There is a gap in climate predictions. It is between the annual and decadal.
I was once told by a very eminent climate scientist that he didn’t care what the observations of the real world were, he believed in models, and only models, and they were enough to work out what is going on. But I wonder?
The basics of climate change are well known. Where the debate lies is in the important details — and the peer-reviewed scientific literature is full of such disagreements, frequently about the same sets of data. This is commonly ignored by the more enthusiastic alarmists.
Empirical data is one thing, but what is needed for developing policies are predictions. Predictions are based on models and their limitations are obviously: what we know about the real world, what we can measure, and how we can transpose them into computer code. Models are not, and never will be, the real world but they are nether-the-less essential and seductive.
There is a gap in climate predictions. It is between the annual and decadal. If scientists could make seamless predictions from weather forecasts over a few days, through seasons to annual and decadal timescales they would be in a far better position to influence policy. But as it is, it almost seems that weather forecasts and short-term seasonal forecasts are a different type of animal than decadal forecasts. Establishing what some call a “seamless climate service delivery system,” has been called one of climate science’s grand challenges.
Some are optimistic that it can be done. According to a recent press release by the University of Exeter such near-term climate predictions are ‘coming of age.’ According to Professor Adam Scaife, jointly from the University of Exeter and Met Office and a lead author on the study in Nature Climate Change.
“There is a lot of work still to do, but just as weather forecasts became a regular operational activity in the 20th century, we are now approaching a similar point for near term climate predictions and these are now being made at a number of scientific institutes worldwide,”
says Professor Adam Scaife, jointly from the University of Exeter and Met Office and a lead author on the study in Nature Climate Change
Having read the paper and supporting literature I do not feel so optimistic. After all, on the one hand the study claims that prospects for success are good, but on the other hand that there are formidable challenges.
We are not yet able to even identify and constrain the key factors that will need to be understood let alone able to determine climate predictability on weekly to decadal scales when we don’t know the uncertainties. Indeed the paper talks of an “envelope of uncertainties” controlled by forced and internal climate variability.
Some believe that near-term predictions have been vindicated by what are called retrospective predictions or hindcasts. Modelers of all disciplines will readily tell you that the past and the future are not the same, for the obvious reason you have to wait for the future. Success in hindcasting is no guarantee of success in forecasting. The only thing that works is to make a prediction and wait to see how good it was.
I know I am not alone in thinking that optimism for near-term climate forecasting belies the difficulties, especially when one reads that it is hoped it will be as good as the seasonal forecasting currently available. You don’t have to have long memories to remember how such forecasts by the UK Met Office got the Government into a pickle after a ‘cold snap.’
Looking at the projects to compare climate models with real-world data shows that climate predictions are difficult, and that they are inaccurate. That is the lesson from the 5th round of the Climate Model Inter-comparison Project (CMIP) . CMIP6 is not very encouraging. It will inform the AR6 report of the IPCC in 2021. It’s just at the start but so far it is repeating the problems of CMIP5 in running too warm.
The design of CMIP6
The basics of climate science are fine, but debates about very important (and predominantly uncertain) factors remain, factors that are essential to understand what is going on and what is to come. As for climate models there are obvious problems and concerns. Disagree? Then just look at how wrong the decadal predictions of the past 20 years or so have been, or how the Met Office needs such enormous error-bars when predicting the global temperature of just next year.