Stratospheric Network for the Assessment of Predictability

During winter and spring, the stratosphere is a dynamically exciting place, with intense and dramatic stratospheric major warming events occurring typically in two out of every three years in the Northern hemisphere (Charlton and Polvani, 2007) and minor warming events occurring more frequently still. It is not surprising, therefore, that there has long been interest in understanding what role the stratosphere might play in influencing tropospheric weather and climate. Following from the studies of Baldwin and Dunkerton (1999,2001) there has been a renewed interest in this topic over the past fifteen years. One particular aspect of this problem, first captured succinctly by Boville and Baumhefner (1990), is the idea that an enhanced representation of the stratosphere in models used for forecasting tropospheric weather on short to medium ranges might enhance the tropospheric skill in those models. Many recent studies have confirmed and enhanced these original ideas (e.g. Charlton et al., 2004; Jung and Barkmeijer, 2006; Kuroda, 2010; Roff et al., 2011) leaving atmospheric scientists with a general picture of the stratosphere-troposphere link as one which can add skill to tropospheric forecasts on timescales of 5-15 days, on large planetary scales and in both the northern and southern hemisphere extra-tropics.

SNAP grew out of discussion in the SPARC Data Assimilation Working Group and aims to answer some of the remaining fundamental questions about stratospheric predictability. In particular:

• Are stratosphere-troposphere coupling effects important throughout the winter season or only when major stratospheric dynamical events occur?
• How far in advance can major stratospheric dynamical events be predicted and usefully add skill to tropospheric forecasts?
• Which stratospheric processes, both resolved and unresolved need to be captured by models to gain optimal stratospheric predictability?

Why a new international network?

Answering these scientific and technical questions requires collaboration between the parts of the scientific community interested in stratospheric predictability (both stratospheric dynamicists and forecast providers) and it requires carefully planned experiments that objectively compare the stratospheric skill of different numerical models and understand its source. SNAP will provide a central forum by which this expertise can be regularly shared and improved. The centrepiece of SNAP will be to design and perform a new intercomparison of stratospheric forecasts, producing and maintaining a rich dataset to be used by a broad community of researchers.

Ten years of diligent work by the stratospheric research community has convinced many operational centres to raise the top of their numerical weather prediction models to include the stratosphere and the time is now ripe to seize the opportunity to understand and quantify stratospheric predictability. This is not a task any one individual research group or forecast organisation can achieve on their own, since representation of the stratosphere in NWP models is still in its infancy. Although many models now place their model lid above the stratopause, there are many fundamental unanswered questions about how to represent stratospheric physics properly and appropriately in those models. Examples of poorly studied or understood stratospheric processes in the context of NWP are the role of varying chemical composition in the stratosphere (particularly ozone) and the optimal way to incorporate small-scale gravity waves which are crucial for many stratospheric processes.

The representation of the stratosphere in NWP models can be compared to that of the ocean in climate models in the 1990s; the first step is to add the physical system to a model but the second and much more demanding task is to consider how best to develop and optimise that system for the task at hand. As in many other areas of atmospheric science, comparing models and collaborating on their improvement is often the best spur to rapid progress, since it allows all the groups involved to understand the best and the worst aspects of the choices which must be made in modelling a complex physical system.