Modelling Interactions Between Weather and Climate p. 1/33 Modelling Interactions Between Weather and Climate Adam Monahan & Joel Culina monahana@uvic.ca & culinaj@uvic.ca School of Earth and Ocean Sciences, University of Victoria Victoria, BC, Canada
Modelling Interactions Between Weather and Climate p. 2/33 Introduction Earth s climate system displays variability on many scales
Modelling Interactions Between Weather and Climate p. 2/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m
Modelling Interactions Between Weather and Climate p. 2/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond)
Modelling Interactions Between Weather and Climate p. 2/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather
Introduction From von Storch and Zwiers, 1999 Modelling Interactions Between Weather and Climate p. 3/33
Modelling Interactions Between Weather and Climate p. 4/33 Introduction From von Storch and Zwiers, 1999
Modelling Interactions Between Weather and Climate p. 5/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather Dynamical nonlinearities coupling across space/time scales
Modelling Interactions Between Weather and Climate p. 5/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather Dynamical nonlinearities coupling across space/time scales modelling climate, weather must be parameterised, not ignored
Modelling Interactions Between Weather and Climate p. 5/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather Dynamical nonlinearities coupling across space/time scales modelling climate, weather must be parameterised, not ignored Classical deterministic parameterisations assume very large scale separation
Modelling Interactions Between Weather and Climate p. 5/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather Dynamical nonlinearities coupling across space/time scales modelling climate, weather must be parameterised, not ignored Classical deterministic parameterisations assume very large scale separation Following Mitchell (1966), Hasselmann (1976) suggested stochastic parameterisations more appropriate when separation finite
Modelling Interactions Between Weather and Climate p. 5/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather Dynamical nonlinearities coupling across space/time scales modelling climate, weather must be parameterised, not ignored Classical deterministic parameterisations assume very large scale separation Following Mitchell (1966), Hasselmann (1976) suggested stochastic parameterisations more appropriate when separation finite This talk will consider:
Modelling Interactions Between Weather and Climate p. 5/33 Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather Dynamical nonlinearities coupling across space/time scales modelling climate, weather must be parameterised, not ignored Classical deterministic parameterisations assume very large scale separation Following Mitchell (1966), Hasselmann (1976) suggested stochastic parameterisations more appropriate when separation finite This talk will consider: stochastic averaging as a tool for obtaining stochastic climate models from coupled weather/climate system
Introduction Earth s climate system displays variability on many scales space: 10 6 m - 10 7 m time: seconds - 10 6 years (and beyond) Loosely, slower variability called climate and faster called weather Dynamical nonlinearities coupling across space/time scales modelling climate, weather must be parameterised, not ignored Classical deterministic parameterisations assume very large scale separation Following Mitchell (1966), Hasselmann (1976) suggested stochastic parameterisations more appropriate when separation finite This talk will consider: stochastic averaging as a tool for obtaining stochastic climate models from coupled weather/climate system two examples: coupled atmosphere/ocean boundary layers, extratropical atmospheric low-frequency variability Modelling Interactions Between Weather and Climate p. 5/33
Modelling Interactions Between Weather and Climate p. 6/33 A statistical mechanical perspective Consider a box at temperature T containing N ideal gas molecules
Modelling Interactions Between Weather and Climate p. 6/33 A statistical mechanical perspective Consider a box at temperature T containing N ideal gas molecules Kinetic energy of individual molecules randomly distributed with Maxwell-Boltzmann distribution p(e)
Modelling Interactions Between Weather and Climate p. 6/33 A statistical mechanical perspective Consider a box at temperature T containing N ideal gas molecules Kinetic energy of individual molecules randomly distributed with Maxwell-Boltzmann distribution p(e) As N mean energy approaches sharp value E = 1 N N E i N E = E p(e) de i=1
Modelling Interactions Between Weather and Climate p. 6/33 A statistical mechanical perspective Consider a box at temperature T containing N ideal gas molecules Kinetic energy of individual molecules randomly distributed with Maxwell-Boltzmann distribution p(e) As N mean energy approaches sharp value E = 1 N N E i N E = E p(e) de i=1 For N large (but finite), E is random E = E + 1 N ζ such that (by central limit theorem) ζ is Gaussian
Modelling Interactions Between Weather and Climate p. 6/33 A statistical mechanical perspective Consider a box at temperature T containing N ideal gas molecules Kinetic energy of individual molecules randomly distributed with Maxwell-Boltzmann distribution p(e) As N mean energy approaches sharp value E = 1 N N E i N E = E p(e) de i=1 For N large (but finite), E is random E = E + 1 N ζ such that (by central limit theorem) ζ is Gaussian Very large scale separation deterministic dynamics
Modelling Interactions Between Weather and Climate p. 6/33 A statistical mechanical perspective Consider a box at temperature T containing N ideal gas molecules Kinetic energy of individual molecules randomly distributed with Maxwell-Boltzmann distribution p(e) As N mean energy approaches sharp value E = 1 N N E i N E = E p(e) de i=1 For N large (but finite), E is random E = E + 1 N ζ such that (by central limit theorem) ζ is Gaussian Very large scale separation deterministic dynamics Smaller (but still large) separation stochastic corrections needed
Modelling Interactions Between Weather and Climate p. 7/33 Hasselmann s ansatz: a stochastic climate model Observation: spectra of sea surface temperature (SST) variability are generically red
Modelling Interactions Between Weather and Climate p. 7/33 Hasselmann s ansatz: a stochastic climate model Observation: spectra of sea surface temperature (SST) variability are generically red Observations from North Atlantic weathership INDIA From Frankignoul & Hasselmann Tellus 1977
Modelling Interactions Between Weather and Climate p. 8/33 Hasselmann s ansatz: a stochastic climate model Simple model of Frankignoul & Hasselmann (1977) assumed:
Modelling Interactions Between Weather and Climate p. 8/33 Hasselmann s ansatz: a stochastic climate model Simple model of Frankignoul & Hasselmann (1977) assumed: local dynamics of small perturbations T linearised around climatological mean state
Modelling Interactions Between Weather and Climate p. 8/33 Hasselmann s ansatz: a stochastic climate model Simple model of Frankignoul & Hasselmann (1977) assumed: local dynamics of small perturbations T linearised around climatological mean state fast atmospheric fluxes represented as white noise
Modelling Interactions Between Weather and Climate p. 8/33 Hasselmann s ansatz: a stochastic climate model Simple model of Frankignoul & Hasselmann (1977) assumed: local dynamics of small perturbations T linearised around climatological mean state fast atmospheric fluxes represented as white noise simple Ornstein-Uhlenbeck process with spectrum dt dt = 1 τ T + γẇ E{T (ω) 2 } = γ 2 τ 2 2π(1 + ω 2 τ 2 )
Modelling Interactions Between Weather and Climate p. 8/33 Hasselmann s ansatz: a stochastic climate model Simple model of Frankignoul & Hasselmann (1977) assumed: local dynamics of small perturbations T linearised around climatological mean state fast atmospheric fluxes represented as white noise simple Ornstein-Uhlenbeck process with spectrum Linear stochastic model dt dt = 1 τ T + γẇ E{T (ω) 2 } = γ 2 τ 2 2π(1 + ω 2 τ 2 ) simple null hypothesis for observed variability
Modelling Interactions Between Weather and Climate p. 8/33 Hasselmann s ansatz: a stochastic climate model Simple model of Frankignoul & Hasselmann (1977) assumed: local dynamics of small perturbations T linearised around climatological mean state fast atmospheric fluxes represented as white noise simple Ornstein-Uhlenbeck process with spectrum Linear stochastic model dt dt = 1 τ T + γẇ E{T (ω) 2 } = γ 2 τ 2 2π(1 + ω 2 τ 2 ) simple null hypothesis for observed variability We ll be coming back to this again later on...
Modelling Interactions Between Weather and Climate p. 9/33 Coupled fast-slow systems Starting point for stochastic averaging is assumption that state variable z can be written z = (x,y)
Modelling Interactions Between Weather and Climate p. 9/33 Coupled fast-slow systems Starting point for stochastic averaging is assumption that state variable z can be written z = (x,y) where x slow variable (climate) y fast variable (weather) d dt x = f(x,y) d dt y = g(x,y) such that
Modelling Interactions Between Weather and Climate p. 9/33 Coupled fast-slow systems Starting point for stochastic averaging is assumption that state variable z can be written z = (x,y) where x slow variable (climate) y fast variable (weather) d dt x = f(x,y) d dt y = g(x,y) such that where is small" τ f(x,y) g(x, y)
Modelling Interactions Between Weather and Climate p. 10/33 A simple example As a simple example of a coupled fast-slow system consider d dt x = x x3 + (Σ + x)y d dt y = 1 τ x y + 1 Ẇ τ
Modelling Interactions Between Weather and Climate p. 10/33 A simple example As a simple example of a coupled fast-slow system consider d dt x = x x3 + (Σ + x)y d dt y = 1 τ x y + 1 Ẇ τ The weather process y is Gaussian with stationary autocovariance C yy (s) = E {y(t)y(t + s)} = x ( 2 exp s ) τ x as τ 0 τx 2 δ(s)
Modelling Interactions Between Weather and Climate p. 10/33 A simple example As a simple example of a coupled fast-slow system consider d dt x = x x3 + (Σ + x)y d dt y = 1 τ x y + 1 Ẇ τ The weather process y is Gaussian with stationary autocovariance C yy (s) = E {y(t)y(t + s)} = x ( 2 exp s ) τ x as τ 0 τx 2 δ(s) Intuition suggests that as τ 0 d dt x x x3 + τ(σ + x) x Ẇ
Modelling Interactions Between Weather and Climate p. 11/33 Averaging approximation (A) As τ 0, over finite times we have x(t) x(t)
Modelling Interactions Between Weather and Climate p. 11/33 Averaging approximation (A) As τ 0, over finite times we have x(t) x(t) with d dt x = f(x)
Modelling Interactions Between Weather and Climate p. 11/33 Averaging approximation (A) As τ 0, over finite times we have x(t) x(t) with where f(x) = d dt x = f(x) f(x,y)dµ x (y) = lim T 1 T T 0 f(x,y(t)) dt
Modelling Interactions Between Weather and Climate p. 11/33 Averaging approximation (A) As τ 0, over finite times we have x(t) x(t) with where f(x) = d dt x = f(x) f(x,y)dµ x (y) = lim T 1 T T averages f(x,y) over weather for a fixed climate state 0 f(x,y(t)) dt
Modelling Interactions Between Weather and Climate p. 11/33 Averaging approximation (A) As τ 0, over finite times we have x(t) x(t) with where f(x) = d dt x = f(x) f(x,y)dµ x (y) = lim T 1 T T averages f(x,y) over weather for a fixed climate state 0 f(x,y(t)) dt To leading order, climate follows deterministic averaged dynamics
Modelling Interactions Between Weather and Climate p. 11/33 Averaging approximation (A) As τ 0, over finite times we have x(t) x(t) with where f(x) = d dt x = f(x) f(x,y)dµ x (y) = lim T 1 T T averages f(x,y) over weather for a fixed climate state 0 f(x,y(t)) dt To leading order, climate follows deterministic averaged dynamics For simple model d dt x = E y x { x x 3 + xy + Σy } = x x 3
Modelling Interactions Between Weather and Climate p. 11/33 Averaging approximation (A) As τ 0, over finite times we have x(t) x(t) with d dt x = f(x) where f(x) = f(x,y)dµ x (y) = lim T 1 T T 0 f(x,y(t)) dt averages f(x,y) over weather for a fixed climate state To leading order, climate follows deterministic averaged dynamics For simple model d dt x = E y x { x x 3 + xy + Σy } = x x 3 i.e. depending on x(0) system settles into bottom of one of two potential wells x = ±1
Modelling Interactions Between Weather and Climate p. 12/33 Analogy to Reynolds averaging Approximation (A) analogous to Reynolds averaging in fluid mechanics: split flow into mean and turbulent components u = u + u
Modelling Interactions Between Weather and Climate p. 12/33 Analogy to Reynolds averaging Approximation (A) analogous to Reynolds averaging in fluid mechanics: split flow into mean and turbulent components u = u + u where u(t) = 1 2T T T u(t + s) ds
Modelling Interactions Between Weather and Climate p. 12/33 Analogy to Reynolds averaging Approximation (A) analogous to Reynolds averaging in fluid mechanics: split flow into mean and turbulent components where u(t) = 1 2T u = u + u T T Full flow satisfies Navier-Stokes equations u(t + s) ds t u + u u = force
Modelling Interactions Between Weather and Climate p. 12/33 Analogy to Reynolds averaging Approximation (A) analogous to Reynolds averaging in fluid mechanics: split flow into mean and turbulent components where u(t) = 1 2T u = u + u T T Full flow satisfies Navier-Stokes equations u(t + s) ds t u + u u = force mean flow satisfies eddy averaged dynamics t u + u u = force u u
Analogy to Reynolds averaging Approximation (A) analogous to Reynolds averaging in fluid mechanics: split flow into mean and turbulent components where u(t) = 1 2T u = u + u T T Full flow satisfies Navier-Stokes equations u(t + s) ds t u + u u = force mean flow satisfies eddy averaged dynamics t u + u u = force u u Note: averaging is projection operator only if scale separation between u and u Modelling Interactions Between Weather and Climate p. 12/33
Modelling Interactions Between Weather and Climate p. 13/33 Central limit theorem approximation (L) For τ 0 averaged and full trajectories differ; this difference can be modelled as a random process
Modelling Interactions Between Weather and Climate p. 13/33 Central limit theorem approximation (L) For τ 0 averaged and full trajectories differ; this difference can be modelled as a random process Basic result: for τ small x(t) x(t) + ζ
Modelling Interactions Between Weather and Climate p. 13/33 Central limit theorem approximation (L) For τ 0 averaged and full trajectories differ; this difference can be modelled as a random process Basic result: for τ small x(t) x(t) + ζ where ζ is the multivariate Ornstein-Uhlenbeck process d ζ = Df(x) ζ + σ(x)ẇ dt
Modelling Interactions Between Weather and Climate p. 13/33 Central limit theorem approximation (L) For τ 0 averaged and full trajectories differ; this difference can be modelled as a random process Basic result: for τ small x(t) x(t) + ζ where ζ is the multivariate Ornstein-Uhlenbeck process d ζ = Df(x) ζ + σ(x)ẇ dt Diffusion matrix σ(x) defined by lag correlation integrals σ 2 (x) = E y x {f [x,y(t + s)]f [x,y(t)]} ds where f [x,y(t)] = f[x,y(t)] f[x]
Modelling Interactions Between Weather and Climate p. 13/33 Central limit theorem approximation (L) For τ 0 averaged and full trajectories differ; this difference can be modelled as a random process Basic result: for τ small x(t) x(t) + ζ where ζ is the multivariate Ornstein-Uhlenbeck process d ζ = Df(x) ζ + σ(x)ẇ dt Diffusion matrix σ(x) defined by lag correlation integrals σ 2 (x) = E y x {f [x,y(t + s)]f [x,y(t)]} ds where f [x,y(t)] = f[x,y(t)] f[x] Note that std(ζ) τ so corrections vanish in strict τ 0 limit
Modelling Interactions Between Weather and Climate p. 14/33 Central limit theorem approximation (L) Averaging solution x(t) augmented by Gaussian corrections such that x(t) itself not affected
Modelling Interactions Between Weather and Climate p. 14/33 Central limit theorem approximation (L) Averaging solution x(t) augmented by Gaussian corrections such that x(t) itself not affected feedback of weather on climate only through averaging; weather can t drive climate between metastable states
Modelling Interactions Between Weather and Climate p. 14/33 Central limit theorem approximation (L) Averaging solution x(t) augmented by Gaussian corrections such that x(t) itself not affected feedback of weather on climate only through averaging; weather can t drive climate between metastable states (L) approximation still has been very successful in many contexts (e.g. Linear Inverse Modelling)
Modelling Interactions Between Weather and Climate p. 14/33 Central limit theorem approximation (L) Averaging solution x(t) augmented by Gaussian corrections such that x(t) itself not affected feedback of weather on climate only through averaging; weather can t drive climate between metastable states (L) approximation still has been very successful in many contexts (e.g. Linear Inverse Modelling) For the simple example model: f (x, y) = (x x 3 + xy + Σy) (x x 3 ) = (x + Σ)y
Modelling Interactions Between Weather and Climate p. 14/33 Central limit theorem approximation (L) Averaging solution x(t) augmented by Gaussian corrections such that x(t) itself not affected feedback of weather on climate only through averaging; weather can t drive climate between metastable states (L) approximation still has been very successful in many contexts (e.g. Linear Inverse Modelling) For the simple example model: f (x, y) = (x x 3 + xy + Σy) (x x 3 ) = (x + Σ)y so σ 2 (x) = (x + Σ) 2 E y x {y(t + s)y(t)} ds = (x + Σ) 2 x 2
Central limit theorem approximation (L) Averaging solution x(t) augmented by Gaussian corrections such that x(t) itself not affected feedback of weather on climate only through averaging; weather can t drive climate between metastable states (L) approximation still has been very successful in many contexts (e.g. Linear Inverse Modelling) For the simple example model: f (x, y) = (x x 3 + xy + Σy) (x x 3 ) = (x + Σ)y so σ 2 (x) = (x + Σ) 2 E y x {y(t + s)y(t)} ds = (x + Σ) 2 x 2 x(t) 1 + ζ such that d dt ζ = 2ζ + (1 + Σ)2 Ẇ Modelling Interactions Between Weather and Climate p. 14/33
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate x(t) solution of stochastic differential equation (SDE) d x = f(x) + σ(x) Ẇ dt
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate x(t) solution of stochastic differential equation (SDE) d x = f(x) + σ(x) Ẇ dt Most accurate of all approximations on long timescales; can capture transitions between metastable states
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate x(t) solution of stochastic differential equation (SDE) d x = f(x) + σ(x) Ẇ dt Most accurate of all approximations on long timescales; can capture transitions between metastable states Noise-induced drift further potential feedback of weather on climate
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate x(t) solution of stochastic differential equation (SDE) d x = f(x) + σ(x) Ẇ dt Most accurate of all approximations on long timescales; can capture transitions between metastable states Noise-induced drift further potential feedback of weather on climate Simple model: d dt x = x x3 + (Σ + x) x Ẇ
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate x(t) solution of stochastic differential equation (SDE) d x = f(x) + σ(x) Ẇ dt Most accurate of all approximations on long timescales; can capture transitions between metastable states Noise-induced drift further potential feedback of weather on climate Simple model: d dt x = x x3 + (Σ + x) x Ẇ Multiplicative noise introduced through:
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate x(t) solution of stochastic differential equation (SDE) d x = f(x) + σ(x) Ẇ dt Most accurate of all approximations on long timescales; can capture transitions between metastable states Noise-induced drift further potential feedback of weather on climate Simple model: d dt x = x x3 + (Σ + x) x Ẇ Multiplicative noise introduced through: nonlinear coupling of x and y in slow dynamics
Modelling Interactions Between Weather and Climate p. 15/33 Hasselmann approximation (N) More complete approximation involves full two-way interaction between weather and climate x(t) solution of stochastic differential equation (SDE) d x = f(x) + σ(x) Ẇ dt Most accurate of all approximations on long timescales; can capture transitions between metastable states Noise-induced drift further potential feedback of weather on climate Simple model: d dt x = x x3 + (Σ + x) x Ẇ Multiplicative noise introduced through: nonlinear coupling of x and y in slow dynamics dependence of stationary distribution of y on x
Modelling Interactions Between Weather and Climate p. 16/33 When does this work? For the (A) approximation to hold all that is required is a timescale separation
Modelling Interactions Between Weather and Climate p. 16/33 When does this work? For the (A) approximation to hold all that is required is a timescale separation For (L) and (N) further require fast process is strongly mixing
Modelling Interactions Between Weather and Climate p. 16/33 When does this work? For the (A) approximation to hold all that is required is a timescale separation For (L) and (N) further require fast process is strongly mixing Informally, what is required is that the autocorrelation function of y decay sufficiently rapidly with lag for large timescale separations the delta-correlated white noise approximation is reasonable
Modelling Interactions Between Weather and Climate p. 16/33 When does this work? For the (A) approximation to hold all that is required is a timescale separation For (L) and (N) further require fast process is strongly mixing Informally, what is required is that the autocorrelation function of y decay sufficiently rapidly with lag for large timescale separations the delta-correlated white noise approximation is reasonable Crucially: none of this assumes that the fast process is stochastic. Effective SDEs (L) and (N) arise in limit τ 0 as approximation to fast deterministic or stochastic dynamics
Modelling Interactions Between Weather and Climate p. 17/33 Relation to MTV Theory" Have assumed that influence of weather on climate is bounded as τ 0
Modelling Interactions Between Weather and Climate p. 17/33 Relation to MTV Theory" Have assumed that influence of weather on climate is bounded as τ 0 MTV" theory considers averaging assuming that weather influence strengthens in this limit
Modelling Interactions Between Weather and Climate p. 17/33 Relation to MTV Theory" Have assumed that influence of weather on climate is bounded as τ 0 MTV" theory considers averaging assuming that weather influence strengthens in this limit An example: dx dt dy dt = x + a τ y = 1 x 1 τ τ y + b Ẇ τ
Modelling Interactions Between Weather and Climate p. 17/33 Relation to MTV Theory" Have assumed that influence of weather on climate is bounded as τ 0 MTV" theory considers averaging assuming that weather influence strengthens in this limit An example: dx dt dy dt = x + a τ y = E y x {y} = τx, std y x = b 2 /2 1 x 1 τ τ y + b Ẇ τ
Modelling Interactions Between Weather and Climate p. 17/33 Relation to MTV Theory" Have assumed that influence of weather on climate is bounded as τ 0 MTV" theory considers averaging assuming that weather influence strengthens in this limit An example: dx dt dy dt = x + a τ y = E y x {y} = τx, std y x = b 2 /2 1 x 1 τ τ y + b Ẇ τ Stochastic terms remain in the strict τ 0 limit dx dt = (1 a)x + abẇ
Modelling Interactions Between Weather and Climate p. 17/33 Relation to MTV Theory" Have assumed that influence of weather on climate is bounded as τ 0 MTV" theory considers averaging assuming that weather influence strengthens in this limit An example: dx dt dy dt = x + a τ y = E y x {y} = τx, std y x = b 2 /2 1 x 1 τ τ y + b Ẇ τ Stochastic terms remain in the strict τ 0 limit dx = (1 a)x + abẇ dt Present analysis will not make MTV ansatz of increasing weather influence as τ decreases; rather than a τ = 0 theory", it is a small τ theory"
Modelling Interactions Between Weather and Climate p. 18/33 Case study 1: Coupled atmosphere-ocean boundary layers Atmosphere and ocean interact through (generally turbulent) boundary layers, exchanging mass, momentum, and energy
Modelling Interactions Between Weather and Climate p. 18/33 Case study 1: Coupled atmosphere-ocean boundary layers Atmosphere and ocean interact through (generally turbulent) boundary layers, exchanging mass, momentum, and energy Idealised coupled model of atmospheric winds and air/sea temperatures: d dt u = Π u c d(w) h(t a, T o ) wu d dt v = c d(w) h(t a, T o ) wv d w e h(t a, T o ) u + σ uẇ1 w e h(t a, T o ) v + σ uẇ2 dt T a = β (T o T a )w β θ(w µ w ) λ a T a + Σ 11 Ẇ 3 + Σ 12 Ẇ 4 γ a γ a γ a d dt T o = β (T a T o )w + β θ(w µ w ) λ o T o + Σ 21 Ẇ 3 + Σ 22 Ẇ 4 γ o γ o γ o
Modelling Interactions Between Weather and Climate p. 19/33 Case study 1: Coupled atmosphere-ocean boundary layers 60 o N 54 o N 48 o N OSP 42 o N 36 o N 30 o N 165 o E 180 o W 165 o W 150 o W 135 o W 120 o W Ocean station P
Modelling Interactions Between Weather and Climate p. 20/33 Case study 1: Coupled atmosphere-ocean boundary layers Atmosphere and ocean interact through (generally turbulent) boundary layers, exchanging mass, momentum, and energy Idealised coupled model of atmospheric winds and air/sea temperatures: d dt u = Π u c d(w) h(t a, T o ) wu d dt v = c d(w) h(t a, T o ) wv d w e h(t a, T o ) u + σ uẇ1 w e h(t a, T o ) v + σ uẇ2 dt T a = β (T o T a )w β θ(w µ w ) λ a T a + Σ 11 Ẇ 3 + Σ 12 Ẇ 4 γ a γ a γ a d dt T o = β (T a T o )w + β θ(w µ w ) λ o T o + Σ 21 Ẇ 3 + Σ 22 Ẇ 4 γ o γ o γ o Atmospheric boundary layer thickness determined by surface stratification h(t a, T o ) = max [ h min, h(1 α(t a T o )) ]
Modelling Interactions Between Weather and Climate p. 21/33 Case study 1: Coupled atmosphere-ocean boundary layers 1100 1000 900 h (m) 800 700 600 500 400 3 2 1 0 1 2 3 T a T o (K) Boundary layer height and surface stability
Case study 1: Coupled atmosphere-ocean boundary layers correlation 1 0.5 0 0 20 40 60 80 lag (days) T a (obs) T a (sim) T o (obs) T o (sim) T o (K) 2 1 0 1 2 4 2 0 2 4 T a (K) Obs Sim 0.8 Obs Sim 0.1 0.08 Obs Sim correlation 0.6 0.4 0.2 p(w) 0.06 0.04 0 0.02 0.2 0 2 4 6 lag (days) 0 0 10 20 30 w (m/s) Observed and simulated statistics (Ocean Station P) Modelling Interactions Between Weather and Climate p. 22/33
Modelling Interactions Between Weather and Climate p. 23/33 Case study 1: Coupled atmosphere-ocean boundary layers Observed timescale separations: τ w τ Ta 0.7
Modelling Interactions Between Weather and Climate p. 23/33 Case study 1: Coupled atmosphere-ocean boundary layers Observed timescale separations: τ w τ Ta 0.7 Take x = T = (T a,t o ), y = (u,v) with
Case study 1: Coupled atmosphere-ocean boundary layers Observed timescale separations: τ w τ Ta 0.7 Take x = T = (T a,t o ), y = (u,v) with Averaging model (A) 5 f 1 ( K / s ) d T = f(t) + ΣẆ dt 5 f 2 ( K / s ) T o (K) 0 0 5 5 0 5 T a (K) 5 5 0 5 T a (K) 5 0 5 10 15 x 10 5 5 0 5 x 10 6 Modelling Interactions Between Weather and Climate p. 23/33
Modelling Interactions Between Weather and Climate p. 24/33 Case study 1: Coupled atmosphere-ocean boundary layers Analytic expression for σ(t a, T o ) σ(t a, T o ) = β γ 2 a + γ 2 o γ o γ a 1 1 γ a γ o ψ(t a T o )
Modelling Interactions Between Weather and Climate p. 24/33 Case study 1: Coupled atmosphere-ocean boundary layers Analytic expression for σ(t a, T o ) σ(t a, T o ) = β γ 2 a + γ 2 o γ o γ a 1 1 γ a γ o ψ(t a T o ) where ψ(t a T o ) = (T a T o + θ) E {w (t + s)w (t)} ds
Case study 1: Coupled atmosphere-ocean boundary layers Analytic expression for σ(t a, T o ) σ(t a, T o ) = β γ 2 a + γ 2 o γ o γ a 1 1 γ a γ o ψ(t a T o ) where ψ(t a T o ) = (T a T o + θ) E {w (t + s)w (t)} ds 0.5 ψ (K m s 1/2 10 4 ) 0 0.5 1 1.5 2 2.5 10 5 0 T a T o (K) 5 10 Modelling Interactions Between Weather and Climate p. 24/33
Modelling Interactions Between Weather and Climate p. 25/33 Case study 1: Coupled atmosphere-ocean boundary layers correlation correlation 1 0.5 0 1 0.5 0 0 20 40 60 80 lag (days) T (reduced) a T (full) a T o (reduced) T (full) o 0 20 40 60 80 lag (days) T (reduced) a T (full) a T o (reduced) T (full) o T o (K) T o (K) 2 1 0 1 2 4 2 0 2 4 2 1 0 1 T a (K) T a (K) Reduced Full Reduced Full 2 4 2 0 2 4 τ = 0.7 τ = 0.35
Modelling Interactions Between Weather and Climate p. 26/33 Case study 1: Coupled atmosphere-ocean boundary layers Sura and Newman (2008) fit observations at OSP to linear multiplicative SDE d T = AT + B(T) Ẇ dt
Modelling Interactions Between Weather and Climate p. 26/33 Case study 1: Coupled atmosphere-ocean boundary layers Sura and Newman (2008) fit observations at OSP to linear multiplicative SDE d T = AT + B(T) Ẇ dt Form of this SDE justified through simple substitution in full coupled dynamics w µ w + σẇ
Modelling Interactions Between Weather and Climate p. 26/33 Case study 1: Coupled atmosphere-ocean boundary layers Sura and Newman (2008) fit observations at OSP to linear multiplicative SDE d T = AT + B(T) Ẇ dt Form of this SDE justified through simple substitution in full coupled dynamics w µ w + σẇ Form of SDE did not account for feedback of stratification on h; physical origin of parameters in inverse model somewhat different than had been assumed
Modelling Interactions Between Weather and Climate p. 27/33 Case study 2: extratropical atmospheric LFV Variability of extratropical atmosphere on timescales of 10 days + descrbed as low-frequency variability (LFV)
Modelling Interactions Between Weather and Climate p. 27/33 Case study 2: extratropical atmospheric LFV Variability of extratropical atmosphere on timescales of 10 days + descrbed as low-frequency variability (LFV) Associated with hemispheric to global spatial scales
Modelling Interactions Between Weather and Climate p. 27/33 Case study 2: extratropical atmospheric LFV Variability of extratropical atmosphere on timescales of 10 days + descrbed as low-frequency variability (LFV) Associated with hemispheric to global spatial scales Northern Annular Mode (NAM) From www.whoi.edu
Modelling Interactions Between Weather and Climate p. 27/33 Case study 2: extratropical atmospheric LFV Variability of extratropical atmosphere on timescales of 10 days + descrbed as low-frequency variability (LFV) Associated with hemispheric to global spatial scales Northern Annular Mode (NAM) Basic structures characterised statistically; dynamical questions remain open From www.whoi.edu
Modelling Interactions Between Weather and Climate p. 27/33 Case study 2: extratropical atmospheric LFV Variability of extratropical atmosphere on timescales of 10 days + descrbed as low-frequency variability (LFV) Associated with hemispheric to global spatial scales Northern Annular Mode (NAM) Basic structures characterised statistically; dynamical questions remain open Empirical stochastic models have been useful for studying LFV From www.whoi.edu
Modelling Interactions Between Weather and Climate p. 27/33 Case study 2: extratropical atmospheric LFV Variability of extratropical atmosphere on timescales of 10 days + descrbed as low-frequency variability (LFV) Associated with hemispheric to global spatial scales Northern Annular Mode (NAM) From www.whoi.edu Basic structures characterised statistically; dynamical questions remain open Empirical stochastic models have been useful for studying LFV Stochastic reduction techniques natural tool for investigating dynamics
Case study 2: extratropical atmospheric LFV Kravtsov et al. JAS 2005 Modelling Interactions Between Weather and Climate p. 28/33
Modelling Interactions Between Weather and Climate p. 29/33 Case study 2: extratropical atmospheric LFV Using PCA decomposition, identify 2 slow modes: propagating wavenumber 4, stationary zonal
Modelling Interactions Between Weather and Climate p. 29/33 Case study 2: extratropical atmospheric LFV Using PCA decomposition, identify 2 slow modes: propagating wavenumber 4, stationary zonal Kravtsov et al. (2005) concluded that regime behaviour resulted from nonlinear interaction of these modes
Modelling Interactions Between Weather and Climate p. 29/33 Case study 2: extratropical atmospheric LFV Using PCA decomposition, identify 2 slow modes: propagating wavenumber 4, stationary zonal Kravtsov et al. (2005) concluded that regime behaviour resulted from nonlinear interaction of these modes Consider reduction to 3-variable, 1-variable (stationary mode) models
Case study 2: extratropical atmospheric LFV Modelling Interactions Between Weather and Climate p. 30/33
Case study 2: extratropical atmospheric LFV Modelling Interactions Between Weather and Climate p. 31/33
Modelling Interactions Between Weather and Climate p. 31/33 Case study 2: extratropical atmospheric LFV Stationary mode potential shows prominent shoulder
Modelling Interactions Between Weather and Climate p. 31/33 Case study 2: extratropical atmospheric LFV Stationary mode potential shows prominent shoulder In this model, appears that multiple regimes produced by nonlinearity in effective dynamics of stationary mode
Modelling Interactions Between Weather and Climate p. 31/33 Case study 2: extratropical atmospheric LFV Stationary mode potential shows prominent shoulder In this model, appears that multiple regimes produced by nonlinearity in effective dynamics of stationary mode Non-gaussianity doesn t require state-dependent noise
Modelling Interactions Between Weather and Climate p. 32/33 Case study 2: extratropical atmospheric LFV MTV approximations with minimal regression tuning
Modelling Interactions Between Weather and Climate p. 33/33 Conclusions Stochastic reduction tools allow systematic derivation of low-order climate models from coupled weather/climate systems
Modelling Interactions Between Weather and Climate p. 33/33 Conclusions Stochastic reduction tools allow systematic derivation of low-order climate models from coupled weather/climate systems Reduction approach is straightforward, although implementation in high-dimensional systems challenging
Modelling Interactions Between Weather and Climate p. 33/33 Conclusions Stochastic reduction tools allow systematic derivation of low-order climate models from coupled weather/climate systems Reduction approach is straightforward, although implementation in high-dimensional systems challenging This approach provides a more general approach than e.g. MTV theory, with
Modelling Interactions Between Weather and Climate p. 33/33 Conclusions Stochastic reduction tools allow systematic derivation of low-order climate models from coupled weather/climate systems Reduction approach is straightforward, although implementation in high-dimensional systems challenging This approach provides a more general approach than e.g. MTV theory, with benefit of less restrictive assumptions
Modelling Interactions Between Weather and Climate p. 33/33 Conclusions Stochastic reduction tools allow systematic derivation of low-order climate models from coupled weather/climate systems Reduction approach is straightforward, although implementation in high-dimensional systems challenging This approach provides a more general approach than e.g. MTV theory, with benefit of less restrictive assumptions drawbacks of more difficult implementation, lack of closed-form solution
Modelling Interactions Between Weather and Climate p. 33/33 Conclusions Stochastic reduction tools allow systematic derivation of low-order climate models from coupled weather/climate systems Reduction approach is straightforward, although implementation in high-dimensional systems challenging This approach provides a more general approach than e.g. MTV theory, with benefit of less restrictive assumptions drawbacks of more difficult implementation, lack of closed-form solution Real systems do not generally have a strong scale separation; an important outstanding problem is a more systematic approach to accounting for this