Session U6: Computational and Theoretical Challenges in Predicting Climate Change

Radiative Transfer in Climate Models

Radiation is a key physics element in both the maintenance of the climate as well as in driving climate change. The absorption of the Sun's radiant energy by the surface-atmosphere system in the ultra-violet, visible and near-infrared spectral regions, and the absorption and emission of infrared radiation by the surface and atmosphere, together govern the planetary energy balance. The importance of radiative processes enters in both the ``forcing'' of the climate system and in the ``feedbacks'' that amplify the response to perturbations of the energy balance. As examples, we will examine the natural and anthropogenic radiative forcings that have occurred over the 20$^{th}$ century, our understanding of the governing processes and the challenges in representing them in climate models. The quantitative description comprises the determination of the forcing at the surface and in the atmosphere due to: emissions of the long-lived greenhouse gases (e.g., carbon dioxide), ozone precursors, and pollution particulates (e.g., sulfate and black carbon); changes in land surface properties; changes in solar irradiance; and particulates arising due to episodic volcanic eruptions. The various types of forcings are governed by fundamentally different underlying mechanisms, have distinct space-time dependencies and uncertainties, and exert varying signatures in terms of the climate system responses.   

Climate Feedbacks and Their Simulation in Coupled Ocean Atmosphere Models

The response of Earth's climate to an increase in greenhouse gases depends on a complex superposition of feedback processes. These processes act to either amplify or dampen the climate's response to an initial perturbation in the Earth's radiative energy budget. Differences in the representation of these feedback processes in current models represent a primary source of uncertainty in model projections of future climate change. Progress in reducing uncertainties in model predictions of climate change therefore requires an accurate assessment of the differences in various feedback strengths between models. In this talk I will review the key climate feedback processes and assess their range of values in current models. Attention will be focused on the feedbacks from water vapor and clouds, which represent the most important climate feedbacks in current models. My presentation will describe the prevailing view behind these feedbacks and review observational evidence used in assessing the fidelity of their representation in current models.   

Predicting climate change: Uncertainties and prospects for surmounting them

General circulation models (GCMs) are among the most detailed and sophisticated models of natural phenomena in existence. Still, the lack of robust and efficient subgrid-scale parametrizations for GCMs, along with the inherent sensitivity to initial data and the complex nonlinearities involved, present a major and persistent obstacle to narrowing the range of estimates for end-of-century warming. Estimating future changes in the distribution of climatic extrema is even more difficult. Brute-force tuning the large number of GCM parameters does not appear to help reduce the uncertainties. Andronov and Pontryagin (1937) proposed \textit{structural stability} as a way to evaluate model robustness. Unfortunately, many real-world systems proved to be structurally unstable. We illustrate these concepts with a very simple model for the El Ni\~{n}o--Southern Oscillation (ENSO). Our model is governed by a differential delay equation with a single delay and periodic (seasonal) forcing. Like many of its more or less detailed and realistic precursors, this model exhibits a Devil's staircase. We study the model's structural stability, describe the mechanisms of the observed instabilities, and connect our findings to ENSO phenomenology. In the model's phase-parameter space, regions of smooth dependence on parameters alternate with rough, fractal ones. We then apply the tools of random dynamical systems and \textit{stochastic structural stability} to the circle map and a torus map. The effect of noise with compact support on these maps is fairly intuitive: it is the most robust structures in phase-parameter space that survive the smoothing introduced by the noise. The nature of the stochastic forcing matters, thus suggesting that certain types of stochastic parametrizations might be better than others in achieving GCM robustness. This talk represents joint work with M. Chekroun, E. Simonnet and I. Zaliapin.