Understanding Climate Risk

Science, policy and decision-making

Step change hypothesis and working paper

with 11 comments

Imagine you didn’t know anything about climate change and the greenhouse effect but were interested and you know a bit about general science. Would you accept the following story?

“Earth’s climate is a large, complex system, affected by forces that produce both linear and nonlinear responses. Shortwave radiation – basically UV – from the sun comes in and heats up the planet, producing infrared radiation. Some UV gets reflected straight back out by clouds, snow and ice and stuff. The land can heat up quite a lot, but it cools back down again and doesn’t store much. If a forest is cleared and replaced by buildings, it will warm up a bit but the effect is only local.”

“But the ocean – that’s another story. It absorbs a lot of radiation, so is taking up heat all the time. Huge streams of energy are entering and leaving the ocean store each year. Some is ‘dry’ or sensible heat, which is ordinary warmth. Some is ‘wet heat’ or evaporated moisture. Energy gets taken up when the moisture is evaporated and it will be released again when the moisture cools, condenses and then gets rained out. In this way, the oceans provide a lot of heat to the land every year, largely as rainfall and a bit of snow.”

“The atmosphere doesn’t store much energy – but does transport it. In the past 60 years, the atmosphere has added about the same amount of heat as the ground surface, while the ocean has warmed by about 30 times that much. The whole atmosphere holds about the same heat as the top 3 metres of ocean (10 feet, if you’re American). So the land might warm and cool a lot in a year, but it doesn’t carry over much energy from one year to the next. The atmosphere can’t carry much and ice will but is too slow. Only the ocean has enough carry-over energy likely to affect climate on decadal timescales (There are exceptions but they’re short-lived).”

“Greenhouse gases in the atmosphere trap some of the heat produced from the incoming solar radiation. Most of this goes into the ocean and this maintains the atmosphere at temperatures roughly 30 °C (54 °F) warmer than they would otherwise be. The atmosphere actually doesn’t store any heat on its own – it is warmed or cooled by the things around it. That was already said above in another way but is really important.”

“The heat produced and stored in the system is forced in two directions. One is to the top of the atmosphere to escape into space so that the energy out equals the energy in. The other is from the equator to the poles, warm to cool, where it also radiates out into space. The transport from the equator to the poles is seriously nonlinear. The rotation of the Earth and drag of the atmosphere produce vortices that include the coriolis effect, but these exist throughout the ocean and the atmosphere. This creates the environment for complex system interactions, in a similar way that pendulums of different length create patterns that may oscillate between organised waves and chaotic motion.”

“The oscillators in the climate system that have the largest effect on climate are formed from the interactions between the atmosphere and the ocean.  They have been call strange attractors because these systems can flip between different fairly stable states. Strange attractors were first described over 50 years ago by Ed Lorenz (more popularly being called the Butterfly Effect). They are produced by store of energy in motion being perturbed by small changes. The ocean is the store and the atmosphere supplies the perturbation. Normally, there are a number of these oscillations in the climate system and they affect things like regional temperature and rainfall, storminess and so on. They can energetically connect areas that are a long way apart, so that ocean warming in one region can affect rainfall in another.”


“These oscillators can rapidly switch between states and some do so on decadal timescales. The stable states in between shifts are called regimes, and they are involved in the transport of energy vertically to the top of the atmosphere, and horizontally from the equator to the poles. Oscillators of different periodicity combine to perturb these regimes, and these are affected by  internal and external sources of energy. Examples are the El Niño Southern Oscillation that flips between El Niño (heating), normal and La Niña (cooling) every few years, and the Atlantic Meridional Oscillation that flips on decadal timescales. These interact with each other, and act as a type of thermostat, being governed by the energy status of the climate at any one time. Under our current state of knowledge individual flips are not really predictable, but their long-term behaviour follows statistical patterns such as power laws.”

“So what happens when more greenhouse gases are added to the atmosphere? If the atmosphere warms in place, then the airmass would gradually get warmer, forming a trend over time. But what if the added energy follows the same pathway as it normally does and becomes stored in the ocean? The atmosphere would not warm straight away and the heat in the ocean would immediately be entrained into these nonlinear systems. Some is mixed into the deep ocean, but sea surface temperatures remain fairly constant until an instability threshold is reached, and the system needs to do some work. It re-organises and a large amount of heat is released into the atmosphere. When the system settles back, higher temperatures at the sea surface and in the atmosphere are maintained.”

“Upon release, this heat would be rapidly transmitted to adjacent land surfaces. Once in the atmosphere, positive feedbacks then lead to warming that respond to the atmosphere’s inherent climate sensitivity. This process leads to step-like warming  where the warming process looks like a stepladder, over long time scales (50+ years) forming a complex trend.”

“Increasing the rate of change by continuing to add greenhouse gases to the atmosphere will forces these step change closer together and probably more locally distributed, much like a boiling pot. At that stage, rising temperatures will look more like an escalator than a stepladder.”

“Reducing greenhouse gas emissions, then reducing greenhouse gases will slow this process down, making the gaps between step changes long and eventually stabilising the atmosphere. Maybe even inducing steps to cause cooling. Can we do this before large ice sheets near the poles melt?”

I have had no trouble explaining most of this to the ordinary person but the climate science community is another thing entirely. The received wisdom of trend analysis and efforts to defend the trend in the climate wars has meant that to many, trends are rusted on science.

This little narrative is the product of many year’s work. It has resulted in a couple of papers in the literature, one on step changes in south-eastern Australia, and another looking at Valuing Adaptation to Rapid Change. A submission of two key papers on the statistics of step changes in observations and models written with Jim Ricketts was rejected without review late last year by one journal and rejected with prejudice early this year by another.

These were then merged and beefed up with a good deal of science philosophy and some of the theoretical reasoning in the above narrative. It will be submitted into an open review journal in the next week. Hopefully they like it enough to open it up for review. Because of the extensive arguments put and the evidence tested, it’s a long paper. Another five just are about ready to go, most quite lengthy.

The problem with publishing these days is that it is difficult to get such things into the literature. Publishing is geared towards making limited advances in 3,000 words. Arguing against a long-held paradigm is not part of the deal.

For that reason, we have decided to release all of the past three years work as working papers, while the later papers are reformatted and submitted to journals. This is for a few reasons:

  1. If the case is put out piecemeal it will be picked apart piecemeal.
  2. If the above story is correct, or close to the mark, then climate is not being analysed appropriately and climate risk is not being characterised properly.
  3. Other work needs doing and this has been burning a hole through the desk for far too long. I might get a life.

The first few working papers are legacy papers that have been rolled into others as the research developed. The first Reconciling anthropogenic climate change and variability on decadal timescales was originally written for a journal submission which fell through. It was developed into four more, three of which will be posted soon. Also posted will be the two papers that were rejected, along with another three earmarked for submission.

Some will frown and disapprove that research is being posted without peer review because that’s not real science, is it? However, the frustration of watching a whole heap of rubbish getting posted about trend analysis in climate on one hand and a whole heap of rubbish about how climate is not a trend on the other, is too much to bear. We are in the middle of the next big shift right now. As this El Niño settles back into a more normal pattern we will find the shallow oceans and atmosphere are appreciably warmer than they were between 1997 and 2014. Things might be stable for a few years, who knows?, but another one will be on the way somewhere down the track. Previous step changes were 1968-70 southern hemisphere, 1979-80 global, 1987-88 northern hemisphere and 1997-98 global.

Every one of these step changes takes the world into new territory. So far this time, it’s been the biggest mangrove die-off ever seen in northern Australia, kelp forests of south-west WA killed off by record warm waters, the largest coral bleaching yet witnessed and unprecedented wildfires in North America. And that’s the short of it.

Some of the above narrative may be wrong, maybe all of it – but if it means that climate science takes nonlinearity a little more seriously and treats it as more than just ‘noise’, it will be worthwhile.

This page will be updated as the working papers are linked in.

Finally, doing a project this size means that I have had a lot of help from a number of people, picking up other work, helping with this project and dealing with someone who’s attention is elsewhere. Celeste Young stepped in and got the Valuing Adaptation to Rapid Change project over the line in 2012-2013 when it was dead in the water and is still doing the same for other projects. Jim Ricketts has coded an objective multi-step bivariate model that allows rapid assessment of step changes over large data sets. My work colleagues have had to adjust as other projects have slowed down. Most of the work has been unfunded, so it has received a large amount of in-kind assistance. My Australian climate colleagues and adaptation experts worldwide have also kept faith as we have worked hard to make the case for viewing climate in a very different light to how it is normally seen.

Working Papers

Jones, R.N. (2015) Reconciling anthropogenic climate change and variability on decadal timescales. Climate Change Working Paper No. 31.

This paper was completed in late 2014, so while reasonably comprehensive, has been superseded by subsequent work (which will be posted soon).

Jones, R.N. (2015) Reconciling anthropogenic climate change and variability on decadal timescales: narratives and hypotheses. Climate Change Working Paper No. 32.

This paper was developed from Climate Change Working Paper No. 31 and serves as an introduction to Climate Change Working Papers No. 33 to 36.

Jones, R.N. (2015) Reconciling anthropogenic climate change and variability on decadal timescales: history and philosophy. Climate Change Working Paper No. 33.

Why are we putting straight lines though complex trends? I try to answer this.

Jones, R.N. and Ricketts, J.H. (2015) Analysing steps in global and regional observed air temperature. Climate Change Working Paper No. 34.

Straightforward analysis of step changes in temperature, showing they are spatially organised. This paper has been merged with the following paper.

Jones, R.N. and Ricketts, J.H. (2015) Analysing steps in modelled global surface air temperature. Climate Change Working Paper No. 35.

Climate models do a pretty good job of reproducing historical step changes. Steps and shifts also explain total warming in the 21st century better than internal trends.

Jones, R.N. (2015) Reconciling anthropogenic climate change and variability on decadal timescales: the challenge. Climate Change Working Paper No. 36.

If climate change is episodic rather than gradual, we will need to re-think the way we analyse and communicate it. Check out the cover. It has elephants.

Jones, R.N. and Ricketts, J.H. (2016) The climate wars and “the pause” – are both sides wrong? Climate Change Working Paper No. 37

This paper was borne out of the frustration that both sides of this debate were so totally focused on trends and winning the war that how the climate changes is secondary.

Jones, R.N. and Ricketts, J.H. (2016) Reconciling the signal and noise of atmospheric warming on decadal timescales. Climate Change Working Paper No. 38

This paper does the full-on step and trend severe testing. So which regime reigns supreme?

Jones, R.N. and Ricketts, J.H. (2016) Atmospheric warming 1997–2014: hiatus, pause or regime? Climate Change Working Paper No. 39

This paper looks more closely at the hiatus, pause or regime in warming. No cherries here.

Edited 1/8 to fix the gap in explanation pointed out by mt. 5/8 Working papers 32 to 36 added. 8/8 Working papers 37 to 39 added


11 Responses

Subscribe to comments with RSS.

  1. I was with you until “These exist in metastable states as interactions between the atmosphere and the ocean.” I can’t find an antecedent for “these”, except “vortices” at the end of the prior paragraph. I am not sure what “metastable” means, but I suppose I could accept this claim being about vortices. But if you refer to “vortices”, then the next two statements make no sense: They [vortices] have been call[ed] strange attractors because these systems can flip between different states. They were first described over 50 years ago by Ed Lorenz (more popularly being called the Butterfly Effect).

    A vortex is not a strange attractor in any sense relevant to climate.

    Then in the next paragraph you go on to talk about “these oscillators”. A vortex is not a strange attractor is not an oscillator.

    So once you get into the weeds, I think this gets a bit messed up.

    That said “Some is mixed into the deep ocean, but sea surface temperatures remain fairly constant until an instability threshold is reached, and the system needs to do some work. It re-organises and a large amount of heat is released into the atmosphere. When the system settles back, higher temperatures at the sea surface and in the atmosphere are maintained.” does seem something like what the system is trying to tell us. It’s a bit premature to call this proven. Someone will have to do the nuts and bolts of connecting El Niño to ocean warming to climate change. It seems to me like a hypothesis that’s worth working on. I’m not sure of the state of the literature on this. Have you consulted any physical climatologists?


    August 1, 2016 at 2:22 am

    • Thanks Michael,

      all I meant by vortices was to say there are rotations going on at the most elementary level and was not intending to drill down that far by that statement. In the sense that pendulums form waves, the climate does similar things by producing rotation. The strange attractors are emergent from this system but there are many different types of rotations within it. Just trying to create some word pictures.

      And it’s all hypothetical – but that last part is what we think is going on. Having been wrong many times on the way so far, I’m sure we will be surprised by what comes up next. My colleague Jim Ricketts will be drilling down into where these changes are happening in spatial detail – that will help a lot, but they are tightly tied into the decadal mechanisms.

      Terry O’Kane and colleagues are working on the details of these things. Their article on the 1970’s shift is looking at the dynamics ENSO regimes and the late 1970’s climate shift: The role of synoptic weather and South Pacific ocean spiciness http://www.sciencedirect.com/science/article/pii/S0021999113007341

      The following El Nino event was associated with a step change in many data sets – how some El Nino events are associated with large scale changes and other are not is an interesting question.

      Roger Jones

      August 1, 2016 at 7:52 am

      • And my weeds may be too verdant, but I’m happy to be corrected on turning this into everyday language as much as possible – hopefully fixed somewhat now (1/8)

        Roger Jones

        August 1, 2016 at 11:47 am

  2. Thanks Roger. Very interesting to read it broken down like that. Could you expand a little on the section I’ve copied below – this appears to be the crux of the issue.

    “Some is mixed into the deep ocean, but sea surface temperatures remain fairly constant until an instability threshold is reached, and the system needs to do some work. It re-organises and a large amount of heat is released into the atmosphere.”

    I haven’t come across ‘work’ in that context since first year physics (at which I was not very good), but could you explain a little about the ‘instability thresholds’, and what the ‘system doing work’ and ‘re-organising’ might involve. Keen to understand this better.


    Nick A

    August 2, 2016 at 9:13 am

    • Hi Nick, that’s the point I’m weakest on. I’ll have a go from what various people have speculated in the literature because there’s still a discussion going on about how these systems work.

      The system is actually at an energetically stable state when it’s in a regime. It might seem paradoxical that the stable state is nonlinear circulation patterns but that’s the way it goes. It’s doing its best to maximise the transport of energy to where it is going. In a normal climate these patterns become periodically unstable, but add extra energy that needs to be transported and that instability increases.

      There are a couple of hypotheses about. One is the maximum entropy production principle, where out of a myriad of pathways, the one selected is the one that maximises entropy at the greatest rate within the limits set. That is, in the climate system, energy will be moved from a to b at as great a rate as possible (hope I have that bit right). If the statistical properties of the system change so that a greater rate of entropy production can be achieved, the system will re-organise to do that.

      The strange attractors in the system are various oscillators around which this type of activity is organised and provide the organised chaos, sometimes referred to.

      Our work is largely based on statistical induction where we are seeing step-like changes in the data. The nature and timing of those changes fits well into this type of thinking and not at all well into a mechanistic model of gradational change. The theoretical advances in this area are quite slow and are not advancing to where a community of practice can use that sort of information to manage risk – they exist as simple models and lots of equations in scientific papers. The statistical reasoning provides a partial explanation of why change isn’t gradual and provides a hook for the physics in terms of characterising risk. We are trying to show that the risk of nonlinear change is present and that more work on the why is needed.

      Some papers that explore this are Ozawa et al 2003


      and Ghil

      Roger Jones

      August 2, 2016 at 10:06 am

      • Thanks Roger. I’ll confess I didn’t quite follow all that, but I’m still prepared to accept the proposition that change is not gradual. Intuitively, it makes sense (though I appreciate that your job is to demonstrate it, not rely on intuition). And from a risk-management perspective, it seems fairly obvious that just extrapolating a straight line is not a great approach to climate adaptation.

        Reminds me of a pot of boiling pureed soup (cf. water). Simmering water is constantly releasing little bubbles. Simmering soup seems to do nothing and then one giant bubble belches out.

        Anyway, good luck!

        Nick A

        August 2, 2016 at 2:02 pm

  3. I’m intrigued to find the reference to my statistical testing philosophy in some of these papers. I don’t know enough to capture the point, but I’m guessing it’s something like this: Existing methods might implicitly contain an assumption of gradual as opposed to step-like changes, thereby precluding them from distinguishing between the gradual vs step-like models severely. As a result, the gradual models haven’t passed severe tests. But I realize there’s a lot more. Perhaps the papers are also suggesting that the step-like models do pass rather severe tests. That would mean that if the step like hypothesis was incorrect, then, very probably you wouldn’t find the kinds of patterns you do; thus the analysis indicates genuine departure from gradual models? Moreover, it’s mentioned that a Bayesian analysis starts with low belief in the step hypothesis. This effectively precludes the chance to discover the falsity of gradual models. I’d really like to understand more, possibly to explain the example in my work, or at least in a blog post: errorstatistics.com
    P.S. My colleague, Aris Spanos, is the one who actually knows how to do the data analysis and modeling with temporal trends. He’s the one who alerted me to this work.


    August 26, 2016 at 6:54 am

    • Hmmm, that means maybe it isn’t explained well enough.

      Climate change is a fascinating problem and is a bit different to general frequentist problems that can be addressed through repeated independent sampling. The biggest difference is that there is only one world and one effect at the global scale, so sampling that using frequentist methods requires using models as sample proxies and doing things like segmenting data series, spatially and temporally. Global (macro) economics has a similar issue.

      Depending on the researcher, linearity of the forced signal is either accepted as justified by theory or accommodated as a working assumption using parsimony, because it’s the simplest statistical model and is robust across many applications. That leads to a working assumption that climate risk changes gradually, but for the agnostic leaves the door open for nonlinear processes to be present. We also know from the theory that the change in heat energy trapped in the earth system by increasing greenhouse gases is roughly linear, because of the way those gases gradually increase in the atmosphere.

      What is in dispute, is whether the atmosphere warms gradually in place or whether the heat is being stored elsewhere in the earth system and released in such a way that warming is nonlinear on decadal timescales. Combinations of the two are mechanistically plausible. This sets up substantive criteria for testing.

      The other way is try a whole heap of statistical models for fit and to use inference from their statistical performance to suggest how the system operates. That is pretty much what the Chicago School of Economics did for economics, formalising it as a ‘philosophy’ (e.g., Friedman), where the theory fits the simplest model. Climate science isn’t quite the same, because we have climate models that simulate the physics (and now chemistry and some biology) of the climate, but to understand what those models are saying, we still need to use statistical tools and inferences.

      I have been using your arguments to try and make our testing framework more rigorous and to strengthen (and explain) the probative elements because the tests by themselves were not convincing to enough people – this blew word limits out of the water.

      Yes, we are saying in the paper now under discussion at Earth System Dynamics that the step hypothesis does pass severe tests using highly probed, rather than highly probable criteria. There are six tests that we say could not all pass if the change was trend-like. Because these tests are all different we could not think of a way to combine them into a single metric or to formalise any degree of severity. We would be interested in your views on this.

      I have one paper that looks at the history of gradual and uniformitarian reasoning in climatology and argue that it creates a legacy that treats climate change as a trend as received wisdom – that it is so fundamental that it is rarely questioned unless you’re questioning the core science, which we are not.

      In laying out Htrend vs Hstep, it became apparent that by severe testing criteria, Htrend has not been severely tested. Coincidentally, Aris Spanos sent me some unpublished analyses from a few years ago that showed warming was nonstationary with respect to a simple trend. Other researchers have published similar analyses.

      Our view is that if this is a complex system affecting all of humanity and everything else on the planet, then to separate the anomaly into a simple change variable and put everything else into a box called ‘noise’ is just not good enough if the results are to be used to inform decision making.

      In light of your work on error testing – we have made heaps along the way to get to this point and there are probably many still in our assumptions and workings.

      Roger Jones

      August 26, 2016 at 1:14 pm

      • Roger: I’ll study this, before replying, but I want to be very clear that I do not view frequentist problems as being addressed through repeated independent sampling, or any kind of actual repetitions. But if you do have a severe test, you make use of arguments about the general capabilities of your methods to discriminate claims.


        August 26, 2016 at 10:34 pm

      • I can’t reply to your reply so I’ll put this here – sorry, I didn’t mean to imply that and maybe it’s at cross purposes, but the issue of independent sampling with climate change is tricky and I expect you have come across similar issues elsewhere.

        Part of this is me stumbling around the various meanings and uses of frequentist – there is a tendency to see frequentist reasoning around an effect, and Bayesian around theory (in the climate literature at least).

        Roger Jones

        August 26, 2016 at 10:44 pm

      • and will say – there are a number of resampling strategies going on that fit into the p-hacking strategy, so sorting the wheat from the chaff is not straightforward

        Roger Jones

        August 26, 2016 at 11:26 pm

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s

%d bloggers like this: