The second climate forecast for the next decade has been published [Advancing decadal-scale climate prediction in the North Atlantic sector, Keenlyside et al, Nature, behind a firewall but available here], and the world’s media – and a fair number of blogs – have jumped all over its suggestion that there might be some regional cooling over the next decade. Richard Black at the BBC headlined his piece “Next decade ‘may see no warming'” , the New York Times‘ Andy Revkin settled for “In a New Climate Model, Short-Term Cooling in a Warmer World”, which becomes “Next decade may see no warming” at frogblog and “Global Warming on hold until 2015 claim Germans” at Kiwiblog. So what’s going on? Is global warming really on hold?
The truth is a little more mundane. There’s a lot of work being done to improve climate models, in particular to allow them to make useful predictions of climate in the relatively near future – over the next decade or two – and to do so at a regional scale. This sort of climate forecast is directly policy relevant because it can provide information on what might happen in a particular place, and give us some time to prepare for it. In other words, this kind of modelling will give us an idea of what we might have to adapt to. It’s an important focus of current research, and will form a key part of the work planned for the next IPCC report. The current generation of global circulation models (GCMs) are pretty good at capturing the large scale features of the climate system and how they might change over the long term, but are not designed for and cannot do “climate forecasts”. [There’s a good discussion of what models can and can’t do here (click on “expand all”), which is being linked to by crank sites as somehow suggesting this is news]. GCMs give us a broad brush picture of likely futures.
At the other end of the business, meteorologists start with a snapshot of the current weather and plug that into their computers. The forecasting software – a complex mathematical model based on the equations that describe how the atmosphere behaves – calculates how atmospheric conditions will evolve over the next few hours and days, and produces a sequence of the charts we get to see in the media. Close to the time the forecasts are made, the predictions are generally pretty accurate, but as they get further out into the future – beyond two or three days – they become steadily more and more prone to error. Small errors in the initial picture of the atmosphere get multiplied as the forecast software chugs away, and the accuracy of the forecast decreases.
The “initial conditions” problem limits the usefulness of traditional weather forecasting models to periods of about two weeks. If you want to produce forecasts for a few months ahead – a seasonal forecast – you have to change the way you build the forecast, and take more factors into account. The most important of these is the behaviour of the oceans, because they hold a lot of heat. They fuel a lot of the weather we experience. If a pool of cold water turns up off the east coast of New Zealand, as it did in late 2006, the weather tends to be cooler than average (remember the icebergs?). So seasonal forecasts have to take into account what’s happening in the oceans. And they don’t produce weather maps. They give us a probabilistic view of what might happen. NIWA’s seasonal forecast for the next three months, for instance, suggests that it will be warmer than average in Northland. It doesn’t try to predict what will happen, and when, because it can’t.
The main features of climate tend to be described over periods of decades – thirty years is the World Meteorological Organisation standard. GCMs are trying to describe future climate states, based on changes in the “forcings” being applied to the system – the increase in greenhouse gases, reduction in ice albedo, and so on. They generate “weather” internally, but that weather isn’t based on a picture of current conditions. It’s realistic, but not real. The results are averages and trends, not specific events, and the drivers are changes in the forcings, specifically the increase in greenhouse gas concentrations. The best models generate an El NiÃ±o/La NIÃ±a cycle, but they don’t forecast when real La NiÃ±as might happen.
If you want to try use a GCM to make a “forecast” of how the climate system will respond in the near future, you have to first feed in the “initial conditions”, and the most important of these on seasonal and annual timescales is the state of the ocean. This new paper is the second to attempt to run a climate model with information based on actual measurements of ocean temperature and heat content. The first, last year, came from the UK’s Hadley Centre. Both papers – and both forecasts – are very much first attempts. This is state of the art stuff. We have no idea how “good” these forecasts are.
It’s also important to note that although we’re learning a lot more about the heat content of the oceans, from the ARGO float programme, from satellite measurements of surface temperature and many other studies, there is still a great deal more to learn. Modelling how the oceans work and how currents ship heat around the globe is a long way behind where we are with the atmosphere – and good ocean models will be crucial to useful medium term regional climate projections.
So has global warming been put on hold for the next few years? This new paper suggests that some regions might cool, but the overall average will continue to climb. It’s possible that the rate of warming might slow down for a while due to the way oceans behave, but we have no idea how likely that might be. Stoat thinks that the furore is mainly a press phenomenon, with an agressive press release prompting some enthusiastic over-interpretation. James Annan is generally dubious, and RealClimate has a post in preparation. [Edited to add: Climate Progress also has a good post on the subject, and Nature Reports gives a good overview.]
Meanwhile, the Arctic sea ice is melting, and until the models can reflect that in today’s conditions, I will remain agnostic on their abilities on decadal timescales and at regional levels.