Near-term regional climate: the range of possibilities

What are the possible regional temperature trends over the coming few decades? Globally, on average, there is expected to be a long-term warming, but this is not necessarily true for any particular location or period. What are the probabilities of a local warming or cooling?

The temperature over the next few decades depends on any forced response to greenhouse gases etc, but also on the particular chaotic variability experienced. To quantify the range of possibilities Deser et al. ran a large ensemble of GCM simulations for the next 50 years, with only small changes to the atmospheric initial conditions. The range of outcomes for North America and Europe was perhaps surprisingly large, with most locations having the possibility of little warming (and sometimes cooling) and also very rapid warming.

Similar simulations have now been done with the FAMOUS GCM, using a 1%/year increase in CO2 concentrations for the next 90 years. Two parallel ensembles have been performed: MICRO — 96 ensemble members with only a tiny random initial perturbation to a single grid point, and MACRO — 30 ensemble members with varying ocean & atmospheric initial conditions. As an example of what can be done with such simulations, there is a global cooling trend of 24 years in one of these ensemble members.

The regional results can be explored in the applet below, for different seasons, regions and trend lengths (N). The maps show mean trends for the first N years for the two ensembles (top row), and also the extreme possibilities for each grid point separately (middle and bottom rows). The right column shows the regional average properties using time-series (with red & blue lines showing the maximum & minimum trends) and histograms. To be clear – the only difference between the individual ensemble members is the initial conditions. Lots of fascinating features – would be interested in hear what readers think!

UPDATE (23/08/13): Additional images added which show the spatial map corresponding to the regional average maximum and minimum trend (grid-point not independent option).

UPDATE (11/09/13): Additional North Africa region added.

UPDATE (02/12/13): Results for precipitation added.

UPDATE (02/01/14): Results for Australasia added.

[Note that the variability in FAMOUS is probably too large, so the ranges are likely to be too broad if considering what this means for the real world.]

Trend length:yrs Slide to change trend length:
Season:   DJF:  MAM:  JJA:   SON:   ANNUAL:
Region:   GLOBAL:   EUROPE:   TROPICS:   N. AMERICA:   N. AFRICA:   AUSTRALASIA:
Grid-point independent:   YES:   NO:     Variable:   TEMP.:   PRECIP:
Selected plot unavailable

About Ed Hawkins

Ed Hawkins (twitter: @ed_hawkins) is a climate scientist in NCAS-Climate at the Department of Meteorology, University of Reading. His research interests are in decadal variability and predictability of climate, especially in the Atlantic region, and in quantifying the different sources of uncertainty in climate predictions and impacts. Ed is a Contributing Author to IPCC AR5 and a member of the CLIVAR Scientific Steering Group.
This entry was posted in GCMs, projections, temperature, uncertainty, variability, visualisation. Bookmark the permalink.

12 Responses to Near-term regional climate: the range of possibilities

  1. In the top right panel, the blue line is the minimum trend found in the ensemble and the red line the maximum trend?

  2. John Russell (Twitter@JohnRussell40) says:

    Thanks, Ed.

    This is a really useful visual aid and I’ll be drawing people’s attention to it a lot in the coming months. Great for showing ‘Sceptics’ how their much hyped ‘pauses’ in warming are exactly what can be expected.

  3. Paul Matthews says:

    The question has to be asked again:
    Why are you using a model that, as you admit, has unrealistically large variations?

    • Ed Hawkins says:

      Firstly, FAMOUS is one of the only models that can be used for this exercise because it is relatively coarse resolution. The idea is to demonstrate that sampling ocean & atmospheric initial conditions is essential if we are to provide a range of probabilities for near-term climate. These large ensemble experiments are not done enough with the more complex models, and we are trying to change that! Although each ensemble size would be much smaller with the state-of-the-art models, there are many more models to sample and bring together.
      cheers,
      Ed.

      • lucia says:

        These large ensemble experiments are not done enough with the more complex models, and we are trying to change that!

        Some models in upcoming AR5 have only 1 run. If an AOGCM is worth running, or worth comparing to data or even other models, for any long term projections, 2 runs should be a minimal requirement as it permits some estimate of variability due to initial conditions.

        I agree sampling over runs is important. But a discussion of the likelihood of “N” year trends of 0C/dec based on a model with known too high variability is misleading. That may be the likelihood in that model. But it tells us little about what people argue about at blogs which is: What is the likelihood on earth? Moreover, people want to know: What is the likelihood conditioned on the claim that the “mean” trend over all possible IC’s for the earth is the value claimed by some authoritative body (e.g. IPCC) or even the value claimed by ‘others’ who are suggesting policy responses based on the full ensemble of models. (These people generally do consider the multi-model mean and often insist that we ‘must’ consider the upper extremes plausible.)

        While it might be interesting to see how IC’s matter (which I never doubted– to I think Shub on Twitter has some issues with that), I can’t take any claim that 24 year 0C/dec trends as ‘statistically plausible” for earth seriously if even you know that model has too high variability!

  4. Ed Hawkins says:

    Hi Lucia,
    I hope I have been careful enough to caveat these results because of the variability. As I have emphasised, the role of different types of initial condition perturbation is what is being explored. I have not claimed that this is what should be expected in the real world. They are fun experiments to explore!
    cheers,
    Ed.

  5. Will Thurston says:

    Hi Ed,

    You may be interested in trend analyses from a 10,000 yr integration of CSIRO Mk. 2 under ‘present’ conditions found here:

    http://link.springer.com/article/10.1007/s00382-005-0102-8
    http://link.springer.com/article/10.1007/s00382-006-0153-5
    http://link.springer.com/article/10.1007/s00382-010-0799-x

  6. Pingback: Another Week of Climate Instability News, August 11, 2013 – A Few Things Ill Considered

  7. Foxgoose says:

    Ed

    You say to Lucia – “I have not claimed that this is what should be expected in the real world. They are fun experiments to explore!”

    But activist John Russell, above, seizes on your results to proclaim – “Great for showing ‘Sceptics’ how their much hyped ‘pauses’ in warming are exactly what can be expected.”

    A lovely little cameo of how activism feeds off and distorts real science in the climate game.

  8. Chris Ho-Stuart says:

    Hi — I’ve only recently discovered this blog, and this tool.

    Can you please explain the Grid-point independent option? What’s the difference between YES and NO?

    • Ed Hawkins says:

      Hi Chris,

      Thanks!

      The grid independent option shows the min/max trends for each grid point, independent of which simulation those trends appear in. So, neighbouring grid point trends could come from different simulations. The other version shows the trends at each grid point in the single simulation which shows the max/min trend for the regional average. Hope this helps?

      cheers,
      Ed.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>