42
Spatio-temporal stochastic models with embedded deterministic dynamics Krzysztof Podgorski 28th April 2011 [email protected]

Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Spatio-temporal stochastic models withembedded deterministic dynamics

Krzysztof Podgorski

28th April 2011

[email protected]

Page 2: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What do we want to do?

• In search: a mathematical model X (p, t) that accounts bothfor motion and stochastic variability.

[email protected]

Page 3: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What do we want to do?

• In search: a mathematical model X (p, t) that accounts bothfor motion and stochastic variability.

[email protected]

Page 4: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What was done? or How to move around stochastics?

• Stochastic differential equations:

L(X , ∂X/∂p, ∂X/∂t,p, t) = dW (p, t)

• Dispersion relations encoded in spectrum:

X (p, t) =∑ω

s(ω) · cos(ω2

gp + ωt

)

[email protected]

Page 5: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What was done? or How to move around stochastics?

• Stochastic differential equations:

L(X , ∂X/∂p, ∂X/∂t,p, t) = dW (p, t)

• Dispersion relations encoded in spectrum:

X (p, t) =∑ω

s(ω) · cos(ω2

gp + ωt

)

[email protected]

Page 6: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What do we do here?

Essentially, we combine together well-established components:

• Deterministic Model- Physical phenomenon- Deterministic flow

• Stochastic Model- Covariance structures- ’Static’ stochastic flow

• Stochastic-Deterministic Model- Embedding deterministic flow into static stochastic flow- Dynamical stochastic flow

[email protected]

Page 7: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What do we do here?

Essentially, we combine together well-established components:

• Deterministic Model- Physical phenomenon- Deterministic flow

• Stochastic Model- Covariance structures- ’Static’ stochastic flow

• Stochastic-Deterministic Model- Embedding deterministic flow into static stochastic flow- Dynamical stochastic flow

[email protected]

Page 8: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What do we do here?

Essentially, we combine together well-established components:

• Deterministic Model- Physical phenomenon- Deterministic flow

• Stochastic Model- Covariance structures- ’Static’ stochastic flow

• Stochastic-Deterministic Model- Embedding deterministic flow into static stochastic flow- Dynamical stochastic flow

[email protected]

Page 9: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

What do we do here?

Essentially, we combine together well-established components:

• Deterministic Model- Physical phenomenon- Deterministic flow

• Stochastic Model- Covariance structures- ’Static’ stochastic flow

• Stochastic-Deterministic Model- Embedding deterministic flow into static stochastic flow- Dynamical stochastic flow

[email protected]

Page 10: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Motivation

A need for models that account for:• Variability in time - non-trivial, physically meaningful dynamics• Stochastic variability

One approach is to have a differential equation that is driven by astochastic noise.We propose a reverse:

Stochastic field driven by deterministic flow.

[email protected]

Page 11: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Motivation

A need for models that account for:• Variability in time - non-trivial, physically meaningful dynamics• Stochastic variability

One approach is to have a differential equation that is driven by astochastic noise.We propose a reverse:

Stochastic field driven by deterministic flow.

[email protected]

Page 12: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Motivation

A need for models that account for:• Variability in time - non-trivial, physically meaningful dynamics• Stochastic variability

One approach is to have a differential equation that is driven by astochastic noise.We propose a reverse:

Stochastic field driven by deterministic flow.

[email protected]

Page 13: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Motivation

A need for models that account for:• Variability in time - non-trivial, physically meaningful dynamics• Stochastic variability

One approach is to have a differential equation that is driven by astochastic noise.We propose a reverse:

Stochastic field driven by deterministic flow.

[email protected]

Page 14: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Motivation

A need for models that account for:• Variability in time - non-trivial, physically meaningful dynamics• Stochastic variability

One approach is to have a differential equation that is driven by astochastic noise.We propose a reverse:

Stochastic field driven by deterministic flow.

[email protected]

Page 15: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Is a stochastic field moving?

[email protected]

Page 16: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Is a stochastic field moving?

0 100 200 300 400 5000

50

100

150

200

250

300

350

400

450

500

y, [m

]

x, [m]

W(x,y,4)

0 100 200 300 400 5000

50

100

150

200

250

300

350

400

450

500

x, [m]

y, [m

]

W(x,y,0)

[email protected]

Page 17: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Measuring velocities of a moving surface

20 25 30 35 40 45 50 55 6040

45

50

55

60

65

70

75

80

85 90 95 100 10510

12

14

16

18

20

22

24

26

28

30

45 50 55 6050

55

60

65

70

75

30 35 40 45 50 55 60 6565

70

75

80

85

90

95

100

105

[email protected]

Page 18: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Stochastic Velocity Field

• One of possible definitions of velocity on random surfaces

v(x , t) = −Xt(x , t)

Xx(x , t)= − ∂X (x , t)t

∂X (x , t)x,

0 50 100 1500

50

100

150

[email protected]

Page 19: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Spatial Static Model

The starting point is the following spectral representation of astationary process

X (p)d=

∫Rn

exp(ip · ω)√

S(ω) dB(ω),

Non-stationary extension:

X (p)d=

∫Rn

exp(ip · ω)√

Sp(ω) dB(ω).

The covariance of X is given by

rS(p,p′) =

∫Rn

exp(i(p− p′) · ω)√

Sp(ω)Sp′(ω) dω,

and is non-stationary.

[email protected]

Page 20: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Spatial Static Model

The starting point is the following spectral representation of astationary process

X (p)d=

∫Rn

exp(ip · ω)√

S(ω) dB(ω),

Non-stationary extension:

X (p)d=

∫Rn

exp(ip · ω)√

Sp(ω) dB(ω).

The covariance of X is given by

rS(p,p′) =

∫Rn

exp(i(p− p′) · ω)√

Sp(ω)Sp′(ω) dω,

and is non-stationary.

[email protected]

Page 21: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Spatial Static Model

The starting point is the following spectral representation of astationary process

X (p)d=

∫Rn

exp(ip · ω)√

S(ω) dB(ω),

Non-stationary extension:

X (p)d=

∫Rn

exp(ip · ω)√

Sp(ω) dB(ω).

The covariance of X is given by

rS(p,p′) =

∫Rn

exp(i(p− p′) · ω)√

Sp(ω)Sp′(ω) dω,

and is non-stationary.

[email protected]

Page 22: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example

Take

Sp(ω) =s2(p)Ln(p)

2πn/2 exp(−L2(p)|ω|2/2

),

The non-stationary covariance of X is given by

rS(p,p′) = s(p)s(p′)(

2L(p)L(p′)L2(p) + L2(p′)

)− n2

exp(−(p− p′)TΣ(p− p′)

2

)where Σ = Σ(p,p′) = 2/

(L2(p) + L2(p′)

)· I.

[email protected]

Page 23: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example

Take

Sp(ω) =s2(p)Ln(p)

2πn/2 exp(−L2(p)|ω|2/2

),

The non-stationary covariance of X is given by

rS(p,p′) = s(p)s(p′)(

2L(p)L(p′)L2(p) + L2(p′)

)− n2

exp(−(p− p′)TΣ(p− p′)

2

)where Σ = Σ(p,p′) = 2/

(L2(p) + L2(p′)

)· I.

[email protected]

Page 24: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Spatio-temporal static fields

Assuming that Φ(·; ds) is a Gaussian field-valued measure that isuniquely characterized by the time dependent spatial covariancesrS(p,p′; s), we can define

X (p, t) =

∫f (t, s; p) Φ(p; ds),

which will have the covariance

r(p,p′; t, t ′) =

∫f (t, s; p)f (t ′, s; p′) · rS(p,p′; s) ds

[email protected]

Page 25: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Spatio-temporal static fields

Assuming that Φ(·; ds) is a Gaussian field-valued measure that isuniquely characterized by the time dependent spatial covariancesrS(p,p′; s), we can define

X (p, t) =

∫f (t, s; p) Φ(p; ds),

which will have the covariance

r(p,p′; t, t ′) =

∫f (t, s; p)f (t ′, s; p′) · rS(p,p′; s) ds

[email protected]

Page 26: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example: Temporal Ornstein-Uhlenbeck field

Taking f (t) = e−λt1[0,∞)(t) gives

X (p, t) =

∫ t

−∞e−λ(t−s) Φ(p; ds)

and if rS(p,p′; s) = rS(p,p′), its covariance is given by

r(p,p′; t) = rS(p,p′) · 12λ

e−λ|t|.

This example corresponds to the autoregression model of order one

X (p, t) = ρX (p, t −∆t) +√

1− ρ2 Φt(p),

Taking a space dependent λ(p) gives

r(p,p′; t) =rS(p,p′)

λ(p) + λ(p′)

{e−λ(p′)·t ; if t > 0,e−λ(p)·t ; if t < 0.

[email protected]

Page 27: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example: Temporal Ornstein-Uhlenbeck field

Taking f (t) = e−λt1[0,∞)(t) gives

X (p, t) =

∫ t

−∞e−λ(t−s) Φ(p; ds)

and if rS(p,p′; s) = rS(p,p′), its covariance is given by

r(p,p′; t) = rS(p,p′) · 12λ

e−λ|t|.

This example corresponds to the autoregression model of order one

X (p, t) = ρX (p, t −∆t) +√

1− ρ2 Φt(p),

Taking a space dependent λ(p) gives

r(p,p′; t) =rS(p,p′)

λ(p) + λ(p′)

{e−λ(p′)·t ; if t > 0,e−λ(p)·t ; if t < 0.

[email protected]

Page 28: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example: Temporal Ornstein-Uhlenbeck field

Taking f (t) = e−λt1[0,∞)(t) gives

X (p, t) =

∫ t

−∞e−λ(t−s) Φ(p; ds)

and if rS(p,p′; s) = rS(p,p′), its covariance is given by

r(p,p′; t) = rS(p,p′) · 12λ

e−λ|t|.

This example corresponds to the autoregression model of order one

X (p, t) = ρX (p, t −∆t) +√

1− ρ2 Φt(p),

Taking a space dependent λ(p) gives

r(p,p′; t) =rS(p,p′)

λ(p) + λ(p′)

{e−λ(p′)·t ; if t > 0,e−λ(p)·t ; if t < 0.

[email protected]

Page 29: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example: Temporal Ornstein-Uhlenbeck field

Taking f (t) = e−λt1[0,∞)(t) gives

X (p, t) =

∫ t

−∞e−λ(t−s) Φ(p; ds)

and if rS(p,p′; s) = rS(p,p′), its covariance is given by

r(p,p′; t) = rS(p,p′) · 12λ

e−λ|t|.

This example corresponds to the autoregression model of order one

X (p, t) = ρX (p, t −∆t) +√

1− ρ2 Φt(p),

Taking a space dependent λ(p) gives

r(p,p′; t) =rS(p,p′)

λ(p) + λ(p′)

{e−λ(p′)·t ; if t > 0,e−λ(p)·t ; if t < 0.

[email protected]

Page 30: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Is there non-trivial dynamics?

• If

r(p,p′; t, t ′) =

∫ ∞−∞

f (t, s) · f (t ′, s) · rS(p,p′; s) ds.

then the dynamics of the field is trivial (i.e. velocities arecentered at zero).

• Thus the field

X (p, t) =

∫f (t, s) Φ(p; ds),

does not exhibit any organized motion.

[email protected]

Page 31: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Is there non-trivial dynamics?

• If

r(p,p′; t, t ′) =

∫ ∞−∞

f (t, s) · f (t ′, s) · rS(p,p′; s) ds.

then the dynamics of the field is trivial (i.e. velocities arecentered at zero).

• Thus the field

X (p, t) =

∫f (t, s) Φ(p; ds),

does not exhibit any organized motion.

[email protected]

Page 32: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Feeding deterministic flow into stochastic field

• Flow ψt,h(p) obtained from a velocity field v(p, t) satisfyingthe transport equation

ψt,h(p) = p+

∫ t+h

tv(ψt,u−t(p), u) du = p+

∫ h

0v(ψt,s(p), t+s) ds,

Construction of the stochastic field at a fixed location p and afixed time t:

Y (p, t) =

∫ ∞−∞

f (t, s) Φ(ψt,s−t(p); ds)

[email protected]

Page 33: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

There is a method in the madness

TheoremLet the spatial covariance rS of the innovations Φ(x , y ; dt) beisotropic (i.e. invariant under rotation), then the distribution ofrandom velocities on the surface of Y (p, t) has its center at thevalue of the deterministic velocity field v(x , y , t).

In other words, stochastic dynamics follows the one represented bythe underlying deterministic field.

[email protected]

Page 34: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

There is a method in the madness

TheoremLet the spatial covariance rS of the innovations Φ(x , y ; dt) beisotropic (i.e. invariant under rotation), then the distribution ofrandom velocities on the surface of Y (p, t) has its center at thevalue of the deterministic velocity field v(x , y , t).

In other words, stochastic dynamics follows the one represented bythe underlying deterministic field.

[email protected]

Page 35: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example: Dynamics driven by shallow water equations

∂u∂t

= −g∂h∂x− u

∂u∂x− v

∂u∂y

+ fv

∂v∂t

= −g∂h∂y− u

∂v∂x− v

∂v∂y− fu

∂h∂t

= −∂(hu)

∂x− ∂(hv)

∂y

• where: u, v denote horizontal velocities, h surface elevation• Used to model: Tsunamis, flows in rivers, internal waves,Jupiter’s Atmopshere

[email protected]

Page 36: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Example cont.: Linearization in one dimension

After linearization and reduction to one dimension.• PDE with constant coefficients:

∂u∂t

= −u0∂u∂x− g

∂h∂x

∂h∂t

= −u0∂h∂x− h0

∂u∂x

,

whereu0, h0, g are constants

[email protected]

Page 37: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

1-dim SWE simulation

FILM

[email protected]

Page 38: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Theorem – figures

Figure: Stochastically distorted flow (left) and deterministic shallowwater flow (right).

[email protected]

Page 39: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

A very incomplete and terribly biased bibliography

• Baxevani, A., Borget, C., Rychlik, I. (2008) Spatial Models for the Variability of the SignificantWave Height on the World Oceans.

• Baxevani, A., Podgórski, K., Rychlik, I. (2003) Velocities for moving random surfaces,• Baxevani, A., Podgórski, K., Rychlik, I. (2011) Dynamically evolving Gaussian spatial fields.• Cox, D.R., Isham, V.S. (1988) A simple spatial-temporal model of rainfall.• Gupta, V.K., Waymire, E. (1987) On Taylor’s hypothesis and dissipation in rainfall.• Longuet-Higgins, M. S (1957) The statistical analysis of a random, moving surface.• Porcu, E., Mateu, J., and Christakos, G., (2009) Quasi-arithmetic means of covariance functions

with potential applications to space-time data.• Schlather, M. (2009) Some covariance models based on normal scale mixtures.• Stein, M.L. (2005) Nonstationary spatial covariance functions• Wegener, J. (2010) Ph.D. Thesis, Lund University.

[email protected]

Page 40: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Final Slide

What does Mother Earth think of our methods?

[email protected]

Page 41: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Final Slide

What does Mother Earth think of our methods?

We asked...

[email protected]

Page 42: Spatio-temporal stochastic models with embedded ... · Isastochasticfieldmoving? 0 100 200 300 400 500 0 50 100 150 200 250 300 350 400 450 500 y, [m] x, [m] W(x,y,4) 0 100 200 300

Final Slide

[email protected]