S P A R C Stratospheric Processes And their Role in Climate A project of the World Climate Research Programme Home Initiatives Organisation Publications Meetings Acronyms and Abbreviations Useful Links

Report on ECMWF-SPARC Workshop on Modelling and Assimilation for Stratosphere and Tropopause

ECMWF, Reading, UK, 23-26 June 2003

Emmanouil K. Oikonomou, SPARC Office, Verrières le Buisson (eoikonomou@aerov.jussieu.fr)

The workshop jointly organized by ECMWF and SPARC consisted of talks on the current state of research about the stratosphere and the tropopause. There were 22 oral presentations. Three working groups were formed (Processes, Data Assimilation (DA), and Modelling) who then reported during a plenary session.

1. Presentations

The workshop opened with a summary by A. Simmons on the representation of the stratosphere in the ECMWF operations and their most recent Re-Analysis Scheme ERA-40. Overall the performance of the system is considered successful (e.g. sudden warming predictions, QBO); however, there are problems such as model and observation biases, handling of tides and performance difference between 3D and 4D-Var systems. A. O'Neill talked about the objectives of SPARC and DA requirements: long-term global data sets; 3-D velocity fields with reduced noise; parameterized mass fluxes; diabatic heating rates; ozone assimilation, aerosols and other tracers.

The first session focused on Radiative Transfer (RT) in the stratosphere, cross-Tropopause (TP) processes, Cirrus clouds and chemical forecast using CTMs. The current issues on representing stratospheric processes into modelling and DA include the momentum budget, Gravity Wave Drag (GWD), improvements in tropical winds, meridional circulation and mixing barriers, and the unbalanced flow component in the upper stratosphere and mesosphere (T. Shepherd). Over the tropics, recent inertial adjustments appear to be a real process suggesting that a lot of the upwelling may be resolved. However, outstanding problems remain: tropical winds; biases of models in polar temperatures; underestimation of GWD when parameterizing it in the presence of a zonal-mean sponge.

There have been recent developments with impact on the ECMWF RT stratospheric scheme, followed by comparisons of ozonesonde profiles with the ERA-40 fields and 10-day model profiles (J-J. Morcrette). There also exists future possibility for ECMWF to provide operational UV-B diagnostics once radiation becomes interactive with prognostic ozone. H. Wernli showed results on the cross-TP processes and quantification by using a Lagrangian approach and based upon ECMWF analyses combined with aircraft data from SPURT. TP folds are very important in the sub-tropics but less important in the extra-tropics. Significant differences were found between ERA-40 and ERA-15 near the TP and there was qualitative agreement between regions of maximal STE and storm tracks. T. Peter reported on results from the LITE project and emphasized the role of ultra thin tropical tropopause clouds, concluding that ECMWF analyses suggest maritime continent to be a major source for stratospheric air.

The way stratospheric chemistry and aerosols are understood and represented in DA CTMs remains a challenge. D. Fonteyn presented 4D-Var DA runs using non-operational MIPAS data with a constrained N₂O budget, which suggest that advection and tendencies are anti-correlated and the assimilation should constantly constrain the N₂O budget. Regarding which species to include into CTMs, the session concluded that complex CTMs may not work in short-time and the overall idea is to keep the schemes simple so that they can be controlled.

There were three talks that covered satellite observations in the UT/LS region including aerosols whose loading in 2002 was at the lowest. L.W. Thomason summarized the current state in gas/aerosol observational and retrieval techniques; these include Occultation coverage, Limb Emission (CLAES/HRDLS) and Lidar observations and modelling. B.J. Kerridge reported on assimilated GOME Ozone data, which agree reasonably well when compared to SAGE-III, HALOE and MIPAS profiles using ECMWF winds. A. Dudhia presented an overview on the Infrared Limb Sounding instruments of MIPAS and HIRDLS; a Ray Tracing technique has been implemented to retrieve pressure and temperature with more of a problem being the transmittance calculations.

The session on DA comprised five talks, mainly on the ECMWF and UKMO DA systems. Approaches in chemical DA range from simple chemistry dynamics in GCMs to sophisticated photochemistry CTMs and coupled GCM/CTMs. The assimilation of satellite retrievals versus radiances in operational Numerical Weather Prediction (NWP) has proven to be less attractive, mainly because satellite retrievals retain characteristics of the a priori information, they constrain information from other observations, they have complicated error structures and the distribution retrievals may be significantly delayed compared to raw radiances. T. McNally presented an overview of the observing systems used at ECMWF, with HIRS and AMSUA proving to be the most reliable platforms for the stratosphere. A. Dethof gave a more in-depth analysis of the ECMWF DA scheme for ozone. R. Swinbank talked about assimilating stratospheric ozone and water vapour in the UKMO model; future plans include the development of an Extended Global Assimilation System (DEGAS). A common conclusion is that the key limitations of all these DA satellite systems are their systematic errors and their vertical resolution. Furthermore, model errors are an important issue in chemical DA that can be considered as an unobserved variable requiring the knowledge of cross-error covariance (R. Ménard). Future challenges in DA focus on: understanding of systematic errors; tuning the error covariance specifically for the type of errors encountered in the stratosphere; making use of improved operational instruments along with synergistic use; coupling dynamics with chemistry in DA systems (e.g. GCM/CTM); and assimilating limb radiances, as for example currently developed by DARC from MIPAS data (W. Lahoz).

The final session was devoted to modelling. V. Peuch talked about developing UV index forecast in Météo-France. The results from CTM MOCAGE have been compared to MOZAIC data. The CTM approach (off-line/semi-online) can provide a flexible solution, e.g. for Chemical Weather Forecast (CWF), DA and climate chemistry. Another DA system is NASA’s GEOS-4 with Limb-sounding temperatures from SABER effectively combined with TOVS and the inclusion of MIPAS ozone and SBUV (S. Pawson). The problems are mainly superfluous subtropical mixing, excessive cross-barrier transport and the fact that local assimilation seems to lead to noise.

Further results were presented from using meteorological analyses in the off-line CTM SLIMCAT and coupled chemistry-climate GCM (UKMO Unified Model) applied to study ozone depletion and past trends in mid-latitudes (M. Chipperfield ). A study of NWP models showed that currently stratospheric forecasting at 6 days is comparable to 3-day forecast in the troposphere, with poorer results obtained when the polar vortex flow is rapidly changing (G. Roff ). ECMWF have further developed a finite-element discretization technique for the vertical and they plan to increase the model vertical resolution; these changes are expected to improve the vertical stratospheric transport and reduce large model errors near the stratopause (A. Untch). Finally, reports on work in-progress included the NWP ICON project of the German Weather Service (DWD) and the MPI for Meteorology (L. Bonaventura ) and issues in isentropic-coordinate and TP modelling (J. Thuburn).

2. Recommendations of the working groups

2.1 Processes (H. Wernli)
Six major issues were pinpointed: Ozone; Vertical resolution and upward extension; Global circulation with particular emphasis to the tap-recorder effect; Cloud parameterization; Water vapour in the stratosphere; Data availability.
Prognostics of the ozone profile near the tropical tropopause remain problematic. Therefore, there is a need to validate against ozenesondes the ECMWF ozone fields and compare to the Fortuin and Langematz climatology. It was also suggested to revise the Cariolle-Dequé parameterization scheme implemented in the ECMWF ozone prognostics. The threshold of 195 K used for heterogeneous chemistry activation should be made altitude-dependent and a chlorine memory effect should also be included into the scheme. The results of the revised scheme can be then compared with the output from CTMs and MIPAS-like observations.

The group concluded that high vertical resolution has a large influence on the accuracy of CTMs. It was proposed that when the ECMWF model top exceeds 60 km the effect of non Equilibrium Ratiative processes might have to be included. The ECMWF RT model, which assumes Local Thermodynamic Equilibrium (LTE), can be validated through sensitivity studies and against models including non-LTE schemes. In addition, at about 85 km chemical heating might have to be taken into account. The tape recorder signal was found to be too fast in the ECMWF model coming from a too large mean vertical velocity at the equator, either due to variance in the vertical velocity, or from excessive numerical diffusion.
There is concern about the quality of the vertical transport in the ECMWF datasets. The production of accurate winds in the tropopause region represents a future aim, as well as to assess whether tracer simulation, such as N₂O, will improve the quality of the analyzed stratospheric winds. Also, any future parameterization upgrades should take into account the supersaturation of cirrus clouds. With respect to the water vapour in the stratosphere, MIPAS is moist compared to ECMWF but dry compared to HALOE. A systematic validation of the ECMWF moisture is needed against all available data sources. Finally, the group recognized the usefulness of archiving the diabatic 3D fields from ERA-40.

2.2 Data Assimilation (A. O’Neill)
There was an overall positive comment on the ECMWF system, with significant steps made towards improving the representation of the stratosphere, the existence of good analysis tools, an improved forecast performance, exploitation of novel datasets and interaction with data producers.

The issues for further attention include: calibrate and retune the stratospheric background error covariances; the Jb formulation; encourage the use of non-assimilated (i.e. independent) observations; deal with systematic errors within the analysis; capture an accurate representation of the Brewer-Dobson circulation and mixing barriers, with the suggestion for future inclusion of longer-term tracer species; impose additional constraints upon the DA system, such as balance and conservation. Recognized as particular weakness in the ECMWF system were the tides and the omission in the specification of explicit correlations among tracers and between the tracers and the dynamics in the error covariance.

The group emphasized the need for a robust system for meteorology and ozone DA, with AMSUA being identified as a particular example of a stable, well-calibrated instrument with excellent time continuity and important for the stratosphere. Other instruments also provide good quality measurements, such as ozone column from GOME/SBUV/TOMS and ozone profiles from ENVISAT/AURA. ECMWF should continue to evaluate the need for limb-viewing data and aircraft observations (e.g. MOZAIC). Radiosondes and remotely sensed winds should provide more comprehensive time and location information and humidity sensors need to be improved, along with the Near Real Time (NRT) aspects of ozone sondes and ground based data. In addition, aerosol measurements are required to support the RT modelling, parameterization and chemistry/aerosol forecast. As a particular problem was seen the lack of organization and communication between the wide variety of research missions.

The group discussed the region and flow dependence of B matrix and it stressed the importance of choosing control variables and representations that facilitate the treatment of crucial regions, such as the UT/LS and shear zones, and the wider issue of STE. ECMWF should consider strategies for incorporating constituent modelling/assimilation for both NWP and environmental monitoring. RT observational operations should keep pace with the variety of measurements approaches (UV) and the new instrument developments.

OSEs/OSSEs (Observing System Experiment/Observing System Simulation Experiments) were suggested to assess remotely-sensed winds and humidity data in the UT/LS. The support for the ERA-40 project should continue and the 1990’s should be used for validation of the Re-analysis. The group further suggested to explore the frequency and resolution of analyses and post-processed products (e.g. theta, fluxes) and their external availability to CTM users. The location of the upper boundary and its impacts to radiance assimilation and systematic errors were also discussed, along with the need to strenghten the assosciation between SPARC modelling and DA activity.

2.3 Modelling (J. Thuburn)
The group identified modelling issues related to dynamics and chemistry (ozone/CH₄/water vapour). The modelling issues regarding dynamics include: Mean meridional circulation and vertical transport; lateral dispersion; Model top, sponge layer and GWD; Inertial instability; Vertical resolution, advection and coordinate; and Mass conservation.

The ECMWF model vertical transport appears to be excessively fast in the stratosphere due to either (a) excessive mean meridional circulation, in which case we need to examine whether the problem might be inherent in the free-running model or exacerbated by the DA, or (b) due to vertical advection errors, in which case a higher order advection scheme, or possibly an alternative model vertical coordinate, should be considered. NASA’s experience also suggests excessive lateral dispersion when using analyzed winds for transport, but not when using winds from a free-running GCM.

In the issue of a model sponge layer, the group suggested: a sponge layer to be turned on from 0.1 hPa and a model lid placed at approximately 0.01 hPa; removal of the zonal-mean component of the sponge-layer; introduction of a momentum-conserving GWD scheme. It was also suggested to diagnose the presence of inertial adjustment in the ECMWF model and to explore the possibility of parameterizing this effect.

Vertical resolution is another important issue and several criteria were discussed for what might constitue it sufficient. Re-examination in the current DA schemes is also required if limb data are to be introduced. Regarding the vertical coordinate, the options of a hybrid isentropic and of a quasi-Lagragian were suggested to be considered, if the problems with vertical transport cannot be resolved by simpler measures. Nevertheless, there is experience suggesting that increments in GCM DA schemes can cause problems in the horizontal circulation. ECMWF is currently going ahead with their plan to move into a 91-level model expanding to 0.01 hPa.

With respect to ozone chemistry the group recommended: (i) to investigate the Ozone prognostics in the current Cariolle scheme since there are clear discrepencies between the forecast ozone fields and the observations; in particular there is an under-estimated Antarctic ozone hole, a negative bias in the tropics and a positive bias in mid-latitudes; (ii) explore better ozone parameterizations since the current scheme is based on a 2D-model output, whereas 3D-CTMs seem to perform better; and (iii) coupling ozone to radiation scheme in the future. There was an overall agreement for inclusion of a long-lived tracer for dynamic diagnostics, such as CH₄, and the recommendation that possible extension of the model into the mesosphere may require inclusion of the H₂O chemical sink at high altitudes.