thespace-timestochasticcascadestructureoftheatmosphere.docVIP

Words: 11345 (without references), 8774, no figs. The Weather and Climate: Emergent Laws and Multifractal Cascades 1. Introduction 1.1 The new synthesis 1.1.1 Two (irreconcilable?) approaches to understanding the atmosphere In the last twenty years there has been a quiet revolution in atmospheric modelling. It’s not just that computers and numerical algorithms have continued their development, but rather that the very goal of the modelling has profoundly changed. Whereas twenty years ago, the goal was to determine the (supposedly) unique state of the atmosphere, today with the advent in Ensemble Forecasting Systems (EFS), the aim is to determine the possible future atmospheric states including their relative probabilities of occurrence: this new goal is stochastic. A stochastic process is a set of random variables indexed by time (Kolmogorov 1933), and this definition includes that of deterministic processes as a special case. At present, the EFS are really hybrids in the sense that they operate by first generating an initial ensemble of atmospheric states compatible with the observations and then they use conventional deterministic forecasting techniques to advance each member in time to produce a distribution of future states. Once the leap was taken to go beyond the forecasting of a unique state to forecasting an ensemble, the next step was to make the subgrid parametrisations themselves stochastic (e.g. (Buizza et al., 1999), (Palmer, 2001)), (Palmer and Williams, 2010)). This is an attempt to take into account the variability of different subgrid circulations. The artificial deterministic / stochastic nature of these hybrids suggests that the development or pure stochastic forecasts would be advantageous, a possibility we explore in ch. 9. Interestingly, the tension between determinism and stochasticity has been around pretty much since the beginning, although for most of the (still brief) history of atmospheric science, the deterministic approaches