Watt about the retrospective predictions?

There is a recent post on Watts Up With That (WUWT) called in retrospect we predicted global warming would slow. It’s about a recent paper by Guemas, Doblas-Reyes, Andreu-Burillo & Asif (Retrospective prediction of the global warming slowdown in the past decade, Nature, 3, 2013). I have a feeling that WUWT has already discussed this paper, so maybe they’re running out of things to say.

Anyway, the basic implication of the post and the gist of most comments is that the term “retrospective prediction” is ridiculous. In other words, it’s easy to predict something if you already know the answer. Let’s think about this a little more and consider it from a scientific perspective. I develop a model that I use to predict the future evolution of something. However, I later find that my prediction isn’t quite what happened. What do I do? Do I throw up my hands and say, “Oh well, I was wrong. I give up. Someone else’s turn.” No, I go back and try to work out what was wrong with my model. Let me stop for a moment though. The issue in this case is that the observed global surface temperatures have risen more slowly than would have been expected based on the model predictions. They are, however, within the range of most models. The models aren’t absolutely wrong. It’s just that the model predictions of the most likely trend is higher than that observed.

Okay, let’s say I start to consider why my model results differ from what’s observed. Normally one would add things that previous models ignored. What do you do then? Well, you ask the question “what would my model have predicted had I included this originally?”. What would I call such a test? Well, I might consider calling it a “retrospective prediction”. Seems like a reasonable term to me; what – in retrospect – would my model have predicted. To me, mocking papers such as this makes those on WUWT seem infantile and ignorant. To go back and work out why the prediction of a set of models differed from what was observed is precisely what science is about. Not doing so would be unscientific.

Let me, however, highlight one paragraph from the abstract of the paper

Here we show successful retrospective predictions of this warming slowdown up to 5 years ahead, the analysis of which allows us to attribute the onset of this slowdown to an increase in ocean heat uptake. Sensitivity experiments accounting only for the external radiative forcings do not reproduce the slowdown. The top-of-atmosphere net energy input remained in the [0.5–1] W m−2 interval during the past decade, which is successfully captured by our predictions.

This is a very important paragraph. It makes the claim that the most reasonable explanation for the slowdown is an increase in ocean heat uptake. Furthermore, their model correctly determines the top-of-atmosphere net energy input. This is crucial. Global warming is fundamentally about the increase in energy in the climate system and is not only about the increase in global surface temperatures. If their model is getting this correct, then that means that it is correctly determining the level of global warming as well as correctly “retrospectively” predicting the rise in global surface temperatures.

As a consequence of what I said above, I was going to be critical of the WUWT post for simplistically associating a slower rise in surface temperatures with a slowdown in global warming. Why? Because there is evidence that global warming has not slowed down; it is simply that a larger fraction of the excess energy is going into the oceans and hence less is available to heat the surface. However, the title of the Guedas et al. (2013) paper is “Retrospective prediction of the global warming slowdown in the past decade”. I think that this is a very unfortunate choice of title and actually adds to the confusion. In a sense its contradictory. The slower than expected rise in global surface temperatures is not – in itself – an indicator of a slowdown in global warming and the main result of this paper is to show exactly that. The slower than expected rise in global surface temperatures is not because there has been a slowdown in global warming, it’s because more energy than expected has gone into the deep ocean. My main criticism of the paper is, therefore, that they could have been more careful in their choice of title.

Advertisements
This entry was posted in Anthony Watts, Climate change, Global warming, Watts Up With That and tagged , , , , , . Bookmark the permalink.

9 Responses to Watt about the retrospective predictions?

  1. Lars Karlsson says:

    So the retrospect prediction effectively provides an explanation to the slowdown of the surface temperature increase.
    One thing I have learned about “sceptics” is that they are extremely uninterested in explanations of that slowdown. They just like to say “slowdown” or “pause” or “hiatus” or whatever. They simply love those words. But as soon as you start mentioning ocean uptake or solar minimium or ENSO, they cover their eyes.

  2. That’s certainly what I took from my reading of their paper. I don’t know quite what changes they implemented in their model, but it seems that their model can now match the energy imbalance and get the “correct” rise in surface temperatures. I agree with the latter part of your comment. Any attempt to explain the “slowdown” is seen by skeptics as some kind of cheat – it’s as if scientists are simply meant to accept that their models have failed and that they are now meant to accept that those who questioned their models were correct all the time. Quite what they’re meant to do once they’ve acknowledged this is unclear.

  3. Retrospective predictions are called hindcasts in weather prediction. Before you introduce a new NWP model version, you always perform hindcasts.

    Now that science is trying to get decadal climate prediction work we will see more of such studies. See for example the MiKlip research program in Germany. In decadal climate prediction, you aim to predict the change in the climate for the coming years starting with the current state of the climate (normal climate runs are so-called free runs and start in a random state). To validate how good these new model systems work, climatology will also make more and more hindcasts. It would slow down the work a bit too much to wait 10 years each time you change a few lines of code.

    The “sceptics” will have to get used to retrospective predictions or we will have to get used to their infantile and ignorant mocking rants.

  4. Interesting, thanks. I can’t see the skeptics on WUWT getting used to this. From what I can tell, they see hindcasting as equivalent to cheating. This has been one of the reasons why it seems clear that most are not scientifically trained and – if they are – they certainly seem to have no real experience of scientific modelling.

  5. They will complain, whatever you do.

    If you validate your model with hindcasting they will say it is cheating, if you do not do so they will complain that the model is not validated.

    If you do some data processing they will say you should use the raw data, if you use raw data, they will say the data has problems.

    If you analyse a large dataset they will say you should go out in the field, if you go out in the field they will say the dataset is too small.

    This could be a very, very long comment, but I will stop here.

    It is of no use to try to pacify these unreasonable people. They will complain as long as the science says the climate is changing.

  6. Exactly, that does indeed seem to be the case. One concern with writing this blog in the way that I am is that it might make it seem that WUWT is a site worthy of consideration, rather than something that should just be ignored. Having seen what they typically write, it should really just be ignored (IMO). The problem is that it seems to have credibility with some (although your posts seem to show that it’s readership is dropping). What I think I find most concerning is that there are indeed policy makers who seem willing to accept what is said on this site (and other related sites) when it seems clear that their views are so blatantly biased against anything that even remotely suggests that we should be concerned about AGW.

  7. I try to link to WUWT as little as possible, if there is another blog discussing WUWT, I prefer to link to that one. Like Wikipedia, secondary sources with interpretation are preferred over primary sources, which need interpretation.

    A problem is that policy makers typically do not know much about science. On the other hand they are typically highly intelligent; it is not that easy to get to the top. Thus in most cases, I think they just use the “sceptics” as an excuse, while they do know better.

  8. Yes, you may well be right about policy makers. I find that slightly more concerning though. If they do actually “believe” what is said on WUWT has merit, then there’s a chance one could convince them otherwise. If they know it’s nonsense but use it because it suits their ideology, it’s harder to see how one can overcome that.

  9. If a larger part of their voters is convinced, it starts to get harder to listen to ideology or funders. This blog is very useful for that.

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