Stratospheric and Tropical Impacts on Subseasonal Forecasts Related to Stratospheric Vortex Disruptions

The roles of stratospheric circulation and tropical diabatic heating anomalies in modulating subseasonal (weeks 3-6) tropospheric forecast skill for extended winter (December-March) is investigated using a linear inverse model (LIM), an empirical dynamical model that approximates chaotic atmospheric nonlinearity in terms of predictable linear dynamics and unpredictable noise. The LIM is constructed from 5-day lag covariances of 7-day running mean anomalies determined from the Japanese Reanalysis (JRA-55) for the 1979-2016 period. Cross-validated hindcasts generated using the LIM are then compared to World Climate Research Program Subseasonal-to-Seasonal Prediction Project ECWMF hindcasts for the 1996-2015 winter season (DJF) and NCEP’s CFSv2 for the 1999-2010 winters. Skill is evaluated for anomalous mean sea level pressure (mslp) and 500 hPa geopotential height.

For Week 3, all three models have comparable skill, with similar regional and temporal variations. The ECMWF model has the highest week 3 skill. For longer leads, skill patterns remain similar, but the LIM skill shows a smaller decline with lead so that it’s skill exceeds the CFSv2 by Week 4 and the ECMWF by week 5. On average, skill is low, with the North Pacific maximum local anomaly correlation for mslp of all three forecast models dropping below 0.5-0.6 for forecast leads greater than week 3. However, forecasts that are initialized in the 10-25 day window following major disruptions of the stratospheric polar vortex exhibit substantially enhanced skill for the LIM and ECMWF models. During these periods, at weeks 3 and 4 the two models have similar skill, while the LIM has greater skill at weeks 5 and 6.

Given that the LIM has comparable skill to the ECMWF model and shows similarly enhanced forecast skill at 4-6 week leads related to stratospheric vortex events, we conduct further diagnosis of the LIM’s forecast operator to quantify sources of predictive skill. A singular value decomposition (SVD) of this operator yields the optimal initial conditions leading to maximum growth of stratospheric anomalies. Because the LIM’s linear dynamical operator includes explicit representations of tropospheric and stratospheric dynamical variables and tropical diabatic heating, the SVD analysis allows us to isolate what types of circulation and heating initial states lead to the large stratospheric vortex disruptions that are associated with enhanced tropospheric predictive skill.