as the best ones. The vertical bar in Fig 4 above denotes when the coverage reaches 50%. Note that for measurements to 700 m, 50% coverage was reached in 1984. The three different curves represent 3 climatologies based on different assumptions about under sampled or unsampled regions of the ocean. The two main features that strike me in Fig 4 is the sharp increase from 1995-2003, and then the flat trend since 2003. Also the sharp increase is more evident in the whole layer 0-1800 m than in the shallow layers near the surface, but note that 50% coverage was achieved for the layer 900-1800 m only since 2005.
This figure from AR5 provides additional perspective on the data sampling:
There are very limited observations of the ocean deeper than 2000 m. The IPCC AR5 provides the following figure:
Figure 3.2: b) Observation-based estimates of annual five-year running mean global mean mid-depth (700–2000 m) ocean heat content in ZJ (Levitus et al., 2012) and the deep (2000 – 6000 m) global ocean heat content trend from 1992–2005 (Purkey and Johnson, 2010), both with one standard error uncertainties shaded (see legend).
Well, the error bars in Figure 3.2b seem rather skimpy, but the OHC increase in the deep ocean (below 2000 m) seems pretty small.
Reanalysis versus observations
So exactly where does the argument come from that the deep ocean is sequestering the ‘missing heat’? It seems to come from the Balmaseda et al paper that is based on ocean reanalysis (this paper was discussed here at Climate Etc.). The main figure of interest:
Figure 1: Ocean Heat Content from 0 to 300 meters (grey), 700 m (blue), and total depth (violet) from ORAS4, as represented by its 5 ensemble members. The time series show monthly anomalies smoothed with a 12-month running mean, with respect to the 1958–1965 base period. Hatching extends over the range of the ensemble members and hence the spread gives a measure of the uncertainty as represented by ORAS4 (which does not cover all sources of uncertainty). The vertical colored bars indicate a two year interval following the volcanic eruptions with a 6 month lead (owing to the 12-month running mean), and the 1997–98 El Niño event again with 6 months on either side. On lower right, the linear slope for a set of global heating rates (W/m2) is given.
Now, the theoretical advantage of ocean data assimilation is that it ‘fills in’ unsampled regions using the model dynamics and thermodynamics. Lets compare the Balmaseda et al. reanalysis with the observational climatologies. Focus first on the 0-700 curves, and compare with the corresponding figures in the AR5 and Lyman & Johnson. Balmaseda et al. shows a large increase from 1983-1992 (between the two volcanoes), whereas most of the observational climatologies show little trend during this period and none show a large trend during this entire period. The strong warming trend shown by the observations during the period 1995-2003 followed by weaker trend since 2003, contrasts with Balmaseda that shows no trend between 1992 and 2000, and then a strong warming trend since 2000.
The most surprising thing about the Balmaseda analysis is that the warming increases with increasing depth (largest warming for the 0-7000 m layer). In comparing Balmaseda with the other figures, pay attention to the different scaling for the OHC. But the bottom line is that there does not seem to be any observational support for this large sequestration of heat in the deep ocean that is shown by the reanalysis.
To gain further insights into ocean heat sequestration, it is useful to look at the regional variations of OHC anomalies. A presentation by Levitus provides some regional analyses of trends over the period 1955-2010. Which is useful, but I am particularly interested in the trends since 2000 (the period of the large sequestration as per the reanalysis), and I haven’t come across any publications on this (does anyone have some references?). Bob Tisdale provides this plot:
Warming trends (0-2000 m) are seen in the Indian Ocean and the South Atlantic, with slight cooling trends in the Pacific and North Atlantic. Now it seems difficult to me to cook up an explanation for this regional variation in trends that relies on external forcing, although I suspect that someone will think of some rationale for aerosol/black carbon forcing to explain this. This most likely reflects natural internal variability. It doesn’t look like an AGW signal to me. More regional analyses for the past decade would be very helpful in trying to sort this out.
Sea level rise
It is very difficult to sort out the causes of OHC variability owing to the short time record. Some further insights into longer term OHC variability can be inferred from this figure from the AR5 on rates of sea level rise:
Figure 3.14: 18-year trends of GMSL rise estimated at 1-year intervals. The time is the start date of the 18-year period, and the shading represents the 90% confidence. The estimate from satellite altimetry is also given, with the 90% confidence given as an error bar. Uncertainty is estimated by the variance of the residuals about the fit, and accounts for serial correlation in the residuals as quantified by the lag-1 autocorrelation.
Note the high values in the early part of the century, nearly as high or as high as the value for the last two decades. Now there are other factors that contribute to sea level rise changes; from the AR5 chapter 3 (a table included in my recent testimony):
- AR5 (1993-2010)
- Thermal expansion 1.1
- Glaciers and ice caps 0.76
- Greenland ice sheet 0.33
- Antarctic ice sheet 0.27
- Land water storage 0.38
- Sum 2.8
- Observed sea level rise 3.2
OHC changes (thermal expansion) accounts for about 1/3 of the total sea level rise. What did this balance look like circa 1930’s to 1950’s? Presumably the land water storage and glacier melt was smaller, so the thermal expansion was more dominant in this early period. Which suggests that ocean heat content was greater in this early period than in the current period, and cannot be attributed to AGW.
Roger Pielke Sr. has often stated that ocean heat content is a much better metric for climate change than surface temperature. I don’t prefer one over the other as an intrinsic metric (they provide two different pieces of information), but I find the ocean heat content data to be a much less mature data set than the surface temperature data set. The sampling particularly of the mid to deep ocean is very sparse prior to 2000. And the oceanographic community is still debating the calibration of MBT and XBT profiles. There is substantial disagreement among the various OHC climatologies, and there are no OHC climatologies prior to 1950. Global sea level trend data suggests substantial thermal expansion in the earlier part of the 20th century, which is an issue that seems insufficiently explored.
Ocean reanalyses can potentially provide new insights into global OHC variations, but ocean reanalysis is in its infancy.
The main issue of interest is to what extent can ocean heat sequestration explain the hiatus since 1998. The only data set that appears to provide support for ocean sequestration is the ocean reanalysis, with the Palmer and Domingues 0-700 m OHC climatology providing support for continued warming in the upper ocean.
All in all, I don’t see a very convincing case for deep ocean sequestration of heat. And even if the heat from surface heating of the ocean did make it into the deep ocean, presumably the only way for this to happen involves mixing (rather than adiabatic processes), so it is very difficult to imagine how this heat could reappear at the surface in light of the 2nd law of thermodynamics.