{"id":333151,"date":"2024-06-17T11:13:29","date_gmt":"2024-06-17T09:13:29","guid":{"rendered":"https:\/\/climatescience.press\/?p=333151"},"modified":"2024-06-17T11:13:32","modified_gmt":"2024-06-17T09:13:32","slug":"gcms-cannot-predict-climate","status":"publish","type":"post","link":"https:\/\/climatescience.press\/?p=333151","title":{"rendered":"GCMs Cannot Predict Climate"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

<\/p>\n\n\n\n

From Watts Up With That?<\/a><\/p>\n\n\n\n

Michael Jonas<\/p>\n\n\n\n

In March, my article \u201cTraffic Lights and Roundabouts \u2013 Why the Climate Models will never work<\/a>\u201d was presented on WUWT. That was a somewhat light-hearted analogy between road traffic and climate, saying in essence that the techniques used in climate models wouldn\u2019t work for road traffic, so why would you trust them to work for climate. The reason for writing it was to give people an argument that could be used in conversation with those whose eyes would glaze over if you tried to talk about the inner workings of climate models.<\/p>\n\n\n\n

Another reason for writing it was that I had been putting off writing a proper critique of climate models, knowing how much work it would be.  Well, comments on \u201cTraffic Lights and Roundabouts\u201d have spurred me into action, and I have now written up a proper analysis, and it has been published \u2013 General Circulation Models cannot predict climate<\/a>.<\/p>\n\n\n\n

The paper is based on Chaos Theory, of course, and two very interesting (to my mind) facts emerged:<\/p>\n\n\n\n

1. Climate is sufficiently complex that its various parts have different \u2018prediction horizons\u2019. A prediction horizon is the length of time beyond which we can no longer accurately forecast a chaotic system\u2019s behaviour. So some parts of a climate model, like hydroclimatic processes (the water cycle) break down very quickly, while other parts, like tropical ocean surface temperatures, can work for quite a long time.<\/p>\n\n\n\n

2. For climate, prediction horizons are nested. If you get past the short term prediction horizon of maybe a few weeks, you hit a new one of maybe a few years. Past that, there is a decadal horizon, then centuries, then millenia, etc. It may even be better to think of prediction horizon as a continuum rather than nested.<\/p>\n\n\n\n

In \u201cTraffic Lights and Roundabouts\u201d, I said that I was not the first person to say that the climate models will not work. In this \u2018GCMs cannot predict climate\u2019 paper, I also recognise that I am not the first person to make many of the points in the paper, and hopefully I have made this clear via references. However, I might be the first to put it all together in a journal. If not, I apologise, I couldn\u2019t find it in the literature.<\/p>\n\n\n\n

In summary, the main point is that the grid-based physical processes and parameteristions in the GCMs cannot predict climate because there is a short prediction horizon for most of what goes on in climate. That is, a tiny error will very quickly increase in size until it has completely swamped the predictions. It has been shown that GCM results can be dramatically improved if a grid-level process is replaced by a higher-level parameterisation (see \u201cseasons\u201d in the paper). My argument is basically that this applies to just about all longer term climate features in the GCMs (I actually think it really is all). In other words, when the physical processes and small-scale parameterisations in the GCMs (I\u2019ll call these their \u201cgrid-level processes\u201d) hit a prediction horizon for a particular feature, the barrier can be overcome by analysing the feature externally and then feeding it back into the model. There is no point at which the model, after being fed with a number of such longer term features, can ever reliably predict any other longer term features, because it necessarily hits a new prediction horizon when it steps outside the areas that it has been given.<\/p>\n\n\n\n

The end result is that the grid-level processes in a GCM cannot predict anything into any kind of longer term future. All longer term features must be analysed externally and then be fed into the GCM if the GCM is to produce reasonable results. But then the grid-level processes in the GCM aren\u2019t predicting anything. If the grid-level processes are still in the GCM, they are now simply \u2018obeying orders\u2019.<\/p>\n\n\n\n

Even longer term features, like ocean oscillations, have their own prediction horizon. Will they speed up or slow down, get stronger or weaker, or even stop for a while \u2013 we don\u2019t know. So there is a limit to how far we can extrapolate them into the future. For example, we are used to the 11-ish year cycle of sunspots, but for several decades within the Maunder Minimum they virtually stopped. Maybe Earthly cycles can do that too. Maybe William Herschel was right, that there really was a causal connection between what we now call the sunspot cycle and wheat prices, it\u2019s just that things changed at the end of the Dalton Minimum. Today\u2019s scientists often claim that William Herschel was wrong, based on the fact that the correlation he observed did not continue, but they do not take into account the fact that the Dalton Minimum did not continue either.<\/p>\n\n\n\n

Many years ago, a well-known climate scientist told me they didn\u2019t know the mechanisms that caused periods like the Medieval Warming Period (MWP) or the Little Ice Age (LIA), so they could not code them into the climate models. My paper says that they can now put in the MWP\/LIA pattern without knowing the mechanisms.<\/p>\n\n\n\n

The paper ends up arguing that a GCM calculates weather at each time step and this is then amalgamated into a final prediction of climate, but a realistic long term climate model would instead calculate climate and then weather would be deduced from the climate.<\/p>\n\n\n\n

The abstract of the paper:<\/p>\n\n\n\n

Abstract<\/strong>        <\/p>\n\n\n\n

This study draws on Chaos Theory to investigate the ability of a General Circulation Model to predict climate. The conclusion is that a General Circulation Model\u2019s grid-level physical processes and parameterisations cannot predict climate beyond maybe a few weeks. If a General Circulation Model is to be used at all, longer term climate features can be analysed externally and fed into the model but they cannot be represented by the model any better than by the external analysis. The external analysis, which is likely to be simpler, has the added advantage that the assumptions that are used, and the uncertainties in the results, are much more likely to be explicitly identified, quantified, and understood. Consequently it would be clear which aspects of the climate are being predicted, and how reliable those predictions are. The longer the timescale is, the less relevant the grid-level physical processes and parameterisations in a General Circulation Model become. Although a General Circulation Model can be made to represent climate over a longer time scale, its grid-level physical processes and parameterisations cannot predict the climate. A General Circulation Model calculates weather at each time step and this is then amalgamated into a final prediction of climate. This process is back to front. A realistic long term climate model would calculate climate and then weather would be deduced from the climate.<\/p>\n\n\n\n

The full paper is here<\/a>.<\/p>\n\n\n\n