1. Training Course – NWP-PR: Weather regimes 1 Weather regimes Franco Molteni European Centre for Medium-Range Weather Forecasts

2. Training Course – NWP-PR: Weather regimes 2 Outline Introduction: Dynamical concepts Examples of recurrent flow patterns Historical overview Impact of external/boundary forcing on atmospheric regimes Detection of regimes in atmospheric and model datasets Time filtering Distinction between internal and intra-seasonal variability PDF estimation and statistical significance Applications to seasonal predictions Diagnostics of regimes in model simulations Predictability of regime frequencies Non-linear impact of ENSO on regime properties

3. Training Course – NWP-PR: Weather regimes 3 Weather regimes and related dynamical concepts Weather regime: A persistent and/or recurrent large-scale atmospheric circulation pattern which is associated with specific weather conditions on a regional scale Flow regime: A persistent and/or recurrent large-scale flow pattern in a (geophysical) fluid-dynamical system Multiple equilibria: Multiple stationary solutions of a non-linear dynamical system

4. Training Course – NWP-PR: Weather regimes 4 Recurrent flow patterns: examples A sequence of 5-day mean fields of 500 hPa geopotential height during boreal winter …

5. Training Course – NWP-PR: Weather regimes 5 Recurrent flow patterns: examples … but each one occurred in a different winter ! 5-9 Jan 1985 4-8 Feb 1986 10-14 Jan 1987

6. Training Course – NWP-PR: Weather regimes 6 Regimes as quasi-stationary states q : barotropic or quasi-geostrophic potential vorticity t q = - V ψ grad q - D (q – q*) steady state for instantaneous flow: 0 = - V ψ grad q - D (q – q*) steady state for time-averaged flow: 0 = - V ψ grad q - D (q – q*) - V ψ grad q

7. Training Course – NWP-PR: Weather regimes 7 Possible regime behaviour: large-scale teleconnections The Pacific / North American (PNA) pattern

8. Training Course – NWP-PR: Weather regimes 8 Possible regime behaviour: Atlantic/Pacific blocking 500 hPa geop. height

9. Training Course – NWP-PR: Weather regimes 9 2002 2004 2003 All-India Rainfall time-series (May-September) Possible regime behaviour: monsoon active/brake phase

10. Training Course – NWP-PR: Weather regimes 10 Flow regimes in non-linear systems Lorenz (1963) truncated convection model dX/dt = σ (Y – X) dY/dt = - X Z + r X –Y dZ/dt = X Y – b Z Unstable stationary states X = Y = Z = 0 X = Y = ± [ b (r-1)] ½, Z = r -1

11. Training Course – NWP-PR: Weather regimes 11 Seminal papers: Charney and DeVore 1979 Multiple steady states of low-order barotropic model with wave-shaped bottom topography

12. Training Course – NWP-PR: Weather regimes 12 Papers on multiple equilibria and quasi-stationary states Orographically forced models : Charney and Straus 1980: Form-grad instability, multiple equilibria and propagating planetary waves in baroclinic, orographically forced, planetary wave systems Charney, Shukla and Mo 1981: Comparison of barotropic blocking theory with observation Legras and Ghil 1985: Persistent anomalies, blocking and variations in atmospheric predictability Benzi, Malguzzi, Speranza, Sutera 1986: The statistical properties of the atmospheric general circulation: observational evidence and a minimal theory of bimodality Thermally forced models: Mitchell and Derome 1983: Blocking-like solutions of the potential vorticity equation: their stability at equilibrium and growth at resonance

13. Training Course – NWP-PR: Weather regimes 13 Seminal papers: Reinhold and Pierrehumbert 1983 Hemispheric weather regimes arising from equilibration of large-scale dynamical tendencies and forcing from transient baroclinic eddies

14. Training Course – NWP-PR: Weather regimes 14 Eddy forcing of blocking: the Imperial College school Green 1977: The weather during July 1976: some dynamical consideration of the drought Illari and Marshall 1983: On the interpretation of eddy fluxes during a blocking episode Shutts 1986: A case study of eddy forcing during an Atlantic blocking episode Haines and Marshall 1987: Eddy-forced coherent structures as a propotype of atmospheric blocking

15. Training Course – NWP-PR: Weather regimes 15 Seminal papers: Vautard and Legras 1988 Regional weather regimes arising from equilibration of large-scale dynamical tendencies and PV fluxes from transient baroclinic eddies

16. Training Course – NWP-PR: Weather regimes 16 Seminal papers: Hansen and Sutera 1986 Bimodality in the probability density function (PDF) of an index of N. Hem. planetary wave amplitude due to near-resonant wave-numbers (m=2-4)

17. Training Course – NWP-PR: Weather regimes 17 Multi-dim. PDF estimation and cluster analysis Searching for densely- populated regions in phase space: Mo and Ghil 1988 Molteni et al. 1990 Cheng and Wallace 1993 Kimoto and Ghil 1993a, b Michelangeli et al. 1995 Corti et al. 1999 Kimoto and Ghil 1993a

18. Training Course – NWP-PR: Weather regimes 18 Impact of external forcing in non-linear systems Lorenz (1963) truncated convection model with additional forcing f (Molteni et al. 1993; Palmer 1993) dX/dt = σ (Y – X) dY/dt = - X Z + r X –Y + f dZ/dt = X Y – b Z Unstable stationary states (for f=0) X = Y = Z = 0 X = Y = ± [ b (r-1)] ½, Z = r -1

19. Training Course – NWP-PR: Weather regimes 19 Impact of external forcing in non-linear systems The properties of atmospheric flow regimes may be affected by anomalies in boundary forcing in two different ways: Weak forcing anomaly: the number and spatial patterns of regimes remain the same, but their frequency of occurrence is changed (Lorenz model paradigm) Strong forcing anomaly: the number and patterns of regimes are modified as the atmospheric system goes through bifurcation points

20. Training Course – NWP-PR: Weather regimes 20 Regime detection: time filtering PDFs along the axis connecting the weakly-unstable steady states in the 1963 Lorenz system (Marshall and Molteni 1993) -------- : unfiltered data _____ : running means δt = 0.6 -------- : running means δt = 0.4 _____ : running means δt = 0.6.......... : running means δt = 0.8

21. Training Course – NWP-PR: Weather regimes 21 PDF of total vs. intra-seasonal variability P(x) = f 1 G (μ 1, σ 1 ) + f 2 G (μ 2, σ 2 ) where: f 1 + f 2 = 1, f 1 μ 1 + f 2 μ 2 = μ* P(x) is bimodal if:| μ 1 - μ 2 | > σ 1 + σ 2 P(x) = 0.5 G ( 0.8, 0.6) + 0.5 G (-0.8, 0.6)

22. Training Course – NWP-PR: Weather regimes 22 PDF of total vs. intra-seasonal variability Let us assume that the total sample is composed by seasons when f 1 =0.7, f 2 =0.3, and seasons when f 1 =0.3, f 2 =0.7 P(x) = f 1 G (0.8, 0.6) + f 2 G (-0.8, 0.6) PDF of total anomaly PDF of intra-seasonal anomaly

23. Training Course – NWP-PR: Weather regimes 23 PDF estimation with the Gaussian kernel method EOFs/PCs of monthly-mean 500 hPa height P(x) = N -1 i G (x i, h σ ) h : kernel width h = 0.3 h = 0.4 h = 0.5

24. Training Course – NWP-PR: Weather regimes 24 Regimes from PDF estimation (Corti et al. 1999)

25. Training Course – NWP-PR: Weather regimes 25 PDF estimation: statistical significance Fraction of uni/multi-modal PDFs obtained from a gaussian distribution

26. Training Course – NWP-PR: Weather regimes 26 Regime diagnostics on CGCM variability System-3 11-member ensembles, init. on 1 nov. 1981/2005 Clusters of low-pass-filtered 200 hPa height Results consistent with Michelangeli et al. 1995

27. Training Course – NWP-PR: Weather regimes 27 A non-linear approach to extra-trop. predictions Cluster analysis of low-freq. ( T>10 d) Z 200 in NCEP re-analysis and COLA AGCM ensembles (Straus, Corti, Molteni 2007)

28. Training Course – NWP-PR: Weather regimes 28 A non-linear approach to extra-trop. predictions Predictability of cluster frequencies (SCM 2007)

29. Training Course – NWP-PR: Weather regimes 29 Does ENSO affect the number of regimes? Ratio of inter-cluster to intra-cluster variance as a function of ENSO indices (Straus and Molteni 2004)

30. Training Course – NWP-PR: Weather regimes 30 Conclusions Flow regime behaviour can be reproduced in a variety of dynamical models of different complexity. Regimes may be defined on a hemispheric or regional domain. The impact of boundary forcing anomalies on regime properties may occur through changes in regime frequencies or bifurcation effects. Detection of regimes in atmospheric and model datasets is dependent on appropriate time-filtering, distinction between total/internal and intra-seasonal variability, proper use of statistical significance tests. Predictability of regime frequencies and variations in the number of regimes as a function of the ENSO phase have been detected in ensembles of GCM simulations, and offer an alternative approach to seasonal prediction.