Julia Slingo of Reading University, UK, detailed how weather forecasting models have increased in resolution but climate models haven’t – they’ve tended to increase their duration, ensemble size, and/or complexity instead, for example including factors such as dynamic vegetation and atmospheric chemistry.

"Do we need to redress the balance now?" asked Slingo. Higher resolution would help show topographic and landscape influences such as mountains and megacities. Warren Washington of the US National Centre for Atmospheric Research was in agreement – he said that better resolution of the atmosphere could allow the inclusion of hurricanes and cyclones. But on the downside, higher resolution means making predictions on a decadal rather than a century-scale.

As tera- and petascale computing facilities as well as new Earth observation datasets are now available, Slingo reckons it’s a good time to examine the future of modelling. "Codes that scale across thousands of processors will be needed to exploit tera/petascale computing," she explained. And in an attempt to overcome the issues that arise from dealing with multiple scales, climate modellers are developing links with the computational fluid dynamics community, which has faced multiscale modelling problems for years.

Slingo raised concerns that the current parameterization approach may not be optimal for handling models that introduce life into the climate system, such as earth system and global environment system models, which incorporate biological and human factors.

In summary, Slingo’s three challenges for the future are: representing the multiscale nature of the climate; developing model codes that will exploit future petascale computing architectures; and how to represent living organisms and human responses within coupled climate systems.

Washington also spoke about some of the new trends in modelling, including the finite spectral element method HOMME that’s under development at NCAR. "It’s promising but not fully tested," he said. Washington believes the future is to use new numerical methods that are more amenable to running on very large computers. He’s also in favour of animations, which "aren’t just a toy, they can be very useful" in revealing patterns that might otherwise be missed.

"The community needs to be more effective at dealing with unethical sceptics," he added. "Ethical sceptics help us improve our models – they play a major service by pointing out weaker science."

Life model
On the biological side, Stephen Pacala of Princeton University, US, explained that researchers don’t understand the fundamental processes that affect how biological systems interact with climate. "We don’t have the Navier Stokes equations for the biosphere," he said. Pacala believes there’s a need to build a sophisticated statistical interface with climate models. "Models disagree about the future of the terrestrial carbon sink," he said. "That’s not surprising as we can’t determine where the missing carbon sink is at the moment."

Models tend to divide the biosphere into five different types of biota, assigning a climate sensitivity to each one. But there’s more variation within a functional type than between them. Pacala reckons that opens too much "wiggle room" when assigning parameter values. "We should take tuning out of the closet and learn something from it,' he said. It’s important to find out more as the uncertainty in the terrestrial biosphere response might halve or double the amout of effort needed to solve the climate problem.

Such research could involve reconstructing the past history of land use. Below-ground microbial communities are also a murky area. "We don’t even know what those organisms are so it’s very easy to ignore them," said Pacala. "There must be lots of redundancy in function between the species because there are so many of them. We don’t know enough about their response. The biggest uncertainty is the race between drying and warming of northern soils."

• The proceedings of the meeting will be published in the Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences in early 2009.