MSDOL : a tool for sequential assimilation of satellite measurements of the stratosphere

Alain Hauchecorne, Jean-Loup Bertaux, Charlie Cot, Slimane Bekki

• Introduction
• Atmospheric model. I: dynamics
• Atmospheric model. II: chemistry & transport
• Assimilation algorithm
• Assimilation algorithm: application to MSDOL
• Assimilation algorithm - Specification of the forecast error variance
• Preparatory phase of the ENVISAT atmosphere chemistry mission validation
• Next steps

FIGURES

• Figure 1
• Figure 2
• Figure 3
• Figure 4
• Figure 5
• Figure 6
• Figure 7
• Figure 8
• Figure 9
• Figure 10
• Figure 11
• Figure 12

Introduction

Objectives:

·       Help satellite experiments validation by assimilation of the measurements into the MSDOL model (*);

·       Create climatological fields from assimilated fields;

·       Perform trend and correlation studies.

MSDOL is a Dynamic Chemistry Transport Model (D-CTM) forced by sequential assimilation of ozone observations and forced with wind data.

The end-product is the 3D structure of the atmosphere as a function of time.

MSDOL has been developed in the frame of the EU Environment & Climate programme

The MSDOL tool consists of:

·       an atmospheric model, coupling the dynamics, the chemistry and the transport in the stratosphere and the mesosphere;

·       a coupling with meteorological analyses (winds and temperature), driving the dynamics in the lower part of the model (up to 10 hPa);

·       a coupling with ENVISAT Òatmospheric chemistryÓ instruments (GOMOS, MIPAS and SCIAMACHY) data: sequential reading and assimilation of available measurements are performed during the run of the model;

·       a tool to simulate GOMOS/MIPAS/SCIAMACHY-like measurements.

Atmospheric model. I: dynamics

Based on a 3D coupled model of the dynamics & chemistry of the middle atmosphere (Rose & Brasseur, JGR 94, 16387, 1989).

·       Resolution:

o     24 pressure levels from 243 to 0.01 hPa;

o     11.25° in longitude, 5° in latitude.

·       Forcing :

o     Heating by O3 and O2, cooling by CO2;

o     Interaction mean flow-gravity waves;

o     Horizontal winds and temperature relaxed toward meteorological analyses (ECMWF, NCEP) in the lowermost region.

Atmospheric model. II: chemistry & transport

·       Chemical model:

o     33 chemical species and 5 chemical families

o     145 chemical or photolysis reactions, including heterogeneous chemistry on PSCs;

o     Tendencies computed through an iterative Euler-backward scheme.

·       Transport:

VanLeer eulerian scheme

o     Mass conservative

o     Positive definite

o     Low computational cost

Assimilation algorithm

·       MSDOL uses sequential assimilation: whenever an observation becomes available , it is used to update the predicted value by the model which is run simultaneously;

·       Optimal interpolation is used to combine observations and outputs of the model;

where

	Analysed state vector on the model grid (ozone mixing ratio)
	Predicted state vector
	Observations vector
	Weight of the innovation vector with respect to the prediction of the model
	Forward model providing a direct relation between the predicted state vector and the observations vector
	Covariance matrix of the forecast error
	Covariance matrix of the observation error
	Time evolution operator (model)

Assimilation algorithm: application to MSDOL

Assumptions:

o     The observation errors are spatially and temporally uncorrelated. This allows to handle each observation separately;

o     The horizontal and vertical forecast error covariances are independent and the vertical forecast error covariance is neglected;

o     Only the time evolution of the diagonal elements (the variances) of the forecast error covariance matrix are computed;

o     The covariances are computed by assuming that the horizontal correlation of the forecast error field may be represented by a simple function of the distance;

o     The operator H is a bilinear interpolation acting on the four points of the model surrounding the location of the observation.

Assimilation algorithm - Specification of the forecast error variance

The time evolution of the variances is represented by a law of the form:

with

	Variance of the forecast error at time k
	Operator representing the transport
	Time step of the model
	Ozone relaxation time
	Relative amplitude of the error model
	Climatological ozone mixing ratio

The first term represents the predictability error (as defined by Daley, MWR 120, 1735, 1992; that is the propagation of the error by the model), the second is the model error (that is the error made by the model over 1 time step).

·       The quality of the analyses depends on the correct estimation of the forecast error covariance matrix;

·       In the present case, the correlation is estimated from the difference between analysed (resulting from the assimilation of MLS data) and forecast ozone fields (Parrish & Derber, MWR 120, 1747, 1992).

Preparatory phase of the ENVISAT atmosphere chemistry mission validation

In order to assess the impact of the assimilation, an experiment using simulated observations has been set up. It is carried out in 4 steps:

STEP 1 : Modelled state of the atmosphere

Ø   Free-running simulation of the atmospheric model

STEP 2 : Simulation of the true state of the atmosphere

Ø   Obtained by adding a white noise to the state vector at each time step

STEP 3 : Generation of the simulated measurements

Ø   From the simulated true atmosphere and accounting for the instrumental characteristics (geometry, sampling and precision of the instrument and of the retrieval).

STEP 4 - Assimilation run

Ø   the independent assimilation of ENVISAT simulated measurements.

Preparatory phase of the ENVISAT atmosphere chemistry mission validation

An experiment using simulated observations (cont'd)

Team

ACRI-ST, (France)

Odile Hembise, Antoine Mangin, Bertrand Théodore

the Service d'Aéronomie (S.A.), (France)

Alain Hauchecorne, Jean-Loup Bertaux, Charlie Cot

the Bologna University, (Italy)

*Massimo Carlotti, Luca Magnani 

the Bremen University, (Germany)

*Klaus Bramstedt, Heinrich Bovensmann

the Finnish Meteorological Institute (FMI), (Finland)

*Erkki Kyrölä, Seppo Hassinen

Preparatory phase of the ENVISAT atmosphere chemistry mission validation

An experiment using simulated observations (cont'd)

Generation of the simulated measurements

Instruments observations characterisation:

·       Precision:

o     GOMOS (vertical distribution of ozone concentration)

o     Precision characterised with the stellar properties (Mv, T°)

o     MIPAS (vertical distribution of ozone volume mixing ratio)

o     Precision characterised with end-to-end simulations (instrumental + inversion)

o     SCIAMACHY (total ozone column)

o     Precision characterised with end-to-end simulations (instrumental + inversion)

·        Geometry:

Preparatory phase of the ENVISAT atmosphere chemistry mission validation

An experiment using simulated observations (cont'd)

Results

RMS difference between the simulated true state and:

Clearly apparent is the gain (by a factor of 2) brought by the assimilation above 100 hPa.

Ozone fields (model, assimilated, true)

GOMOS assimilation

after 3 days of assimilation (42.6 hPa)

	Model field
	Assimilated field
	True ozone field

Ozone fields (model, assimilated, true)
GOMOS assimilation
after 3 days of assimilation (27.6 hPa)

	Model field
	Assimilated field
	True ozone field

	Model field
	Assimilated field
	True ozone field

	Model field
	Assimilated field
	True ozone field

	Model field
	Assimilated field
	True ozone field

	Model field
	Assimilated field
	True ozone field

	Model field
	Assimilated field
	True ozone field

	Model field
	Assimilated field
	True ozone field

Next steps

* Assimilation and validation of ODIN data as a rehearsal of the ENVISAT validation (Spring 2001)

* Assimilation and validation of ENVISAT data (Winter 2001)