Title: Quality of model and Error Analysis in Variational Data Assimilation
1Quality of model andError Analysis in
Variational Data Assimilation
- François-Xavier LE DIMET
- Victor SHUTYAEV
- Université Joseph FourierINRIA
- Projet IDOPT, Grenoble, France
- Russian Academy of Sciences
- Institute of Numerical Mathematiques
2Prediction What information is necessary ?
- Model
- law of conservation mass, energy
- Laws of behaviour
- Parametrization of physical processes
- Observations in situ and/or remote
- Statistics
- Images
3Forecast..
- Produced by the integration of the model from an
initial condition - Problem how to link together heterogeneous
sources of information - Heterogeneity in
- Nature
- Quality
- Density
4Basic Problem
- U and V control variables, V being and error on
the model - J cost function
- U and V minimizes J
5Optimality System
- P is the adjoint variable.
- Gradients are couputed by solving the adjoint
model then an optimization method is performed.
6Errors
- On the model
- Physical approximation (e.g. parametrization of
subgrid processes) - Numerical discretization
- Numerical algorithms ( stopping criterions for
iterative methods - On the observations
- Physical measurement
- Sampling
- Some pseudo-observations , from remote
sensing, are obtained by solving an inverse
problem.
7Sensitivity of the initial condition with respect
to errors on the models and on the observations.
- The prediction is highly dependant on the initial
condition. - Models have errors
- Observations have errors.
- What is the sensitivity of the initial condition
to these errors ?
8Optimality System including errors on the model
and on the observation
9Second order adjoint
10(No Transcript)
11(No Transcript)
12Models and Data
- Is it necessary to improve a model if data are
not changed ? - For a given model what is the best set of
data? - What is the adequation between models and data?
13A simple numerical experiment.
- Burgers equation with homegeneous B.C.s
- Exact solution is known
- Observations are without error
- Numerical solution with different discretization
- The assimilation is performed between T0 and T1
- Then the flow is predicted at t2.
14(No Transcript)
15(No Transcript)
16Partial Conclusion
- The error in the model is introduced through the
discretization - The observations remain the same whatever be the
discretization - It shows that the forecast can be downgraded if
the model is upgraded. - Only the quality of the O.S. makes sense.
17Remark 1
- How to improve the link between data and models?
- C is the operator mapping the space of the state
variable into the space of observations - We considered the liear case.
18(No Transcript)
19(No Transcript)
20Remark 2 ensemble prediction
- To estimate the impact of uncertainies on the
prediction several prediction are performed with
perturbed initial conditions - But the initial condition is an artefact there
is no natural error on it . The error comes from
the data throughthe data assimilation process - If the error on the data are gaussian what
about the initial condition?
21Because D.A. is a non linear process then the
initial condition is no longer gaussian
22Control of the error
23(No Transcript)
24Choice of the base
25Remark .
- The model has several sources of errors
- Discretization errors may depends on the second
derivative we can identify this error in a base
of the first eigenvalues of the Laplacian - The systematic error may depends be estimated
using the eigenvalues of the correlation matrix
26Numerical experiment
- With Burgers equation
- Laplacian and covariance matrix have considered
separately then jointly - The number of vectors considered in the correctin
term varies
27With the eigenvectors of the Laplacian
28(No Transcript)
29(No Transcript)
30(No Transcript)
31(No Transcript)
32(No Transcript)
33(No Transcript)
34(No Transcript)
35(No Transcript)
36(No Transcript)
37- Model error estimation controlled system
- model
- adjoint system(to calculate the gradient)
38Reduction of the size of the controlled problem
Suppose is a base of the
phase space and is time-dependent
base function on 0, T, so that
then the controlled variables are changed to
with controlled space size
39Optimality conditions for the estimation of
model errors after size reduction
If P is the solution of adjoint system, we
search for optimal values of
to minimize J
40- Problem how to choose the spatial base
? - Consider the fastest error propagation direction
- Amplification factor
- Choose as leading eigenvectors of
- Calculus of
- - Lanczos Algorithm
41Numerical experiments with another base
- Choice of correct model
- - fine discretization domain with 41 times
41 grid points - To get the simulated observation
- - simulation results of correct model
- Choice of incorrect model
- - coarse discretization domain with 21
times 21 grid points
42The difference of potential field between two
models after 8 hours integration
43Experiments without size reduction (108348)
the discrepancy of models at the end of
integration
before optimization
after optimization
44 Experiments with size reduction (38048)
the discrepancy of models at the end of
integration
before optimization
after optimization
45 Experiments with size reduction (3808)
the discrepancy of models at the end of
integration
before optimization
after optimization
46Conclusion
- For Data assimilation, Controlling the model
error is a significant improvement . - In term of software development its cheap.
- In term of computational cost it could be
expensive. - It is a powerful tool for the analysis and
identification of errors