pres esa 05

Assimilation of EPS tropospheric ozone for air quality forecastB. Sportisse, M. Bocquet, V. Mallet, I. Herlin, JP. Berroir, H. Boisgontier: 

Assimilation of EPS tropospheric ozone for air quality forecast B. Sportisse, M. Bocquet, V. Mallet, I. Herlin, JP. Berroir, H. Boisgontier ESA EUMETSAT EPS/MetOp RAO Workshop, May 2006

Project objectives: 

Project objectives Applicative context: air quality forecast at European Scale. Air quality -> boundary layer concentrations Objective: to assess the feasibility of assimilating O3 columns and profiles, to be provided by MetOp (IASI, GOME): numerical experiments to be performed before the launch of MetOp to perform experiments with real O3 data when available validation and quantification by comparison with forecast obtained without assimilation, or with assimilation of ground stations measurements

Outline: 

Outline The POLYPHEMUS air quality modelling system Preliminary studies: Relative weight of boundary layer ozone in the IASI 0-6km O3 column Sensitivity of boundary layer ozone to IASI O3 column Data assimilation feasibility Based on twin numerical experiments Based on simulated IASI columns Perspectives: assimilation of real data

The Polyphemus air quality modelling system: 

The Polyphemus air quality modelling system Developed by CEREA (joint ENPC-EdF laboratory, Clime team shared between ENPC and INRIA) Modular architecture: Access to raw input data Input data processing: physical parameterizations, … Numerical heart: CTM, its linear tangent and adjoint models High level applications: Direct forecast Sensitivity studies Impact studies Ensemble forecast Data assimilation … Visualisation

POLYPHEMUS architecture: 

POLYPHEMUS architecture

POLYPHEMUS applications: ozone forecast: 

POLYPHEMUS applications: ozone forecast European scale run ECMWF met. Fields EMEP emissions Extensively validated http://www.enpc.fr/cerea/polyphemus

POLYPHEMUS applications: ensemble forecast: 

POLYPHEMUS applications: ensemble forecast O3 vs time for different physical parameterization Need of better constraining models Data assimilation

Preliminary studies: 

Preliminary studies Is assimilation of EPS O3 information within regional CTMs feasible? Problem: EPS O3 mainly representative of upper troposphere Air quality application interested in boundary layer ozone Feasible if: Boundary layer has a significant weight in the 0-6km column Boundary layer ozone, as forecast by Polyphemus, is sensitive to changes in upper tropospheric ozone

Relative weight of boundary layer O3: 

Relative weight of boundary layer O3 Computed from a reference situation (validated Polyphemus analysis – July 2001) Average contribution of boundary layer 03 to 0-6km column: 14% (more during day because boundary layer is higher) Irregularly scattered in space and time Mean at 0h Mean at 15h

Sensitivity of boundary layer O3: 

Sensitivity of boundary layer O3 Experiment: Consider a reference run Apply perturbation to upper tropospheric ozone: at initial date, every 24h Run forecast with perturbated O3 Compare forecast with reference Conclusion, from initial perturbation. Max. sensitivity: 50% Max difference with reference observed 27h after simulation Conclusion, from cyclic pertubation Perturbated runs quicly diverges from reference Overall conclusion: Yes, upper tropospheric ozone can be used to constrain boundary layer O3

Sensitivity of boundary layer ozone: 

Sensitivity of boundary layer ozone -50% perturbation applied on UT O3 Graph: mean relative difference at ground level between perturbation and reference

Sensitivity of boundary layer O3: 

Sensitivity of boundary layer O3 Same experiment Relative difference 27h after perturbation

Sensitivity of boundary layer O3: 

Sensitivity of boundary layer O3 Perturbation applied every 24h

Feasibility of data assimilation: 

Feasibility of data assimilation Numerical experiments: Columns generated from reference situation, 5% perturbation Assimilation within perturbated runs: Method: Optimal Interpolation Different choices of model/observation covariance matrices Assimilation every 24h 2 types of perturbated runs: Perturbation of initial condition: assimilation should speed up the convergence to reference Pertubation of physical parameterization (Kz): perturbation always different from reference, assimilated run should lie in between.

Feasibility of assimilation: perturbation of initial condition: 

Feasibility of assimilation: perturbation of initial condition Relative difference with reference Blue: without DA Others: DA with different values of obs. error. Quicker convergence to reference

Feasibility of assimilation: simulated IASI column: 

Feasibility of assimilation: simulated IASI column Context: cooperation with SA (C. Clerbaux) Simulation of IASI radiances from: Polyphemus reference (0-5km) UGAMP climatologies (6-60km) Interpolation (5-6km) LBLRTM code + IASI instrument model Inversion of IASI radiances: Atmosphit model (for evaluation, 2 dates) SA NN operational code

Simulation of IASI measurements: 

Simulation of IASI measurements Example of profiles/inversion error (Atmosphit)

Simulation of IASI measurements: issues: 

Simulation of IASI measurements: issues Preparing the assimilation of real data: Realistic noise (instrument, inversion procedure) Set up of observation operator Set up of observation covariance matrices Influence of a priori information: 6-60km (O3, T, p) LST Ongoing, awaiting results

Conclusions: 

Conclusions Twin numerical experiments tend to show that assimilation of 0-6km is feasible and can improve the analysis of boundary layer O3, in the case of high quality measurements Realistic accounting for instrument and inversion noise: experiment with simulated IASI columns going on Waiting for real data…