I previously discussed the net TOA radiative imbalance (rtmt) in models, and it seems pretty clear now that there is some flux adjustment made to the TOA radiation in these models. This makes it a bit trickier to compare this variable alone to recent papers measuring the imbalance, as we actually would need to take into account the evolution since pre-industrial times, perhaps zeroing this imbalance in 1880 a la Hansen et. al (2005).
Nevertheless, I was interested to find that Climate Explorer now contains radiation fields for the CMIP5 models, which makes it a bit easier to perform the global averaging, rather than manually doing it from PCMDI files. I have chosen rcp45, which appears to be the “central” scenario, although for this comparison it shouldn’t make much difference what scenario we choose because I only use the very early portion.
The dashed lines represents TOA imbalance estimates from Loeb et. al (2012) [L12], von Schuckmann and Le Traon (2011) [SL11], and Douglass and Knox (2012) [DK12]. The SL11 estimate has been multiplied by 0.7 to adjust for the percent of surface area covered by ocean. I also include an estimate of TOA imbalance from the new NOAA OHC data down to 2000 meters.
Here’s a look at the 5 and 10 year averages (click to enlarge):
As you can see, there is a lot of variability in the 5-year averages for the actual observations, although this is likely the result of the limited coverage of ocean floats down to those depths up until recently. The other thing that is curious is the amount of models with an apparent negative or neutral net imbalance up until ~1970. This may be a consequence of my zeroing at 1880 (although I’m not sure how else to corral these, as their absolute values are all over the place). I”m inclined to think that there was likely a positive downward imbalance in the early 20th century that led to the increase in temperature (and hence an imbalance even in 1880), but so far I haven’t looked into this.
I should also caveat that the DK12 paper only uses OHC down to 700 meters, which seems to lead to an underestimate in the TOA imbalance. Given the increasing amount of ARGO floats in recent times, I think those estimates using values down to 2000 meters probably yield the more correct result.
Also, part of my interest in performing this analysis was to compare the TOA imbalance in models vs. observations. Thankfully, through Dr. Allan’s website (a co-author on the L12 paper), I was able to access the supplement for the L12 paper. The last page (figure s3) has a comparison that I was interested in between their analysis and the CMIP3 models:
I was able to get a similar imbalance for ukmo_hadcm3 using thetaO values down to about 3000m from the A1B model run, so I’m thinking that is the method used rather than looking at the raw rtmt values. Several models seem to show a significantly higher imbalance than the Loeb et al estimate. Using a lower estimate (e.g. 0.38 W/m^2/K) would probably place the “likely” line below a few more models, but not significantly so. It will be interesting to see how the TOA imbalance evolves in upcoming years.
Data and code for this post can be retrieved here.