A while back, there was quite a “kerfluffle” (as Lucia called it) regarding the comparisons between the GISS model projections for Ocean Heat Content (OHC) and those calculated by NODC. However, the thing that struck me at the time was that I could not easily find any available averages for the 700 m OHC projections for any of the model runs. If I recall correctly, RealClimate, Tamino, and Bob Tisdale were simply using linear extrapolations.
Since recently I had been doing my own OHC calculations from the NOAA/NODC data, I decided I would register for an account at PCMDI, download the gridded ocean potential temperature data, and perform the averaging myself to derive estimates of OHC for some model projections. Today, I will be showing the five runs from GISS-ER under the SRESA1B scenario (720 ppm stabilization) from CMIP3. I’d also like to do GISS-EH, but those files are a lot bigger, and taking longer to download and process. I’m assuming they have a higher resolution.
Anyhow, the package that includes the data and scripts mentioned in this post is available here. There are a few points I want to note:
- I convert from average temperature to Ocean Heat Content simply by multiplying by a factor (specific heat * volume * density) which assumes a constant value for all of those as we increase depth, which is not strictly accurate. However, as we saw in that previous post on OHC, the results are pretty darned close (r^2 = .998 between my NODC calculations and that available from Climate Explorer), and I will be using my calculated values from observations to compare against my calculations from the projections, so it will be more of an apples-to-apples comparison, even if the factor is slightly off.
- I start everything in 2004, because that is when the GISS-ER projections start for the SRESA1B scenario. I have baseline everything to the overlapping period…that is, 1st quarter of 2004 through the 2nd quarter of 2011.
- If you want to personally do a more in-depth conversion from layer temperatures to OHC, I have also included the intermediate calculations of globally-averaged ocean potential temperatures for the top 16 layers (down to 700m).
Anyhow, here is the comparison between the 5 GISS-ER model runs and the observations:
For kicks, I would note that the simple OLS trend for the overlapping period is 0.089 +/- 0.163 J*10^22/year for the observations, compared to the 0.715 +/- .054 J * 10^22/year for the model mean. On the one hand, I’ll note that this is a short period, and there are some other things – such as a prolonged solar minimum – taking place here. On the other hand, I’d like to point out that calculating the smallest trend over ANY 7.5 year period in the runs of the chart above yields the following results:
Run1: 0.673 J * 10^22/year
Run2: 0.714 J * 10^22/year
Run3: 0.583 J * 10^22/year
Run4: 0.555 J * 10^22/year
Run5: 0.633 J * 10^22/year
So, the smallest trend for any 7.5 year period in the 36 years for any of the five runs is 0.555 J*10^22/year, which is clearly much larger than our currently observed value. I’ll need to take a look at Meehl et. al 2011, and there are certainly other models, but it seems to me – at least at initial glance – that there is nothing in the GISS-ER runs to suggest that we could potentially see a slow-down in the OHC of the upper 700 meters that we are currently seeing. I’d welcome other opinions.
Troy's Scratchpad 
It’s interesting that the ocean heat content runs are almost perfectly monotonic, in spite of different runs having different internal variability, presumably. This suggests to me that the models depictions of internal variability are not realistic energetically-that is to say, no anomalous build-up stall in the build-up of Joules seems to occur in models, there is virtually only a long term trend. In order to catch up to the models we will need a strong natural warming to occur on top of anthropogenic effects, to counteract a presumably natural brief cooling. Well, either that or the models long term trends will prove wrong.
Comment by timetochooseagain — October 6, 2011 @ 2:51 pm
Yes, GISS-ER does not seem to have much in the way of variability in OHC, which is why it strikes me as a bit surprising that we’ve heard so much about how the recent flattening is not a big deal. I’ll take a look at other models, but as you say, GISS-ER does not seem to realistically simulate this portion. Perhaps other models do better.
Comment by troyca — October 6, 2011 @ 5:37 pm
Troy: Can you divide the Model ER outputs into the ocean basins? And do you also have the GISS Model-ER hindcast data. That is, from the start of the NODC data in 1955, can you isolate the GISS Model runs per ocean basin, North Atlantic, South Atlantic, etc:
1. to see if the models runs pick up the disparity that exists between the North Atlantic OHC and the OHC of the other ocean basins,
2. to see if the models pick up the shift in the late 1980s in the North Pacific OHC (north of 20N),
3. to show that the models miss the upward shifts in the tropical Pacific OHC caused by the multiyear La Niña events. (We already know the GISS Model ER does not model ENSO.)
Comment by Bob Tisdale — October 6, 2011 @ 5:23 pm
Hi Bob,
What you describe might be a bit of an undertaking. I can get the hindcast data easy enough (just need to wait for the download, which at ~1.2 GB per run can be a while), but isolating ocean basins may be a bit more difficult. If there are basin masks that use the same resolution as GISS-ER I could probably do it quick enough, but otherwise there may be some messy mapping…I haven’t really looked into it. I’m currently using all my bandwidth downloading the other ensemble members.
If you have time, I would recommend signing up with PCMDI, and then you can grab the data. It comes in NetCDF format, for which there is a package in R, so you can get up and reading it pretty quick (I’ve included a script in that package that shows the initial read).
Comment by troyca — October 6, 2011 @ 5:45 pm
Thanks for the suggested direction. I’ll look into it.
Comment by Bob Tisdale — October 6, 2011 @ 6:23 pm
troyca: Do you have the hindcast of the Model-ER OHC data? If so, can you compare it to the NODC OHC data for the period of 1955 to 2003, with the base years for anomalies set at 1955 to 2003? In effect, I’m asking if you can confirm which of the following curves is the actual NODC OHC data?
It’s from this post:
Comment by Bob Tisdale — October 10, 2011 @ 2:00 am
Correction: I wrote above, In effect, I’m asking if you can confirm which of the following curves is the actual NODC OHC data?
It should read, In effect, I’m asking if you can confirm which of the following curves is actually the GISS Model-ER data?
I need another cup of coffee!
Comment by Bob Tisdale — October 10, 2011 @ 2:02 am
I’ll check it out. The GISS-ER 20th century experiment has 9 runs, so I’ll need to download and process them all before I can get the average. Thankfully, they are lower resolution than the other models, and take substantially less time to download. Hopefully I’ll have an answer soon enough, and I’ll put up all the run data in addition to the mean.
Comment by troyca — October 10, 2011 @ 9:13 am
[…] I mentioned in my last post, the GISS-ER projections for ocean heat content did not seem to include any period of flattening […]
Pingback by ECHAM-MPI model runs, Ocean Heat Content, and Katsman and van Oldenborgh 2011 « Troy's Scratchpad — October 11, 2011 @ 9:31 am
[…] my first post on GISS-ER Ocean Heat Content projections, Bob Tisdale asked me if I could show the hindcast data […]
Pingback by GISS-ER Ocean Heat Content, 20th Century Experiment and A1B projections « Troy's Scratchpad — October 13, 2011 @ 12:46 pm