Quantcast
Channel: Blog Posts – Geophysical Fluid Dynamics Laboratory
Viewing all articles
Browse latest Browse all 50

27. Estimating TCR from recent warming

$
0
0

GISTEMP annual mean surface temperatures (degrees C)
for the Northern and Southern Hemispheres.

Here’s an argument that suggests to me that the transient climate response (TCR) is unlikely to be larger than about 1.8C.  This is roughly the median of the TCR’s from the CMIP3 model archive, implying that this ensemble of models is, on average, overestimating TCR

Formally, we define the TCR of a model by increasing the CO2 at the rate of 1%/year and looking at the global mean surface warming at the time of doubling. I have discussed the relevance of the TCR for attribution of 20th century warming and for warming scenarios over the next century in several earlier posts (3,4,6).  Gregory and Forster (2008) - GF08 -  is a good reference on this topic.  The discussion below assumes that, for the time scales of relevance here, the forcing and response are more or less proportional with negligible time lag  (i.e. F = \beta T +N were  F is the forcing and N the ocean heat uptake, but N = \gamma T, so T = \xi F where \xi = 1/(\beta +\gamma)).  TCR is then obtained by multiplying \xi by the forcing for CO2 doubling.   TCR is smaller than the equilibrium response to CO2 doubling (the climate sensitivity) because of the effects of heat uptake — but note also the complication discussed in post #5:  the strength of the radiative restoring can change (it typically decreases in models) as the deep ocean equilibrates to a change in forcing.  I won’t discuss equilibrium sensitivity further here.

The figure at the top of the post shows the time series of surface temperature averaged over the two hemispheres, from GISTEMP.  The Southern Hemisphere (SH) has warmed relatively steadily over the past century, while the Northern Hemisphere (NH) warmed more rapidly in the first part of the century and from 1970-2000, with the familiar cooling episode in between.  I expect the response to the WMGG’s to be roughly separable in space and time: A(t)B(x,y).  One might conceivably see a slow drift in the pattern of the response, and some changes in structure near the sea ice edge, but it is hard to see how multi-decadal swings in the spatial structure could emerge from the forced response to WMGGs.  GCM simulations are consistent with this expectation.

So this structure could be due either to the response to other forcing agents, aerosols in particular, or to internal variability.  The major source of internal variability on these time scales is thought to be the pole-to-pole overturning circulation in the Atlantic ocean.  Variations in the strength of this circulation alter the temperature difference between the hemispheres.   In models, the mean NH temperature is a lot more responsive than the SH to this variability, so a stronger than average overturning warms the NH more than it cools the SH, resulting in a global mean warming and providing a consistent picture for the relatively steady SH trend.  On the other hand, aerosol forcing is predominately located in the Northern Hemisphere, also providing a natural explanation for the relative shape of these curves to the extent that the time variation of the forcing matches features in the observed NH temperature variations.

It is important to sort out whether the non-WMGG forced response or internal variability is dominant in this regard, or if they both contribute substantially.  But here I want to see what this plot implies about the TCR, irrespective of which of these sources is dominant. To do this, I am going to focus on the latest period, since 1980 or so, in which the rate of NH warming has been unusually large compared to that in the SH. Focusing on this most recent 30 year period has advantages because it is the satellite era, so we have more observations of things like total solar irradiance that help us reduce the mechanisms that we need to consider.

GF08 discusses the estimate of TCR that one obtains by making the simple assumption that neither internal variability nor aerosols affect the trend over the period since 1980.  The WMGG forcing from 1980 to 2010 is 1.1 W/m2 using standard expressions (see this NOAA site), and is fairly linear in time.  With a warming of 0.5K in global mean temperature T, this would require a value of \xi (in T = \xi F) of about 0.45C/(Wm-2).  GF08 remove volcano years before regressing T against F, and one could also remove ENSO as do Lean and Rind (2009) and Foster and Rahmstorf (2011), in order to reduce the scatter before estimating \xi, but this doesn’t change the overall trend in temperature much and so doesn’t change the central estimate of \xi.   A value of \xi = 0.45, multiplied by the standard CO2-doubling forcing of 3.7 W/m2, gives a value of about 1.8C for TCR.

GF08 use an estimate of the internal variability in 30-year trends (obtained from a GCM) to expand the uncertainty in this estimate beyond that coming from the regression itself; they assume that the system is equally likely to have been in a warming phase of multidecadal variability as a cooling phase over this period, so their uncertainty range remains centered around the TCR value of 1.8C.

But the rapid warming of the NH with respect to the SH over this 30 year period requires an explanation other than WMGGs.  One possible explanation is that aerosol forcing has decreased over this period, enhancing NH warming.  But if that is the case, the aerosol reduction is providing some of the global mean warming as well, so the total WMGG+aerosol forcing over this period would be enhanced, reducing the value of \xi.  If instead internal variability is the culprit in the large recent differential warming of the hemispheres, we reach the same conclusion — this variability would have contributed not just to the differential warming but to the global mean warming (see, for example, Knight et al (2005) or Zhang et al (2007)), requiring us to lower our estimate of \xi as before.

By how much should we lower this estimate?  You need to quantify and distinguish between the aerosol and internal variability sources to go much further. My personal best estimate is currently about 1.4C — I won’t try to justify this further here, but it is close to the central estimate for TCR in the recent paper by Gillett et al (2012).

A TCR of 1.4K corresponds to a value of \xi \approx  0.38K/Wm-2 and 1/\xi \approx 2.65 Wm-2/C.  Assuming a typical GCM heat uptake efficiency, \gamma \approx 0.7 W/m2 (I would really like to have a simple theory for this number), this gives a radiative restoring strength of \beta \approx 1.95 Wm-2/C.  This is roughly the value you get from fixed relative humidity models with no cloud feedback (see post #25).  You need some positive cloud feedback or greatly reduced heat uptake to get up to a TCR of 1.8C.  With estimated current WMGG radiative forcing of about 2.8 W/m2, and with a climate resistance of 0.38C/Wm-2, you still need aerosol forcing of about -0.7 W/m2 to get the century-long global warming down to 0.8C.

It’s a simple story, based on a lot of assumptions.  Analysis of GCMs with this argument in mind might help focus attention on aspects of model simulations that constrain TCR — or it might indicate weaknesses in the argument,  allowing models to be consistent with the recent rate of warming in both hemispheres while simultaneously possessing a TCR larger than 1.8C.

[The views expressed on this blog are in no sense official positions of the Geophysical Fluid Dynamics Laboratory, the National Oceanic and Atmospheric Administration, or the Department of Commerce.]


Viewing all articles
Browse latest Browse all 50

Latest Images

Trending Articles



Latest Images