Examining Sources of Variability in Model Skill during the 7 January 2013 Sudden Stratospheric Warming Event

The purpose of this study is to investigate the sources of variability in stratospheric forecast skill amongst several operational models initiated at various lead times prior to the 7 January 2013 sudden stratospheric warming (SSW). This study extends upon a previous analysis by the Stratospheric Network for the Assessment of Predictability (SNAP), which concluded that a change in lead time from 15 to 10 days increased model skill of this weak vortex event by roughly 50 percent. The cause of such variability in predictive skill remains an important question and will be investigated further in this study.

It is hypothesized that model error in forecasting thermodynamic processes (e.g. latent heating) within the SSW precursor tropospheric blocking resulted in systematic biases and variability in dynamical forcing (e.g. upward wave flux) into the polar stratosphere in forecasts of the 7 January 2013 SSW event. Both medium-range and sub-seasonal forecast periods will be interrogated for model variability and initialized at several different lead times. Medium-range forecasts will be analyzed using NCEP Global Ensemble Forecast System Reforecast Version 2 data, and sub-seasonal forecast periods will be investigated using Sub-seasonal to Seasonal (S2S) Prediction Project data. Both elliptical diagnostics of the polar vortex – its eccentricity, center longitude and latitude, and area encompassed by the vortex – and zonal mean metrics will be used to assess model errors and biases in the stratosphere. Total diabatic heating and negative potential vorticity advection via the irrotational wind will be calculated to analyze the thermodynamic processes linked to tropospheric blocking. For each model and ensemble member, forecast skill will be assessed via a combination in the ability to forecast the stratospheric regime change, the dynamic wave forcing, and the thermodynamic forcing within the tropospheric block. Forecast error and uncertainty of both thermodynamic and dynamic forcing will also be assessed for different model configuration (e.g. parameterization, resolution) composite groups.