Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction...
Transcript of Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction...
![Page 1: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/1.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Regularization in Variational Data Assimilation
Melina Freitag
Department of Mathematical Sciences
University of Bath
ICIAM 2011, VancouverMinisymposium MS49: Variational Data Assimilation
18th July 2011
joint work with C.J. Budd (Bath) and N.K. Nichols (Reading)
Melina Freitag Regularization in Variational Data Assimilation
![Page 2: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/2.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Weather forecast for today
Melina Freitag Regularization in Variational Data Assimilation
![Page 3: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/3.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
The MetOffice weather forecast for today
Melina Freitag Regularization in Variational Data Assimilation
![Page 4: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/4.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Introduction
4DVar and Tikhonov regularisation
Application of L1-norm regularisation in 4DVarMotivation: Results from image processingL1-norm regularisation in 4DVar
Examples
Melina Freitag Regularization in Variational Data Assimilation
![Page 5: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/5.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Outline
Introduction
4DVar and Tikhonov regularisation
Application of L1-norm regularisation in 4DVarMotivation: Results from image processingL1-norm regularisation in 4DVar
Examples
Melina Freitag Regularization in Variational Data Assimilation
![Page 6: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/6.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Data Assimilation in NWP
Find an estimate xi at time i for the true state of the atmosphere xTruthi .
Observations yi
• Satellites
• Ships and buoys
• Surface stations
• Aeroplanes
Melina Freitag Regularization in Variational Data Assimilation
![Page 7: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/7.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Data Assimilation in NWP
Find an estimate xi at time i for the true state of the atmosphere xTruthi .
A priori information xB
i
• background state (previousforecast)
Observations yi
• Satellites
• Ships and buoys
• Surface stations
• Aeroplanes
Melina Freitag Regularization in Variational Data Assimilation
![Page 8: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/8.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Data Assimilation in NWP
Find an estimate xi at time i for the true state of the atmosphere xTruthi .
A priori information xB
i
• background state (previousforecast)
Models
• an operator linking state space andobservation space (imperfect)
yi = Hi(xi)
Observations yi
• Satellites
• Ships and buoys
• Surface stations
• Aeroplanes
Melina Freitag Regularization in Variational Data Assimilation
![Page 9: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/9.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Data Assimilation in NWP
Find an estimate xi at time i for the true state of the atmosphere xTruthi .
A priori information xB
i
• background state (previousforecast)
Models
• an operator linking state space andobservation space (imperfect)
yi = Hi(xi)
• a model for the atmosphere(imperfect)
xi+1 = Mi+1,i(xi)
Observations yi
• Satellites
• Ships and buoys
• Surface stations
• Aeroplanes
Melina Freitag Regularization in Variational Data Assimilation
![Page 10: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/10.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Data Assimilation in NWP
Find an estimate xi at time i for the true state of the atmosphere xTruthi .
A priori information xB
i
• background state (previousforecast)
Models
• an operator linking state space andobservation space (imperfect)
yi = Hi(xi)
• a model for the atmosphere(imperfect)
xi+1 = Mi+1,i(xi)
Observations yi
• Satellites
• Ships and buoys
• Surface stations
• Aeroplanes
Assimilation algorithms
• find an (approximate) state of theatmosphere xi at times i (usuallyi = 0)
• xAi : Analysis (estimation of the
true state after the DA)
• forecast future states of theatmosphere
Melina Freitag Regularization in Variational Data Assimilation
![Page 11: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/11.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Observations
Melina Freitag Regularization in Variational Data Assimilation
![Page 12: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/12.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Schematics of Data Assimilation
Figure: Background state xB
Melina Freitag Regularization in Variational Data Assimilation
![Page 13: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/13.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Schematics of Data Assimilation
Figure: Observations y
Melina Freitag Regularization in Variational Data Assimilation
![Page 14: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/14.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Schematics of Data Assimilation
Figure: Analysis xA (consistent with observations and model dynamics)
Melina Freitag Regularization in Variational Data Assimilation
![Page 15: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/15.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Data Assimilation in NWP
Under-determinacy
• Size of the state vector x: 432 × 320 × 50 × 7 = O(107)
• Number of observations (size of y): O(105 − 106)
Melina Freitag Regularization in Variational Data Assimilation
![Page 16: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/16.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Data Assimilation in NWP
Under-determinacy
• Size of the state vector x: 432 × 320 × 50 × 7 = O(107)
• Number of observations (size of y): O(105 − 106)
Assumptions
• background error εB = xB − xTruth and covariance matrix
B = (εB − εB)(εB − εB)T
• observation error εO = y − H(xTruth) and covariance matrix
R = (εO − εO)(εO − εO)T
• Non-trivial errors: B, R are positive definite
• Unbiased errors: xB − xTruth = y − H(xTruth) = 0
• Uncorrelated errors: (xB − xTruth)(y − H(xTruth))T = 0
Melina Freitag Regularization in Variational Data Assimilation
![Page 17: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/17.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Optimal least-squares estimator
Cost functionSolution to the optimisation problem xA = arg minJ(x) where
J(x) =1
2(x − xB)T B−1(x − xB) +
1
2(y − H(x))T R−1(y − H(x))
= JB(x) + JO(x)
⇒Three-dimensional variational data assimilation (3DVar)
Melina Freitag Regularization in Variational Data Assimilation
![Page 18: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/18.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Optimal least-squares estimator
Cost functionSolution to the optimisation problem xA = arg minJ(x) where
J(x) =1
2(x − xB)T B−1(x − xB) +
1
2(y − H(x))T R−1(y − H(x))
= JB(x) + JO(x)
⇒Three-dimensional variational data assimilation (3DVar)
Interpolation equations
xA = xB + K(y − H(xB)), where
K = BHT (HBHT + R)−1 K . . . gain matrix
⇒ Optimal interpolation
Melina Freitag Regularization in Variational Data Assimilation
![Page 19: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/19.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Four-dimensional variational assimilation (4DVar)
Minimise the cost function
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2
nX
i=0
(yi − Hi(xi))T R
−1
i (yi − Hi(xi))
subject to model dynamics xi = M0→ix0.
Figure: Copyright:ECMWF
Melina Freitag Regularization in Variational Data Assimilation
![Page 20: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/20.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Four-dimensional variational assimilation (4DVar)
Minimise the cost function
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2
nX
i=0
(yi − Hi(xi))T R
−1
i (yi − Hi(xi))
subject to model dynamics xi = M0→ix0.
Figure: Copyright:ECMWF
Melina Freitag Regularization in Variational Data Assimilation
![Page 21: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/21.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Four-dimensional variational assimilation (4DVar)
Minimise the cost function
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2
nX
i=0
(yi − Hi(xi))T R
−1
i (yi − Hi(xi))
subject to model dynamics xi = M0→ix0.
Figure: Copyright:ECMWF
Melina Freitag Regularization in Variational Data Assimilation
![Page 22: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/22.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Minimisation of the 4DVar cost function
• Use Newton’s method in order to solve ∇J(x0) = 0, that is
∇∇J(xk0)∆xk
0 = −∇J(xk0)
xk+10
= xk0 + ∆xk
0
k ≥ 0
Melina Freitag Regularization in Variational Data Assimilation
![Page 23: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/23.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Minimisation of the 4DVar cost function
• Use Newton’s method in order to solve ∇J(x0) = 0, that is
∇∇J(xk0)∆xk
0 = −∇J(xk0)
xk+10
= xk0 + ∆xk
0
k ≥ 0
• Use approximate Hessian - Gauß-Newton method
∇J(x0) = B−1(x0 − xB0 ) −
nX
i=1
Mi,0(x0)T HTi R
−1
i (yi − Hi(xi)),
and
∇∇J(x0) = B−1 +n
X
i=1
Mi,0(x0)T HTi R
−1
i HiMi,0(x0).
Melina Freitag Regularization in Variational Data Assimilation
![Page 24: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/24.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Outline
Introduction
4DVar and Tikhonov regularisation
Application of L1-norm regularisation in 4DVarMotivation: Results from image processingL1-norm regularisation in 4DVar
Examples
Melina Freitag Regularization in Variational Data Assimilation
![Page 25: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/25.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
4DVar minimises
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2
nX
i=0
(yi − Hi(xi))T R
−1
i (yi − Hi(xi))
subject to model dynamics xi = Mi0x0
Melina Freitag Regularization in Variational Data Assimilation
![Page 26: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/26.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
4DVar minimises
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2
nX
i=0
(yi − Hi(xi))T R
−1
i (yi − Hi(xi))
subject to model dynamics xi = Mi0x0
or
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2(y − H(x0))T R−1(y − H(x0))
whereH = [HT
0 , (H1M10(t1, t0))T , . . . (HnMn0(tn, t0))T ]T
y = [yT0 , . . . ,yT
n ]T
and R is block diagonal with Ri, i = 0, . . . , n on the diagonal.
Melina Freitag Regularization in Variational Data Assimilation
![Page 27: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/27.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Solution to the optimisation problemCost function
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2(y − H(x0))T R−1(y − H(x0))
Gauß-Newton method
∇∇J(xk0)∆xk
0 = −∇J(xk0)
xk+10
= xk0 + ∆xk
0
Melina Freitag Regularization in Variational Data Assimilation
![Page 28: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/28.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Solution to the optimisation problemCost function
J(x0) =1
2(x0 − xB
0 )T B−1(x0 − xB0 ) +
1
2(y − H(x0))T R−1(y − H(x0))
Gauß-Newton method
(B−1 + HT R−1H)∆xk0 = −B−1(xk
0 − xB0 ) + HT R−1(y − H(x0))
xk+10
= xk0 + ∆xk
0
Melina Freitag Regularization in Variational Data Assimilation
![Page 29: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/29.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Variable transformSet
B = σ2BCB
R = σ2RCR
b = C− 1
2R (y − H(x0)
A = C− 1
2R HC
12B
µ2 =σ2
R
σ2B
Gauß-Newton method
(B−1 + HT R−1H)∆xk0 = −B−1(xk
0 − xB0 ) + HT R−1(y − H(x0))
xk+10
= xk0 + ∆xk
0
Melina Freitag Regularization in Variational Data Assimilation
![Page 30: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/30.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Variable transformSet
B = σ2BCB
R = σ2RCR
b = C− 1
2R (y − H(x0)
A = C− 1
2R HC
12B
µ2 =σ2
R
σ2B
Gauß-Newton method
(µ2I + AT A)C− 1
2B ∆xk
0 = −µ2C− 1
2B (xk
0 − xB0 ) + AT b
xk+10
= xk0 + ∆xk
0
Melina Freitag Regularization in Variational Data Assimilation
![Page 31: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/31.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Variable transformSet
zk = C− 1
2B (xk
0 − xB0 )
Gauß-Newton method
(µ2I + AT A)(zk+1 − zk) = −µ2zk + AT b
Melina Freitag Regularization in Variational Data Assimilation
![Page 32: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/32.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Variable transformSet
zk = C− 1
2B (xk
0 − xB0 )
Gauß-Newton method
(µ2I + AT A)(zk+1 − zk) = −µ2zk + AT b
Normal equations
Melina Freitag Regularization in Variational Data Assimilation
![Page 33: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/33.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Variable transformSet
zk = C− 1
2B (xk
0 − xB0 )
Gauß-Newton method
(µ2I + AT A)(zk+1 − zk) = −µ2zk + AT b
Normal equations
Least squares solution
‚
‚
‚
‚
»
A
µI
–
(zk+1 − zk) +
»
−b
µzk
–‚
‚
‚
‚
2
2
→ min
at each Gauß-Newton method step
Melina Freitag Regularization in Variational Data Assimilation
![Page 34: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/34.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Relation between 4DVar and Tikhonov regularisation
Variable transformSet
zk = C− 1
2B (xk
0 − xB0 )
Gauß-Newton method
(µ2I + AT A)(zk+1 − zk) = −µ2zk + AT b
Normal equations
Least squares solution
‚
‚
‚
‚
»
A
µI
–
(zk+1 − zk) +
»
−b
µzk
–‚
‚
‚
‚
2
2
→ min
at each Gauß-Newton method step or
‖Azk+1 − (Azk + b)‖22 + µ2‖zk+1‖2
2
Tikhonov regularisation
Melina Freitag Regularization in Variational Data Assimilation
![Page 35: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/35.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Ill-posed problemsGiven an operator A we wish to solve
Az = c
it is well-posed if
• solution exits• solution is unique• is stable (A−1 continuous)
Melina Freitag Regularization in Variational Data Assimilation
![Page 36: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/36.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Ill-posed problemsGiven an operator A we wish to solve
Az = c
it is well-posed if
• solution exits• solution is unique• is stable (A−1 continuous)
but ..In finite dimensions existence and uniqueness can be imposed, but
• discrete problem of underlying ill-posed problem becomes ill- conditioned
• singular values of A decay to zero
Melina Freitag Regularization in Variational Data Assimilation
![Page 37: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/37.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Ill-posed problemsGiven an operator A we wish to solve
Az = c
it is well-posed if
• solution exits• solution is unique• is stable (A−1 continuous)
but ..In finite dimensions existence and uniqueness can be imposed, but
• discrete problem of underlying ill-posed problem becomes ill- conditioned
• singular values of A decay to zero
• Tikhonov regularization
z =arg min˘
‖Az − c‖2 + µ2‖z‖2¯
=(AT A + µ2I)−1AT c
=(VΣT UT UΣVT + µ2VVT )−1VΣT UT c
=Vdiag
„
s2i
s2i + µ2
1
si
«
UT c = zµ =n
X
i=1
s2i
s2i + µ2
uTi c
si
vi
Melina Freitag Regularization in Variational Data Assimilation
![Page 38: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/38.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Outline
Introduction
4DVar and Tikhonov regularisation
Application of L1-norm regularisation in 4DVarMotivation: Results from image processingL1-norm regularisation in 4DVar
Examples
Melina Freitag Regularization in Variational Data Assimilation
![Page 39: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/39.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Results from image deblurring: L1 regularisation
Figure: Blurred picture
Melina Freitag Regularization in Variational Data Assimilation
![Page 40: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/40.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Results from image deblurring: L1 regularisation
Figure: Tikhonov regularisation min˘
‖Ax − b‖22 + α‖x‖2
2
¯
Melina Freitag Regularization in Variational Data Assimilation
![Page 41: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/41.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Results from image deblurring: L1 regularisation
Figure: L1-norm regularisation min˘
‖Ax − b‖22 + α‖x‖1
¯
Melina Freitag Regularization in Variational Data Assimilation
![Page 42: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/42.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
3 Regularisation Methods
4DVar
minz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖22
Melina Freitag Regularization in Variational Data Assimilation
![Page 43: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/43.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
3 Regularisation Methods
4DVar
minz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖22
L1-norm regularisation
minz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖1
Melina Freitag Regularization in Variational Data Assimilation
![Page 44: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/44.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
3 Regularisation Methods
4DVar
minz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖22
L1-norm regularisation
minz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖1
Total Variation regularisation
minz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖22 + β‖Dx
k+10
‖1
where xk+10
= C12Bzk+1 + xB
0and D is a matrix approximating the derivative of
the solution.
Melina Freitag Regularization in Variational Data Assimilation
![Page 45: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/45.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
Solveminz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖22
using Least squares and
minz
k+1‖Azk+1 − c‖2
2 + µ2‖zk+1‖22 + β‖Dxk+1
0‖1
using quadratic programming (see Fu/Ng/Nikolova/Barlow 2006).
Melina Freitag Regularization in Variational Data Assimilation
![Page 46: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/46.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
Considerminz
k+1‖Azk+1 − c‖2
2 + β‖Dxk+10
‖1
where xk+10
= C12Bzk+1 + xB
0
Melina Freitag Regularization in Variational Data Assimilation
![Page 47: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/47.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
Considerminz
k+1‖Azk+1 − c‖2
2 + β‖Dxk+10
‖1
where xk+10
= C12Bzk+1 + xB
0
minz
k+1‖Azk+1 − c‖2
2 + β‖DC12Bzk+1 + DxB
0 ‖1
Melina Freitag Regularization in Variational Data Assimilation
![Page 48: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/48.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
Considerminz
k+1‖Azk+1 − c‖2
2 + β‖Dxk+10
‖1
where xk+10
= C12Bzk+1 + xB
0
minz
k+1‖Azk+1 − c‖2
2 + β‖DC12Bzk+1 + DxB
0 ‖1
Set
v = βDC12Bzk+1 + βDxB
0 .
and split v into its positive and negative part:
v = v+ − v−
where
v+ = max(v, 0)
v− = max(−v, 0)
Melina Freitag Regularization in Variational Data Assimilation
![Page 49: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/49.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
With
v = βDC12Bzk+1 + βDxB
0
andv = v+ − v−
the solution to
minz
k+1‖Azk+1 − c‖2
2 + β‖DC12Bzk+1 + DxB
0 ‖1
is equivalent to
Melina Freitag Regularization in Variational Data Assimilation
![Page 50: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/50.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
With
v = βDC12Bzk+1 + βDxB
0
andv = v+ − v−
the solution to
minz
k+1‖Azk+1 − c‖2
2 + β‖DC12Bzk+1 + DxB
0 ‖1
is equivalent to
minz
k+1,v+,v−
n
1T v+ + 1T v− + ‖Azk+1 − c‖22
o
subject to
βDC12Bzk+1 + βDxB
0 = v+ − v−
v+,v− ≥ 0.
Melina Freitag Regularization in Variational Data Assimilation
![Page 51: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/51.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
minz
k+1,v+,v−
n
1T v+ + 1T v− + ‖Azk+1 − c‖22
o
subject to
βDC12Bzk+1 + βDxB
0 = v+ − v−
v+,v− ≥ 0.
or
Melina Freitag Regularization in Variational Data Assimilation
![Page 52: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/52.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Least mixed norm solutions
minz
k+1,v+,v−
n
1T v+ + 1T v− + ‖Azk+1 − c‖22
o
subject to
βDC12Bzk+1 + βDxB
0 = v+ − v−
v+,v− ≥ 0.
or
minw
1
2wT Gw + gT w
ff
subject toEw = e and Fw ≥ 0.
where
G =
2
4
2AT A
0
0
3
5 , g =
2
4
−2AT b
1
1
3
5 , F =
2
4
0
−I
−I
3
5
E =h
βDC12B
−I I
i
w =ˆ
zk+1 v+ v−˜T
e = −βDxB0
Melina Freitag Regularization in Variational Data Assimilation
![Page 53: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/53.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Outline
Introduction
4DVar and Tikhonov regularisation
Application of L1-norm regularisation in 4DVarMotivation: Results from image processingL1-norm regularisation in 4DVar
Examples
Melina Freitag Regularization in Variational Data Assimilation
![Page 54: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/54.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Example 1 - Linear advection equation
ut + uz = 0,
on the interval z ∈ [0, 1], with periodic boundary conditions. The initial solution isa square wave defined by
u(z, 0) =
(
0.5 0.25 < z < 0.5
−0.5 z < 0.25 or z > 0.5.
This wave moves through the time interval, the model equations are defined bythe upwind scheme
Un+1
j = Unj −
∆t
∆z(Un
j − Unj−1),
Un+10
= Un+1
N ,
where j = 1, . . . , N , ∆z = 1
Nand n is the number of time steps. We take
N = 100, ∆t = 0.005.
Melina Freitag Regularization in Variational Data Assimilation
![Page 55: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/55.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Setup
• length of the assimilation window: 40 time steps
• perfect observations, noisy and sparse observations
• R = 0.01.
• B = I and B = 0.1e−
|i−j|
2L2 , where L = 5
• use Matlab quadprog.m
Melina Freitag Regularization in Variational Data Assimilation
![Page 56: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/56.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
4DVar - perfect and full observations, B = I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 80
Melina Freitag Regularization in Variational Data Assimilation
![Page 57: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/57.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
L1 - perfect and full observations, B = I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 80
Melina Freitag Regularization in Variational Data Assimilation
![Page 58: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/58.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
4DVar - noisy and sparse observations, B = I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 80
Melina Freitag Regularization in Variational Data Assimilation
![Page 59: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/59.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
L1 - noisy and sparse observations, B = I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 80
Melina Freitag Regularization in Variational Data Assimilation
![Page 60: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/60.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
4DVar - noisy and sparse observations, B = 0.1e−|i−j|
2L2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 80
Melina Freitag Regularization in Variational Data Assimilation
![Page 61: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/61.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
L1 - noisy and sparse observations, B = 0.1e−|i−j|
2L2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 40
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−1.5
−1
−0.5
0
0.5
1
1.5
TruthImperfect modelFinal solution
Figure: t = 80
Melina Freitag Regularization in Variational Data Assimilation
![Page 62: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/62.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Example 2 - Burgers’ equation
ut + u∂u
∂x= u + f(u)x = 0, f(u) =
1
2u2
with initial conditions
u(x, 0) =
(
2 0 ≤ x < 2.5
0.5 2.5 ≤ x ≤ 10.
Discretising
x(j) = 10(j − 1/2)∆x; U0(x(j)) =
(
2 0 ≤ x(j) < 2.5
0.5 2.5 ≤ x(j) ≤ 10.
where j = 1, . . . , N , ∆x = 1
Nand n is the number of time steps. We take
N = 100, ∆t = 0.001.
Melina Freitag Regularization in Variational Data Assimilation
![Page 63: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/63.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Exact solution and model error
Exact solution - method of characteristicsRiemann problem
u(x, t) =
(
2 0 ≤ x < 2.5 + st
0.5 2.5 + st ≤ x ≤ 10,
where s = 1.25
Numerical solution - model error
• the Lax-Friedrichs method (smearing out the shock)
Un+1
j =1
2(Un
j−1 + Unj+1) −
∆t
2∆x(f(Un
j+1) − f(Unj−1)).
• the Lax-Wendroff method (oscillations near the shock).
Un+1
j = Unj −
∆t
2∆x(f(Un
j+1) − f(Unj−1))+
∆t2
2∆x2
“
Aj+ 12(f(Un
j+1) − f(Unj )) − Aj− 1
2(f(Un
j ) − f(Unj−1))
Melina Freitag Regularization in Variational Data Assimilation
![Page 64: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/64.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Visualisation - Truth trajectory and numerical solution
Lax-Friedrichs method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Friedrich
Figure: t = 0Lax-Wendroff method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Wendroff
Figure: t = 0
Melina Freitag Regularization in Variational Data Assimilation
![Page 65: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/65.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Visualisation - Truth trajectory and numerical solution
Lax-Friedrichs method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Friedrich
Figure: t = 25Lax-Wendroff method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Wendroff
Figure: t = 25
Melina Freitag Regularization in Variational Data Assimilation
![Page 66: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/66.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Visualisation - Truth trajectory and numerical solution
Lax-Friedrichs method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Friedrich
Figure: t = 50Lax-Wendroff method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Wendroff
Figure: t = 50
Melina Freitag Regularization in Variational Data Assimilation
![Page 67: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/67.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Visualisation - Truth trajectory and numerical solution
Lax-Friedrichs method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Friedrich
Figure: t = 100Lax-Wendroff method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Wendroff
Figure: t = 100
Melina Freitag Regularization in Variational Data Assimilation
![Page 68: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/68.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Visualisation - Truth trajectory and numerical solution
Lax-Friedrichs method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Friedrich
Figure: t = 200Lax-Wendroff method
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
x
U(x)
TruthLax−Wendroff
Figure: t = 200
Melina Freitag Regularization in Variational Data Assimilation
![Page 69: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/69.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Setup
• length of the assimilation window: 100 time steps
• noisy and sparse observations
• R = 0.01.
• B = 0.1e−
|i−j|
2L2 , where L = 5
• use Matlab quadprog.m
Melina Freitag Regularization in Variational Data Assimilation
![Page 70: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/70.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Lax-Friedrichs method
Melina Freitag Regularization in Variational Data Assimilation
![Page 71: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/71.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
4DVar - noisy and sparse observations, B = 0.1e−|i−j|
2L2
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 0
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 50
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 100
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 200
Melina Freitag Regularization in Variational Data Assimilation
![Page 72: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/72.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
L1 - noisy and sparse observations, B = 0.1e−|i−j|
2L2
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 0
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 50
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 100
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 200
Melina Freitag Regularization in Variational Data Assimilation
![Page 73: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/73.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Lax-Wendroff method
Melina Freitag Regularization in Variational Data Assimilation
![Page 74: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/74.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
4DVar - noisy and sparse observations, B = 0.1e−|i−j|
2L2
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 0
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 50
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 100
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 200
Melina Freitag Regularization in Variational Data Assimilation
![Page 75: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/75.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
L1 - noisy and sparse observations, B = 0.1e−|i−j|
2L2
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 0
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 50
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 100
0 1 2 3 4 5 6 7 8 9 10−2
−1
0
1
2
3
4
TruthImperfect modelFinal solution
Figure: t = 200
Melina Freitag Regularization in Variational Data Assimilation
![Page 76: Regularization in Variational Data Assimilationmamamf/talks/vancouver.pdf · Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples Regularization in Variational](https://reader034.fdocuments.net/reader034/viewer/2022052022/6036cfb4b61e27127d2e1454/html5/thumbnails/76.jpg)
Outline Introduction 4DVar and Tikhonov L1-norm regularisation in 4DVar Examples
Conclusions, questions and further work
• L1-norm regularisation recovers discontinuity better than 4DVar
• Further work: analysis of methods; tests in 2D, 3D
• multiscale methods, other regularisation approaches
Melina Freitag Regularization in Variational Data Assimilation