HAL Id: hal-02708028
https://hal.inria.fr/hal-02708028
Submitted on 1 Jun 2020
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Oceanic Dynamics under Location Uncertainty: Towards a consistent stochastic modeling
Long Li, Werner Bauer, Etienne Mémin
To cite this version:
Long Li, Werner Bauer, Etienne Mémin. Oceanic Dynamics under Location Uncertainty: To-
wards a consistent stochastic modeling. Workshop Conservation Principles, Data and Uncertainty
in Atmosphere-Ocean Modelling, Apr 2019, Potsdam, Germany. �hal-02708028�
Oceanic Dynamics under Location Uncertainty
Towards a consistent stochastic modelling
Fluminance Group
Inria Rennes France & Universit´e Rennes 1 France
Long Li Werner Bauer Etienne Mémin
Objective
Why stochastic modeling ?
Why consistent model ?
Take into account small-scale / unresolved processes
Uncertainty Quantification, Ensemble Forecasts
Both scales must respect appreciate physical laws
Losing consistency may provide wrong statistics
Outline
3 Summary
4
2 Experimental Evaluations
Structure—preserving of Rossby wave
Ensemble forecasting verification of SQG MU
Time—statistics of wind—driven circulation
1 Governing Equations
Location Uncertainty Principles
Barotropic Quasi—Geostrophic Model under LU
homogeneous / heterogeneous
Location Uncertainty Principles
small-scale unresolved large-scale
resolved
dX t = u(X t , t)dt + (X t , t)dB t
Stochastic flow :
a = 4 T = E ⇥
dB t ( dB t ) T ⇤
/dt
Variance tensor :
space-time white noise symmetric
kernel correlated in space
uncorrelated in time
Functional process :
(x, t)dB t = Z
D
˘ (x, y, t)dB t (y)dy
multiplicative noise
subgrid diffusion corrected drift
u 1
2 r · a
Balanced Energy
d t ✓ + u ? · r ✓dt + dB t · r ✓ 1
2 r · (a r ✓)dt = 0 D t ✓ = 4
Transport of a random tracer : ( r · = 0 )
d dt
Z
D
1
2 ✓ 2 = 0
( r · u ? = 0 )
( Mémin 2014; Resseguier et al. 2017 )
Barotropic Quasi–Geostrophic Model under LU ( Resseguier et al. 2017; Li et al. 2019 )
Evolution of the potential vorticity with source processes : D t q = S 1 ( r u)dt + S 2 ( r u)dB t
Evolution of the stream function :
q = /L 2 R + f
Strong incompressible constraints :
u = r ? , dB t = r ? 'dB t , r · u ? = 0 Conservation of total energy : d
dt Z
D
1 2
h kr k 2 + ( /L 2 R ) 2 i
= 0
Transport of a prognostic passive tracer : D t q 0 = 0
Conservation of tracer energy : d
dt Z
D
1
2 (q 0 ) 2 = 0
h Sketch : Z
D
dB t · r (q f ) = Z
D
S 2 ( r u) dB t i
Outline
3 Summary
4
2 Experimental Evaluations
Structure—preserving of Rossby wave
Ensemble forecasting verification of SQG MU
Time—statistics of wind—driven circulation
1 Governing Equations
Location Uncertainty Principles
Barotropic Quasi—Geostrophic Model under LU
Structure–preserving of Rossby wave
Flowchart : q
( Potential Vorticity )
q′
( Passive Tracer )
Illustration of conservations :
Passive Tracer Tracer Energy Total Energy
ψ
( Stream Function )
Ensemble forecasting verification of SQG MU
Time
Initial uncertainty Forecast uncertainty Model
PIC
LU
Illustration of ensemble prediction systems :
Ensemble forecasting verification of SQG MU
Reliability — Ensemble Spread match MSE of the Ensemble Mean :
( Resseguier et al. 2019 )
Ensemble forecasting verification of SQG MU
LU
homogeneous
PIC
homogeneous
LU
heterogeneous
PIC
heterogeneous
Day 10 Day 20 Day 30 Reliability
Model
Reliability — Ranked Histogram of Ensemble Members :
( Resseguier et al. 2019 )
Time–statistics of wind–driven circulation
DNS 256 ⇥ 512 with Re = 450 and Ro = 0.0016
Time–statistics of wind–driven circulation
DNS 256 ⇥ 512 LU 32 ⇥ 64 LES LU 32 ⇥ 64
Truth
Outline
3 Summary
4
2 Experimental Evaluations
Structure—preserving of Rossby wave
Ensemble forecasting verification of SQG MU
Time—statistics of wind—driven circulation
1 Governing Equations
Location Uncertainty Principles
Barotropic Quasi—Geostrophic Model under LU
Summary
A consistent stochastic model :
Physical conservation laws satisfied Better ensemble spread represented
Time-averaged profil well described on coarse mesh
Future works :
Multi-layers QG model under LU
( Q-GCM Projects : Hogg et al. 2003 )
Data Assimilation with Particle Filters
( Cotter et al. 2018 )
Thanks for Your Attention !
Ensemble of random stream functions
Homogeneous parameterization
low-pass band filter ? white noise
Heterogeneous parameterization
y( m )
x(m)
t = 0 d ay s
0 2 4 6 8
x 10 5 0
5
10 x 10 5
1 0.5 0
0.5 x 10 1 3
Neighbor Centered neighbor
Random selection
y( m )
x(m) t = 0 d ay s
0 2 4 6 8
x 10
50
5 10 x 10
51 0.5 0 0.5 x 10 1
3y( m )
x(m) t = 0 d ay s
0 2 4 6 8
x 10
50
5 10 x 10
51 0.5 0 0.5 x 10 1
3y( m )
x(m) t = 0 d ay s
0 2 4 6 8
x 10
50
5 10 x 10
51 0.5 0 0.5 x 10 1
3y( m )
x(m) t = 0 d ay s
0 2 4 6 8
x 10
50
5 10 x 10
51 0.5 0 0.5 x 10 1
3y( m )
x(m) t = 0 d ay s
0 2 4 6 8
5