Report on the Joint SPARC Workshop on Data Assimilation and
International Polar Year (IPY)

4-7 September 2007, Toronto, Canada

S. Polavarapu, Environment Canada, saroja.polavarapu@ec.gc.ca

E. Farahani, University of Toronto, Canada, elham@atmosp.physics.utoronto.ca

G. Manney, Jet Propulsion Laboratories, manney@mls.jpl.nasa.gov

N. McFarlane, SPARC IPO, Canada, Norm.McFarlane@ec.gc.ca

T. G. Shepherd, University of Toronto, Canada, tgs@atmosp.physics.utoronto.ca

Introduction

The SPARC Data Assimilation Working Group (DAWG) was initiated in 2002 to foster exchange between the data assimilation and stratospheric research communities.  Data assimilation impinges on many aspects of SPARC goals such as the documentation of the stratosphere and the need to reduce uncertainties in climate models.  By confronting models with measurements, model deficiencies can be highlighted.  In addition, there is a hope of being able to use the assimilation process to identify free parameters in parameterized processes such as gravity wave drag or deep convection.  One of the main vehicles for this exchange is the annual data assimilation (DA) workshop.  These workshops have typically alternated locations between Europe and North America.  The 2006 workshop was held in Noordwijk, the Netherlands and a summary of this meeting is found in SPARC Newsletter No. 28.  The 2007 workshop was held in Toronto, Canada during 4-7 September.

The themes of the 2007 workshop (which were identified during the 2006 workshop) were: stratosphere-troposphere coupling, the mesosphere (including stratosphere/mesosphere coupling), and International Polar Year (IPY).  Data assimilation for large atmospheric models is primarily performed at operational weather centres.   At operational centres, the value of forecast improvements is often heavily weighted toward tropospheric impacts.  Thus, the data assimilation community is very interested in better understanding stratosphere-troposphere coupling — more specifically, in how improving stratospheric analyses can impact tropospheric forecasts.  The second theme was motivated by the fact that operational centres such as ECMWF, GMAO and the UK Met Office are moving their model lids into the mesosphere (at least 80 km) in order to properly analyse the entire stratosphere.  Thus the assimilation community expressed a desire to better understand the dynamics of the mesosphere in order to improve analyses in that region.  Initial experience with extended model domains that include the mesosphere is that large analysis increments can be generated in the mesosphere which can lead to model instabilities. Finally, IPY was a natural theme since the SPARC DA Working Group is putting a substantial effort into the collection of global analyses during March 2007 to March 2009 (the IPY period) for the SPARC-IPY project.

Stratosphere-Troposphere coupling

In his overview presentation, A. Scaife used climate models to show that the stratosphere plays an important role in the North Atlantic Oscillation (NAO) signal. He showed that the observed increase in the NAO from the 1960s to 1990s was strongly influenced by changes in the stratosphere. Because ENSO (El Nino/Southern Oscillation) events frequently weaken the stratospheric polar vortex they can also produce a negative NAO response. The presence of a model stratosphere can therefore have an influence on seasonal forecasts.  Figure 1 shows that while climate models underestimate blocking frequency, blocking appears to be sensitive to the representation of the model stratosphere. The troposphere-stratosphere HadAM3 model reproduces the observed maxima in blocking in both the Pacific and Atlantic sectors.  M. Baldwin noted that stratosphere-troposphere coupling is evident at high latitudes and easily diagnosed using the NAM (Northern Annular Mode) index.  A negative NAM index corresponds to a strong polar vortex while a positive index corresponds to a weak polar vortex, and the surface signature corresponds to the Arctic Oscillation.  There is a time delay for the stratospheric signal to propagate to the surface, but the impact can last for months (Baldwin and Dunkerton 2001). The strength of the stratospheric vortex also affects the

position of storm tracks, with strong polar vortices being associated with more northern storm tracks.  Interestingly, C. Li showed how wintertime stratospheric NAM anomalies were correlated with the summertime Mei-Yu precipitation anomaly in east Asia. Since the NAM index is a good indicator of stratosphere-troposphere coupling, it would be interesting to be able to apply it as a standard diagnostic of analyses or forecasts.  However, it is computationally intensive to compute as it requires daily 3D fields. To that end, M. Baldwin proposed a 1D version of the NAM index and showed that it captured the essence of the 2D NAM index.

The impact of an improved stratosphere on the troposphere was also studied by D. Jackson and C. Mathison (in a presentation given by M. Keil).  Specifically, use of EOS-MLS data improved ozone analyses which then impacted tropospheric forecasts.  In addition, R. Errico showed that stratospheric singular vectors can be used to identify rapidly growing perturbations.  (Singular vectors identify the most rapidly growing perturbations to a given flow for a specified time period and vector norm.)  For a 5 day optimization period and a background flow that develops a stratospheric warming, stratopheric singular vectors were found to be baroclinic, like tropospheric ones, but unlike their tropospheric counterparts, nonlocal structures were found.  Thus initial perturbations may be located at some distance horizontally from where their ultimate impact is felt.

Figure 1: Atmospheric blocking frequency in the Hadley Centre HadAM3 model calculated according to Tibaldi and Molteni (1990) using 500hPa geopotential height. Modelled climatological blocking frequency in the standard model (upper panel - gray) and troposphere-stratosphere model (upper panel - black).  Blocking climatologies are calculated over the 40 year period from 1960 onwards and 3 realisations are shown from each model. Observed blocking frequencies from ERA 40 reanalyses are shown as blue dots. The lower panel is as the upper panel but with the mean background climatologies of the  models exchanged.  This shows that the model blocking frequency is much higher when the stratosphere is properly represented and that this is due largely to a change in the mean state.  Of course these results are for one particular model and may be model dependent.  (Courtesy of Adam Scaife, Met Office) For better resolution, please contact the SPARC IPO

Top

The role of assimilation in the UTLS region

Because there simply aren’t enough observations to define the full state, data assimilation products are a blend of measurements and model forecasts.  Therefore, it is interesting to understand where the analysis is begin determined by the model and where the analysis is being determined by the assimilation.  K. Wargan was able to show that ozone assimilation primarily moves the modes of the ozone PDFs by correcting model biases, while the basic structure of the ozone fields is determined by the model.  Figure 2 shows a comparison of model and assimilations to aircraft measurements in the UTLS region.  The measurements show a multi-modal structure since some of the air being sampled is tropospheric (the lower ozone mode) while some is stratospheric (the higher ozone mode).   The free-running model correctly represents the existence of the two modes and the overall structure of the PDF, while the assimilation is able to correct the position of the stratospheric mode.  (The tropospheric mode is also slightly improved.)  Figure 3 shows that the shape of the horizontal spectrum of ozone is similar for both model and assimilation, with assimilation producing an offset which primarily acts to correct the ozone bias of the model. Again this suggests that the horizontal structure of the ozone field in the assimilation is mainly being driven by transport, which is captured by the model. This point was also made by G. Manney in comparisons of ozone morphology in the UTLS from the Aura Microwave Limb Sounder (MLS) and a number of transport models (both online and offline) and assimilation systems (with and without ozone assimilation).  Further evidence that the horizontal structure of tracer gradients is well defined by model forecasts was found by M. Hegglin, who showed that CMAM-DAS was able to maintain latitudinal gradients of N2O, NOy and O3, as well as transport barriers (when compared to ACE and SPURT aircraft measurements), despite using a 3D-Var scheme. (Note that Scheele et al. (2005) found that 4D-Var assimilated winds are better able to preserve latitudinal gradients of age-of-air than 3D-Var assimilated winds.)  This result points to the benefit of online advection where constituents are advected by wind fields which are adjusted every time step (rather than every few hours, as occurs in offline advection with a CTM).  This result may seem somewhat surprising since online advection has previously been recommended only for regions where chemistry dominates transport in determining species distributions.

Figure 2: Distribution of ozone within the 360±2.5K isentropic layer in July 2005, 20N-90N from the Measurement of OZone and water vapour by AIrbus in-service airCraft (MOZAIC; Marenco et al 1998) (black), collocated assimilation of EOS Aura data (gray), and a control model run with ozone data withheld (blue). A bimodal structure, indicative of the presence of both, tropospheric and stratospheric air, is seen in all three datasets. In this example assimilation corrects model’s biases by changing the (relative) position of the distribution modes. The result uses assimilation of ozone data from EOS Aura’s OMI (total column) and MLS (215 – 0.14 hPa profiles) into NASA’s GEOS-4 Data Assimilation System (Stajner et al., 2008). (Figure courtesy of Kris Wargan.) For better resolution, please contact the SPARC IPO

The mesosphere

With the recent availability of measurements from ACE and EOS-MLS that extend into the mesosphere, the performance of assimilation in the stratopause region can be assessed. G. Manney examined the performance of operational systems such as ECMWF and GEOS-4 and -5 as well as research assimilation systems (CMAM-DAS and NOGAPS-ALPHA) during a prolonged Stratospheric Sudden Warming (SSW) event in early 2006.  Figure 4 shows that while the operational systems had difficulty in capturing the timing or the height of the reformation of the stratopause on 1 February 2007, the CMAM-DAS fared better, even though no mesospheric data was assimilated.  The NOGAPS-ALPHA model also did well (except immediately below the model top), but for a different reason — in NOGAPS-ALPHA, SABER and EOS-MLS temperature data were assimilated in the mesosphere.  Thus, the higher lid of the CMAM model combined with reasonable mesospheric dynamics may provide a natural boundary condition for stratospheric assimilation.  This implies that study of the upper stratosphere requires a model with a very high top, and either a reasonable mesosphere or assimilation of mesospheric measurements.

An overview of mesospheric dynamics was provided by C. McLandress.  In contrast to the lower atmosphere, here tides and gravity waves are important and lead to large and rapid dynamic variability.  Thus good representation of the mesosphere requires that a model properly depict the various tidal modes and support a realistic gravity wave spectrum.  Models may also employ gravity-wave drag (GWD) schemes to parameterize the effects of subgrid scale gravity waves on the mean flow.   In the tropics, equatorial wave spectra are largely controlled by convective parameterizations (Horinouchi et al. 2003).  Thus convection schemes may impact the zonal wind oscillations in the tropics, as well as modelled tides.   Issues for assimilation include not only the large variability of the mesosphere, but also sampling issues for sun-synchronous satellite orbits.  R. Lieberman found that combining SABER and EOS-MLS measurements could help improve the analysis of the diurnal tides.  However, averaging over longer time intervals reduces the “added benefit” of the second measurement source.  Capturing tides with assimilation is also complicated by the fact that harmonics of up to four per day contribute to the tidal field, as noted by W. Ward and Z. Chen.  V. Yudin reviewed the problem of generation of biases in the stratosphere and mesosphere due to inconsistent vertical scales in background error vertical correlations and weighting functions or Jacobians for nadir temperature sounders, and suggested the use of a singular vector decomposition of Jacobians to ensure a natural tapering of increments.  Yudin also suggested that mesospheric assimilation might proceed through an initial extraction of fast tidal amplitudes followed by zonal mean wind and planetary wave analysis. 

Obtaining a good mesospheric analysis may be aided by the upward transfer of information.  This notion of the slaving of large-scale aspects of the mesosphere to the lower atmosphere was considered by S. Ren for the case of the 2002 Southern Hemisphere SSW.  Information can be propagated by the model’s GWD scheme, as data assimilation can help define the large scale tropospheric and stratospheric flow which then filters upward propagating wave fluxes.  The slaving of fast motions to slow ones was examined with a low-order model by L. Neef. When the true state was unbalanced (as in the mesosphere) and consisted of both time scales of motion, 4D assimilation schemes showed some advantage over 3D ones. 

The predictability of the mesosphere was also considered by Y. Nezlin who showed that for a perfect model and perfect observations, the propagation of information into the mesosphere can be quantified and that the information is primarily on the largest scales.  K. Hoppel also showed that there is some value in performing mesospheric data assimilation.  Figure 5 shows that forecasts initialized from an analysis are better than those initialized from climatology, even in the mesosphere.  In addition, the forecasts are better whether in the winter or summer hemisphere where the dynamics are very different.

The structure and evolution of the polar vortex across the stratopause were described by L. Harvey, who showed the evolution of a diagnostic of the polar vortex edge and anticyclones in three dimensions up through ~70 km from GEOS-4/5 analyses and the WACCM GCM during both Arctic and Antarctic winters.  Preliminary comparisons of vortex edge diagnostics and MLS CO along orbit cross-sections showed encouraging agreement into the mesosphere.  G. Manney also showed good agreement in vortex structure and MLS CO across the stratopause; mesospheric tracer data such as those from MLS are thus shown to be valuable for verification of analysis characteristics in the mesosphere and stratopause region.

Top

Figure 3: One dimensional ozone power spectra computed from 4000 km MOZAIC flight segments (black), collocated assimilation of Aura data (gray), and model (blue) in a) March and b) July 2005. The aircraft data were averaged to match the resolution of the model. Steeper decline of model and assimilation spectra indicates that they exhibit less small scale variability than MOZAIC data. Note that in July, the assimilation has more variability than the model at all scales; that is consistent with increased difference between the ozone distribution modes shown in Figure 2.  (Courtesy of Kris Wargan, GMAO). For better resolution, please contact the SPARC IPO

 

Figure 4:  Pressure-time sections at 70N of zonal mean temperature from (top to bottom, left to right) MLS, SABER, GEOS-5, ECMWF, CMAM-DAS and NOGAPS-ALPHA, from 1 January 2006 through 31 March 2006.  Overlaid contours on analyses panels are 70N zonal mean zonal winds from -60 to 90 m/s by 30 m/s, with easterlies and zero values in black, westerlies in white.  NOGAPS-ALPHA run shown here assimilated  MLS and SABER temperatures up to 0.01 hPa.  CMAM-DAS run assimilates no observations above 1 hPa. (Courtesy of Gloria Manney, JPL) For better resolution, please contact the SPARC IPO

Limitations of DA in the tropical stratosphere

Part of the challenge of obtaining good tropical analyses is that balance relationships are more complex.  Such balance relationships can be used in the extra-tropics in background error covariances to spread information from the mass field to the wind field and vice versa.  H. Körnich suggested that improvement in tropical analyses may be possible by accounting for tropical waves in the background error covariance estimates.  An analysis of tropical waves in free model runs (CMAM and GEM) revealed that tropical waves can be identified, but that the variances due to different modes depend on height, the model used, the QBO phase and tides.  Körnich also noted that while taking tropical modes into account in background error covariances may be beneficial, wind measurements are needed for the covariances to be really effective.  This reflects the fact that wind rather than temperature controls tropical dynamics, as noted in the previous paragraph. Thus, new missions such as ADM-Aeolus or SWIFT which propose to measure winds could help address the issue of poor quality tropical analyses.  M. Reszka discussed another means of improving mass-wind balance—that of enforcing a strong constraint (Charney balance and the quasi-geostrophic omega equation) on analysis increments.  While this approach may help in the extra-tropics, further extensions (including the estimation of diabatic forcing from convective parameterizations) are needed for this approach to adequately deal with tropical balance issues. 

Figure 5: This figure shows the RMS error of the NOGAPS-ALPHA temperature forecast, verified against the NAVDAS assimilation. Forecasts (a) & (c) were initialized from the assimilation.  Forecasts (b) & (d) were initialized from the assimilation below 10 hPa, and above 10 hPa with a zonal mean climatology based on UARS-URAP and CIRA climatologies. Temperature data from the AURA-MLS and SABER instruments were assimilated between 32 hPa and 0.01 hPa.  (Courtesy of Karl Hoppel, NRL.) For better resolution, please contact the SPARC IPO
Figure 6: Zonal mean STD of 6-h forecast errors in the CMAM-DAS computed with the use of simulated truth. All observations from the actual network were simulated by perturbing the truth with random values based on observation error covariances.   Panel (a) shows the temperature error standard deviation in K while panel (b) shows the wind error standard deviation in m/s. (Courtesy of Yulia Nezlin, U Toronto) For better resolution, please contact the SPARC IPO

Top

Chemical Data Assimilation

While operational forecast centres are primarily concerned with weather forecasts, in the future their interests will be much more far reaching.  S. Lu noted that NCEP plans to obtain global estimates of the distribution of atmospheric aerosols as well as forecasts of chemical species and of dust.  Their plans for aerosol forecasting and assimilation, which are motivated by the desire to capture aerosol radiative impacts, were described.   As operational weather centres move toward operational environmental forecasts in the future, it is important to determine which feedbacks must be simulated online and which ones can be neglected for computational expediency.  For example, should chemical data assimilation be performed online in a GCM or offline with a CTM?  With online chemistry assimilation, species are updated using measurements and then used for input to the radiation scheme to potentially improve the dynamical analyses.  R. Ménard, using an NWP forecast model (GEM) with full online chemistry, found that the impact of the ozone radiative feedback on temperature analyses was not significant if only temperature data was assimilated.  However, if both ozone and temperature were assimilated, then the ozone radiative feedback on temperature was significant.  Ozone analyses were significantly improved as was temperature predictability in the lower stratosphere.  M. Parrington showed that assimilation of TES ozone profiles had an impact on outgoing longwave radiation in a chemistry climate model (AM2-Chem) which could be as large as 15% compared to the case without ozone assimilation.  An ozone climatology was used in the radiation calculations so the impact was due to changes in circulation due to the use of assimilated ozone. C. Long showed that by assimilating OMI in addition to SBUV/2 ozone measurements with the NCEP system, not only was the total ozone analysis improved (as expected), but there was also an impact on tropical forecasts.

As noted earlier in the UTLS section, model forecasts can capture the horizontal structure of constituent distributions. This means that the use of model trajectories in a 4D-Var system could potentially help improve analyses in regions where measurements are sparse, such as the tropics.  The prospect of improving winds through constituent assimilation in a 4D-Var system was discussed by J. de Grandpré, who showed an improvement in zonal wind bias in the tropics when assimilating O3, CH4 and N2O from MIPAS.  However, in the extra-tropics, species assimilation in 4D-Var had a negative impact on ozone analysis and predictability.

With the recent availability of multiple species measurements from instruments such as those on ENVISAT, EOS-AURA or ACE, the challenges of ssimilating multiple species can be tackled.  Just as there are mass-wind balances to consider when assimilating dynamic variables, multiple species assimilation may also need to consider how the adjustment of one species affects another.  S. Chabrillat (in a presentation given by Y. Rochon) found that assimilation of a short lived species, NO2, helps the NO2 analysis but has a very negative impact on HNO3 and probably ClONO2.  However, A. Robichaud noted that assimilation of NO2 alone was able to improve 6-hour forecasts of HNO3 as well as capture the mesospheric/thermospheric descent of NOx during an energetic particle precipitation event.

Data assimilation can ideally provide feedback on the quality of not only the models but also the observing system.  F. Baier, using some observing system simulation experiments, found that a better distribution of ozone sondes was preferable to more frequent observations at existing locations.  S. Chabrillat (presented by Y. Rochon) found that MIPAS-IMK assimilations generally compared better to independent measurements than MIPAS-ESA assimilations.  J. Schwinger showed that assimilation can be used to validate one sensor by assimilating a second one with good spatial coverage.  The example of MIPAS ozone assimilation for HALOE validation was used.  This type of cross-validation requires that assumptions made in the assimilation process regarding the validity of specified biases and covariances be checked first.  The payoff is that coincidence of different sensors is no longer an issue since analyses are globally defined. 

IPY – International Polar Year

The objective of the SPARC-IPY project is to obtain a description of the two polar vortices (in terms of dynamics, chemistry and microphysics) during the March 2007 to March 2009 period.  There is special emphasis on the coupling of the stratosphere and mesosphere as well as the stratosphere and troposphere.  To achieve this goal, SPARC will acquire and archive measurements and assimilation products during the IPY period.  The current contents of the archive of assimilation products include analyses from ECMWF, NCEP, Met Office and GMAO.  The archive will also include analyses from CMAM-DAS and GEM-BACH as well as KNMI ozone analyses.  This archive, recently described in SPARC newsletter no. 29, was set up and monitored by D. Pendlebury.  The web interface for data access will be hosted by the SPARC data centre website. 

An important feature of Arctic polar dynamics is SSWs.  M. Baldwin explained their impact on tropospheric weather, and S. Ren and K. Hoppel considered their vertical coupling with the mesosphere in assimilation experiments. Also L. Harvey and G. Manney showed aspects of changes in the mesospheric flow during SSWs in operational analyses and satellite data. T. Chshyolkova noted that while analyses are very useful for diagnosing and understanding the dynamics of polar vortices, vertical extension of operational products is needed to better understand the coupling of the stratosphere and mesosphere during these events.  Another area where assimilation needs improvement is in the depiction of polar ozone depletion.  C. Benson noted that large AIRS observation-minus-forecast residuals are often associated with the presence of PSCs as observed by POAM III. 

showed once this factor is taken into account the agreement on ozone measured by different ground-based instruments and ACE FTS improved.Three poster presentations highlighted stratospheric observations at PEARL. R. Batchelor showed HF, HCl column measurements by a new Bruker-IFS at PEARL in which low values of HF reflects descent inside the polar vortex while decrease in HCl column amounts suggest conversion to active chlorine which results in O3 chemical depletion. A. Fraser focused on observations by a UV-Visible spectrometer at PEARL and showed slant column densities of O3, NO2, BrO and OClO and their comparison with ACE FTS and MAESTRO measurements during the past three ACE Arctic validation campaigns. The total ground-based columns are expected to agree with satellite measurements within error for ozone and within ~15% for NO2. W. Ward studied the wave environment in Arctic region and the coupling of the dynamics between atmospheric layers and locations. On his wavelet spectra plot meteor radar wind signatures over PEARL show a strong diurnal signature whereas this signature was absent in Saskatoon data.

In her ACE satellite mission overview talk, K. Walker presented the ozone evolution mapped by ACE FTS during winter/spring 2005 and their comparison to MLS and SAGEIII.  Also a first global picture of phosgene (COCl2), the product of chlorocarbon decomposition, measured by ACE FTS was presented.

R. Collins’ presentation demonstrated that high-resolution temperature data from a network of Arctic lidar observatories can aid in the study of the coupled tropospheric, stratospheric and mesospheric circulation.  A statistically significant long-term cooling of the middle atmosphere over the past 19 years at Haute Provence was shown. Also several prototype studies on vortex and anticyclone interactions manifested in temperature fields were presented together with observations of zonal wind reversal in the zonal mean during stratospheric warming events. These studies will be continued during the IPY. Also looking at temperature data, Y. Cho showed that the UKMO assimilated data indicates the negative relationship between the lower stratosphere and stratopause temperature. The relationship between the lower-upper stratosphere and MLT region temperature also can be seen in the SABER satellite measurements. As part of the IPY project the SPARC Data Center also hosts polar observational data sets in addition to the above-mentioned analyses products. Based on the discussion in the IPY session, the data archive was decided to be a hybrid of a web portal and an online library which will serve both as a home to observational data with no current permanent archive and as an archive for monthly mean data sets which are mature (e.g. radar and lidar observations). There were also discussions concerning logistical issues regarding providing data with high temporal resolution for specific period. Data providers are responsible for including meta data statements while a medium-level quality control is done by SPARC scientific data managers. Also the possibility of organising a second SPARC-IPY workshop in the Arctic circle with focus on status of current polar observations was proposed. Finally, the development of an outreach programme in collaboration with other related IPY activities such as IASOA was discussed.

NRT availability of research satellite measurements

N. Livesey and A. Lambert (presented by G. Manney) discussed plans for near real time (NRT) availability of EOS-MLS data.  Data assimilators in the audience showed considerable interest in NRT data access.  The question of why NRT data provision was not considered at an earlier stage in the life of the satellite mission was also raised, in view of the demand for MLS data. This is an emerging issue with all research satellite missions, as data assimilation centres are increasingly showing an appetite to assimilate research satellite products including species measurements. While these products are still useful after the fact for validation, their use is enormously enhanced if they are available in NRT in order to be used in the operational cycles. Furthermore this offers tremendous benefits to the measurement team, as the statistical analysis inherent in ongoing assimilation is one of the most effective ways of identifying changes in measurement characteristics. Unfortunately, this opportunity tends to fall between the cracks, as the space agencies do not consider NRT availability as part of their mandate for research satellites. Yet the additional cost involved is a relatively small fraction of the overall cost of the mission, so this is a lost opportunity for atmospheric science. SPARC needs to work as an advocate of the principle that NRT availability should be a basic requirement of all research satellite products.

Next meeting

There will be no SPARC-DA workshop in 2008.  Instead all participants are encouraged to attend the SPARC General Assembly to be held in Bologna, Italy during 31 August to 5 September 2008.  The next SPARC-DA workshop will be held in 2009, most likely in the fall.

Acknowledgements

We are grateful to the Fields Institute for the use of their facilities and for their creation of the workshop website.  All presentations are viewable (with sound) at http://www.fields.utoronto.ca/programs/scientific/07-08/data_assim/ .  Environment Canada provided substantial funding.  WCRP and the SPARC International Project Office provided travel support for some participants.

K. Wargan acknowledges for their strong support the European Commission, Airbus and the Airlines (Lufthansa, Austrian, Air France) who carry free of charge the MOZAIC equipment and perform the maintenance since 1994. MOZAIC is supported by INSU-CNRS (Institut National des Sciences de l'Univers - Centre National de la Recherche Scientifique, France), Météo-France, and FZJ (Forschungszentrum Jülich, Germany).

Work at the Jet Propulsion Laboratory, California Institute of Technology was done under contract with the National Aeronautics and Space Administration.

References

Baldwin,M.P. and T. J. Dunkerton, 2001: Stratospheric Harbingers of Anomalous Weather Regimes, Science,294 (5542), 581-584. DOI: 10.1126/science.1063315

Horinouchi, T., S. Pawson, K. Shibata, U. Langematz, E. Manzini, M.A. Giorgetta, F. Sassi, R.J. Wilson, K. Hamilton, J. deGrandpre, A.A. Scaife, 2003: Tropical comulus convection and upward-propagating waves in Middle-Atmospheric GCMs, Journal of the Atmospheric Sciences60 (22),  2765-2782.

Kistler, R.,  E. Kalnay,  W. Collins,  S. Saha,  G. White,  J. Woollen,  M. Chelliah,  W. Ebisuzaki,  M. Kanamitsu,  V. Kousky,  H. van den Dool,  R. Jenne, and M. Fiorino, 2001: The NCEP–NCAR 50–Year Reanalysis: Monthly Means CD–ROM and Documentation.  Bulletin of the American Meteorological Society, 82, 247–267. DOI: 10.1175/1520-0477(2001)082

Marenco, A., V. Thouret, P. Nédélec, H. Smit, M. Helten, D. Kley, F. Karcher, P. Simon, K. Law, J. Pyle, G. Poschmann, R. Von Wrede, C. Hume, T. Cook,. 1998: Measurement of ozone and water vapor by Airbus in-service aircraft: The MOZAIC airborne program, An overview, J. Geophys. Res., 103(D19), 25,631–25,642. (98JD00977)
 

Stajner, I., K. Wargan, S. Pawson, H. Hayashi, L.-P. Chang, R. C. Hudman, L. Froidevaux, N. Livesey, P. F. Levelt, A. M. Thompson, D. W. Tarasick, R. Stübi, S. B. Andersen, M. Yela, G. König-Langlo,F. J. Schmidlin, and J. C. Witte (2008), Assimilated Ozone from EOS-Aura:  Evaluation of the Tropopause Region and Tropospheric Columns, J. Geophys. Res., submitted.

Scheele, M.P., P.C. Siegmund, P.F.J. van Velthoven, 2005:  Stratospheric age of air computed with trajectories based on various 3D-Var and 4D-Var data sets.  Atmospheric Chemistry and Physics, 5, 1-7.

Tibaldi, S and F. Molteni, 1990: On the operational predictability of blocking, Tellus A, 42, p.343.

Zagar, N., 2004: Assimilation of Equatorial Waves by Line-of-Sight Wind Observations. J. Atmos. Sci. 61, 1877-1893.