Troy's Scratchpad

May 8, 2012

Northern and Southern Hemisphere Warming in Models and GISTEMP

Filed under: Uncategorized — troyca @ 8:31 am

Since I’ve already done the work for this and posted it in a comment at Dr. Held’s blog post, I decided it would be appropriate to post a bit on it here as well (at the very least to supply the data and code). 

The basic subject of that post (at least my takeaway) was figuring out what the spatial structure of recent warming (in particular, Northern Hemisphere vs. Southern Hemisphere from 1980-2010) can tell us about the transient climate response (TCR).  This is something I’d been curious myself about as well, although more with respect to forcing histories: if aerosols were playing a large role in suppressing the overall warming, then I would have expected this suppression to occur primarily in the NH (due to the regional nature of the aerosols), and therefore the SH would experience more warming relative to the NH (although it is complicated by the greater land mass in the North).  But what we see is the opposite: a NH trend that is far higher than the SH trend (about 3.3 times according to GISTEMP). My original thought was that the NH/SH ratio would be correlated with the aerosol forcing across the models, and that I might be able to use that to estimate the actual aerosol forcing we’ve experienced.  However, that is an investigation for a future time, as this forcing data is not easily available at Climate Explorer and I’ll need to perform some bulk downloads from PCMDI.  Instead, based on Dr. Held’s prompt, I downloaded the NH and SH average temperatures for the CMIP3 models from Climate Explorer in an effort to compare the ratio directly to the published TCRs of the models.

First, here’s a comparison of the NH and SH warming in GISS to the CMIP3 multi-model-mean:


The NH trend for MMM is very close to that of GISS (0.24 K/decade compared 0.25 K/decade), but the MMM overestimates the SH trend by a factor of 2.5 (0.17 K/decade compared to 0.07 K/decade).  Still, in general the models show a slightly larger NH than SH trend, which is at least qualitatively consistent with what we see in GISTEMP.   

Now, here are the actual results for the NH/SH ratio (along with the overall warming trend).  The TCR is from AR4 WG1 table 8.2.  I’ve included the min and max values of the NH/SH ratio for those models with multiple runs in order to get some idea of the spread:


It appears that no models (or even any model runs) seem to capture this differential warming.  But what does this tell us about the TCR?

At first glance, there doesn’t seem to be much of a relationship between TCR and this NH/SH ratio (recall, we expect a negative one):


Of course, we wouldn’t expect much of a relationship if we didn’t take into account the TCR vs. overall warming trend as well, which does show something of a relationship:


I ran a multiple regression of TCR against the two variables (NH/SH ratio and overall warming trend).  This at least gives us what we qualitatively expect for TCR – a positive coefficient for overall trend, versus a negative coefficient for  NH/SH warming ratio.  Plugging in the values from GISSTEMP directly into the resulting coefficients indicate a TCR of around 1.4K.  However, the relationship is nowhere near significant, particularly given the high degrees of freedom with only a few (18?) models with TCR available.  Thus, the spread of "uncertainty" in that estimate for TCR is so large as to basically be worthless.  Some better method of combining these values would be needed, or additional data to help constrain the estimate.  Perhaps knowing which of the models include volcanic forcings would help.  Or, maybe the differential warming will simply be a better predictor of the aerosol forcings, which could then be used to constrain TCR (or ECS for that matter).   


  1. Surely the point is if the NH SAT is being (in part) driven by internal dynamical processes rather than external forcing then trying to estimate TCR from it is fatally flawed. At best you can get an upper limit which always has to be accompanied by a caveat that the figure assumes SAT is fully driven by external forcing. There is some research that suggests that assumption is wrong

    Comment by HR — May 9, 2012 @ 5:40 pm

    • This is true, to the extent that the magnitude of NH warming (or the NH/SH) is significantly affected by internal variation, it would hinder a diagnosis of TCR in this manner.

      From the look at that spread, the models do not seem to show a particularly high degree of variation in these quantities over the timescale. This may suggest that there is a still a possible relationship to be found between the anthropogenic aerosols and NH/SH ratio among models, even if it would not apply directly to the “real world”. Perhaps you could get an estimate for the aerosol forcing that is dependent on assumed internal variation.

      Comment by troyca — May 10, 2012 @ 8:36 am

  2. Troy,

    I forgot you had this blog. I asked you on Dr. Held’s blog if you knew how possible it would be to get this data for CMIP5 runs. Nevermind that anyway. You mean historical volcanic forcing right? Plotting the annual average anomaly for each model would answer this, right? I think you see it in your multi-model mean for Pinatubo anyway. Interesting that GISTEMP seems to show a year lag in the SH Pinatubo effect vs the NH whereas the models show no lag vs the NH.

    Comment by BillC — May 10, 2012 @ 7:00 am

    • Hi BillC,

      Yes, Climate Explorer has CMIP5 runs as well, although I would need to search for the TCR in those models (I won’t have much time in the upcoming week to do this either).

      And yes, I agree that plotting individual model runs would likely reveal which include the volcanic forcing. Of course, I think that information is also available somewhere at PCMDI when I have time to dig it out.

      Comment by troyca — May 10, 2012 @ 8:40 am

  3. […] looking at this experiment, I also want to harken back to a previous post on the NH vs. SH rate of warming.  Here is how the two hemispheres warm in this […]

    Pingback by Changes in feedback strength with time: a look at the 1pct2xCO2 GFDL CM2.1 experiment « Troy's Scratchpad — November 7, 2012 @ 7:17 pm

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at

%d bloggers like this: