Real Climate has a post titled Updates to model-data comparisons which includes a plot of the variations in upper ocean content anomalies from the period 1955 through 2009 .
I asked Josh Willis the following with respect to the plot in the Real Climate post
Real Climate has posted a plot of ocean heat content, which we have
discussed before, that shows a sudden jump in the 2002-2003 time frame;
This jump is not seen it other metrics, including the surface temperatures
(which they show) or the lower tropospheric temperatures (e.g. see
see Figure 7 TLT
Can you comment on the realism of this jump? Would you be willing to let me
post your reply, if you do comment?
Most of their trend agreement with the models is due to this single jump.
Josh Willis’s reply [reproduced with his permission]
There is still a good deal of uncertainty in observational estimates of ocean heat content during the 1990s and into the early part of the 2000s. This is because of known biases in the XBT data set, which are the dominant source of ocean temperature data up until 2003 or 2004. Numerous authors have attempted to correct these biases, but substantial difference remain in the “corrected” data. As a result, the period from 1993 to 2003 still has uncertainties that are probably larger than the natural or anthropogenic signals in ocean heat content that happen over a period of 1 to 3 years. However, the decadal trend of 10 to 15 years seems to be large enough to see despite the uncertainties. Because Argo begins to become the dominant source of temperature data in about 2004, the period from 2000 to 2005 is especially worriesome because of the transition from an XBT-dominated estimate of ocean heat content.
You might also comment that there is another easily available estimate besides that of Levitus et al. (the one shown in this blog entry). The other long-term estimate is from Domingues et al. and can be downloaded from CSIRO: