Predictability of the Arctic Oscillation From Stratospheric Influences in Operational Forecasts Junsu Kim 1, Arun Kumar 2, and Thomas Reichler 1 ( 1 Univ. of Utah, 2 NOAA) a.gov/products/precip/C Wlink/daily_ao_index
Questions To what extent do current prediction systems exploit stratospheric signals? How much practical skill is associated with it? How important is a good stratospheric component? 2
Hindcasts or retrospective forecasts, Hamill et al. (2006, BAMS) Fixed 1998 version of NCEP/GFS, T62 Low-top: L28, 7 levels above 100, 2.5 hPa 1979 – present, daily forecasts out to day member ensemble +daily forecasts; each SSW event is captured ˗ forecasts go only out to day 15 ˗ no data on stratospheric levels are archived NOAA/CDC Reforecasts 3
Validation Against NCEP/NCAR reanalysis Ensemble means Northern Annular Mode (NAM) derived from –Sea level pressure (SLP) → AO –Z850, 700, 500, 250, and 150 → NAM x
SLP errors 5 Day JFM AMJ JAS OND Reforecasts were bias- corrected as a function of lead time and day of year Reforecast Climatology Drift
Case Study: Winter 2004 JAN1 FEB1 MAR1 Day 8.5 Day 14.5 Day AO NCEP/NCAR reanalysis NAM index by level and day range 1 sdev NNR CDC
Strong SSW Events Reanalysis ( ) NAM 10 <-3 → 16 events Discard 6 events which exhibit no downward propagation → 10 strong events Reforecasts Reduced amplitudes Basic patterns are retained 7 Day 15 Forecasts (EM) Reanalysis Days after event
SSW Composite 10 strong events NCEP/NCAR Reanalysis 8
Day 15 Day 0 Day 10 Day 5 SSW Composite 9 Reforecasts
Correlation Lagged Correlation AO (61 days, centered at day 19) NAM 10 / AO Evolution Index Days after SSW event Composite NAM/AO Index Reforecast signal arrives too late (ca. 10 days) at surface NAM 10 reanalyses AO reanalyses AO reforecasts day 15
Basic Predictability ACC SLP vs. AO ACC SLP AO AO Winter 11
AO Predictability Day of year Lead time (day) r Shading: Daily correlations Contours: Monthly running means Increased predictability in winter Low predictability in summer Fast decline in spring; final warmings? Gradual increase in fall Correlation: Reanalysis vs. Reforecast r=0.6
Correlation Anomalies Index 13 0 Correlations SSW NAM 10 reanalyses AO reforecasts AO reanalyses AO Predictability During SSWs t [days] Composite of 10 SSWs 31 days AO r(reanalysis; reforecasts) Anomalies: February climatology removed Enhanced AO predictability at longer leads and after SSW event
Conclusion Despite low stratospheric resolution (7 levels), reforecasts show skill in simulating downward propagating signals These signals arrive too late at surface with the delay increasing with lead time SSW events improve AO predictability 0-15 days after the event and at leads of 8-15 days Beside the 10 SSW events examined here, additional skill from stratospheric influences may be realized during other warm events or during cold events