 A good question is, what actually is a climate model? So a climate model is a piece of, or the usual use of the word for a climate model is a piece of computer code, a list of instructions to a computer that encapsulate our very best understanding of the way that the atmosphere of our planet and the ocean work in a physical sense. The basic underpinning laws that climate models are built from includes the basic Newton's laws of motion, conservation energy, conservation of mass, basic physical principles that physicists discovered a very long time ago. So all these models share lots of characteristics and of course the physics of the models are very very similar because they're all based on the fundamental same equations of fluid dynamics. It's the same equation that we knew that we created since hundreds of years. And actually you can write down the very fundamental equations actually on a single piece of paper. Solving them is a lot harder and that's actually what a climate model does. All a climate model is is a million lines of computer code running on a really really big computer system. It's the way that climate models work is that they divide the world up into a series of boxes. So it's very like Lego if you like. The model is like if you would construct your world out of little Lego blocks and it's basically the size of the Lego blocks. So you can buy a really expensive Lego Star Wars ship, my son has those. They're very big, expensive takes a poor parent's two days to build them and they have lots of detail or you can buy a little car for three year old which doesn't have that much detail but it's made out of big big blocks. So you can imagine sort of building up Lego and each one of those Lego blocks represents perhaps a box in which the climate model has a value for temperature, it has a value for the amount of air or water within that box, it has a value for how fast the air or water within that box is moving and how much moisture is contained within it if it's the atmosphere. So you can imagine you've got sort of this matrix if you like surrounding the world of these boxes that go up in the atmosphere down in the ocean. Now the model can't actually tell you anything about the climate on a scale that is smaller than one of these grid boxes. The very highest resolution models are perhaps the order tens of kilometers but most models you're talking hundreds of kilometers so there are lots of processes in the atmosphere in particular that actually occur in reality on a much smaller scale than that for example clouds themselves are much smaller than the size of one of these grid boxes and so we have to make approximations to how some of those processes work. You build parameterizations or representations of processes which need to be resolved at scales that we can't explicitly model in the climate models, things like clouds, things like convection in the atmosphere, things like eddies in the ocean, things like land surface processes. The first ever weather forecast that was carried out on a computer that I'm aware of was carried out by a guy called Charney and actually, well he did a 24 hour weather forecast, actually it took him 24 hours to do that forecast so it wasn't particularly useful but it turned out that if you actually, there's some archive photography of what that machine actually looked like and actually it looks very similar to a modern day supercomputer, it's about the size of a room, it's got lots of leads everywhere and it's got a few people looking around, technicians looking after it, it actually looks very similar to a modern day supercomputer but actually if you work it out it turns out that in fact the amount of computer power in that first supercomputer that did that first weather forecast in the 70s, your mobile phone is probably about 30,000 times more powerful than that supercomputer and a modern day supercomputer is about 30,000 times more powerful than your mobile phone so there are many orders of magnitude, that gives you a flavour of how supercomputing has moved on from the 70s just today, so just in 40 years or so. There's probably no parts of a climate model, of a modern climate model that still reflects what was done in the 70s, almost everything's been rebuilt or rewritten I think, the resolutions increased from something around 700 by 500 kilometres to 100 by 100 kilometres, the detail in the vertical has increased dramatically, oceans have been properly and fully coupled, the land surface has been completely revised to incorporate a whole suite of processes, sea ice models have improved dramatically, cloud parameterizations have improved, we've resolved most of the water vapour feedback problem, it's like asking what's the relationship between a Formula One Grand Prix car in 2014 compared to 1970 and the answer is there probably isn't a single widget that's shared. The resolution is getting higher, in other words the boxers are getting smaller and smaller but the computer power is also increasing and so we're able to simulate sort of the same amount of time if you like with one of these models. So one of our grand challenges in climate models is to dramatically improve the spatial detail that climate models use and that's really a computational problem, we just need bigger supercomputers to really resolve the detail of those things. The things that climate models struggle to capture well would include some extreme events, they struggle with the location of the storm tracks, they struggle with the detail of cloud fields, they struggle with some major challenges we don't really represent at all, the processes which might trigger abrupt climate change so methane release or permafrost melt. A climate model is never going to be able to completely reproduce the weather of the last 200 years however when you average together all these weather events what you end up with is climate and what we think is if we also run lots of climate model simulations as well as and compare the average of those with the average of many years for example of observed weather we can get quite a good comparison between a climate of a model and the climate of the real world and compare those and so that is a fundamental test that we can use to test our climate models. If you ask what do the climate models struggle to represent in terms of the simulation of whether it will warm for a doubling of CO2 my answer would be nothing at all because they do that really well. If you ask what the climate models struggle to predict at the scale of a region and its response to doubling of CO2 in terms of rainfall lots of things they don't get the details of the clouds, the convection, the rainfall processes, the detailed synoptics blocking a whole range of things because the spatial resolution that we use for climate models is probably too coarse to capture a lot of those kinds of key phenomena. There's a whole variety of reasons why we're confident in the skill of climate models for the problems that they were designed for. First of all they're built upon physical principles and those physical principles are known unless of course Newton was stupid which I don't think he was. So we have basic fundamental theory not sourced from climate science but sourced from basic chemistry, basic physics, basic biology and applied mathematics that says the core of climate models is sand. Secondly of course they use routinely in other applications like weather forecasting and so we effectively can evaluate a lot of our science routinely and weather forecasting is becoming increasingly accurate irrespective of what some of your listeners might think. If they actually write a diary of a five-day forecast and then check off how those five-day forecasts evolved they'll find that they are shockingly accurate nowadays. So that's the second test. Thirdly we routinely test our models against observations over the last century and earlier and they do extremely well in that respect. And finally we can test our models against perturbations. So we can for instance simulate a volcanic eruption for example and check that the climate models respond appropriately to what a volcano does to the atmosphere. So there are multiple lines of evidence and they all point to the climate models being reliable for what they were designed. There's a lot of myths out there on what we think or how we build our understanding of what will happen in the future. There are many lines of evidence that are used to understand how climate might change in the future. And if you could take climate models away we would still be just as worried about the future climate. Climate models merely inform and embellish and add colour and flavour to the future of the climate, the projections of future climate. But we would be just as worried based on theory and data. Climate models are one strand of evidence for future climate change but by no means do they underpin our concerns. What underpins our concerns is physics.