Investigation of Chemical and Dynamical Changes in the Stratosphere up to and During the EOS Observing Period

1.0 Background

1.1 Previous reports

Our 1996 report is located on the world-wide web at http://hyperion.gsfc.nasa.gov/EOS/report_1996/report.html. The 1995 report is located at http://hyperion.gsfc.nasa.gov/EOS/report_1995/report.html.

Upper Atmosphere Research Satellite (UARS) data combined with aircraft mission data is the best prototype for the EOS CHEM instruments. Preparing for EOS CHEM data and chemical assimilation is the main focus of this investigation. Thus the tools this IDS has developed and our experience with UARS data is directly applicable to EOS program. A large part of our team is focused on the UARS data set (Schoeberl, Douglass, Jackman, Rood, Geller, Newman, and Lait) as described below. Our group also has a large involvement in aircraft expeditions. Members of our IDS team participated in the AASE, AASE II, ASHOE/MAESA, TRACE A, STRAT, POLARIS, TOTE/VOTE, PEM Tropics, and SPADE aircraft missions and analysis of that data. We believe that aircraft and satellite data are complimentary and analysis of aircraft data is consistent with the interdisciplinary aspect of our proposal.

Model and data analysis tool development has been an important activity under this IDS. For example, an interactive 2D chemical model was developed under this IDS with new results discussed below. We have also participated in the 3D chemical modeling effort and have spun off a Lagrangian Chemical model which is being used to simulate ozone loss in the polar regions (see 1996 report). A lot of progress has been made on the stratospheric assimilation, and using trajectory models for age-of-air calculations as a diagnosis of the assimilation system.

In tropospheric chemistry we have focused mostly on development of better methods for satellite retrieval of tropospheric ozone as described below (Section 2.8).

In the 1999-2002 time frame, the ESA ENVISAT satellite is expected to be major source of new atmospheric trace constituent measurements. In particular the SCIMACHY instrument is expected to measure a significant number of these trace species. In preparation for SCIMACHY we have been analyzing the data from its predecessor instrument, GOME (Global Ozone Monitor) on the ERS-2 platform. The GOME instrument was a significant instrumental advance in that the instrument measures the complete backscattered spectrum from 240 to 800 nm. This new data set has lead to the development of new algorithms for the retrieval of atmospheric trace constituents. Specifically, we have been working with the GOME Team to adapt the NASA total ozone and ozone profile algorithms to the GOME radiance data, we can help separate instrumental from algorithimic artifacts in the retrieved products. We have helped the GOME Team identify algorithmic artifacts, such as how their total ozone values depend on the climatology used in their retrieval. We demonstrated that there were radiance jumps in their spectra that make ozone profile retrievals impossible without some type of spectral recalibration and adjustment. Future work will focus on the development of new data products that can be measured using the spectral information from GOME and SCIMACHY.

2.1 Trajectory models

The three-dimensional chemical model was developed under funding from the Atmospheric Chemistry Modeling and Analysis Program. The chemical package and transport scheme eventually will be used in the chemical assimilation effort. Thus, scientific testing of the 3-D CTM is essential for the success of the IDS assimilation program and also allows us to move toward our goal of using the 3-D chemical model for interpreting long-term trends.

There has been substantial progress in the 3-D chemistry and transport model in the last year. Full chemistry integrations of the model include ozone, reactive constituents important to the ozone evolution, and long lived source gases (Douglass et al., 1997). Because the winds and temperatures used in this off-line simulation are taken from the GEOS-1 data assimilation system, the calculated constituent evolution is sensibly compared with observations. Simulations are complete for the following time periods: Nov. 15, 1991 - Sept. 1, 1992; Nov. 15, 1995 - April 22, 1996; March 18, 1997 - present.

The simulations are directed towards several goals. Both models and observations show interannual variability in lower stratospheric ozone loss in the northern hemisphere (e.g., Manney et al., 1994; Manney et al., 1996; Chipperfield et al., 1996). Simulations for 1991-92 and 1995-96 are being compared with observations to determine if the model is capable of correctly reproducing the interannual variability. In addition, the 95-96 simulation is being compared with VOTE/TOTE aircraft observations described below. Aircraft data was also collected during the Atmosphere-Ocean Chemistry Experiment (ACE) during April 1996. Flights were made through a midlatitude cut off low and observations of this event were presented by Moody et al. [1997]. Fields from the global simulation are being compared with these observations to determine if the model constituent distribution in the upper troposphere evolves realistically during this event, thus providing insight into the modeling of processes associated with stratosphere/troposphere exchange.

The current simulation which was begun March 18, 1997 is being run in near real time in support of the POLARIS mission. In addition, each day a five day forecast of the ozone field is calculated using the result of the full chemistry integration to initialize the forecast, the forecast winds from the DAO, and a parameterization for photochemical production and loss. In general the forecast fields for March 20, 1997 - May 18, 1997 compare well with observations from ADEOS TOMS data, although the quality of the comparison degrades after about a month of integration. Causes for the loss of forecasting skill are being considered currently.

Part of this IDS investigation consists of providing timely feedback to the Goddard Data Assimilation Office (DAO) about remaining problems in the stratosphere Data Assimilation System (DAS). Response to this feedback generates improved stratospheric assimilations. User feedback is an important part of improving a DAS as users bring a wide range of experience, diagnostic tests, transport and trajectory models, and independent (unassimilated) observations. Improving the multi-year transport in the DAS is important for assessing potential climate change issues. This IDS investigation is continuing its close work with the DAO in its effort to improve the transport characteristics of the GEOS-2 DAS.

This past year has seen the development, testing, and validation of the GEOS-2 (Goddard Earth Observing System -2) Data Assimilation System (DAS). Many of the features in this new DAS have been specifically designed and tuned for improving stratospheric data assimilation in response to feedback provided by this IDS investigation. The GEOS-2 DAS is based on the new 70 level GEOS-2 GCM (General Circulation Model) coupled with the new Physical-Space Statistical Analysis System (PSAS). The GCM now includes an orographic gravity wave drag parameterization and Rayleigh friction tuned to eliminate the cold pole bias of the GEOS-1 GCM. The upper boundary of the GEOS-2 GCM is at 0.01 hPa (about 80 km), though the number of data analysis levels in the DAS remains at 18 with the top analysis level at 0.4 hPa (about 55 km).

In the 1996 EOS IDS report we have given an account of the development of a Kalman filter for stratospheric constituent assimilation of limb sounding observations (Lyster et al. 1997). Also shown was its potential in providing useful global field estimates from limited coverage observations. Since that report, the Kalman filter has undergone a statistical validation of its output error estimates against the statistics of the observed-minus-forecast assimilation residuals (Menard et al. 1997). This statistical validation study has enabled us to identified several deficiencies in the original algorithm and, hence, resulted in several modifications of the scheme.

A key feature of the Kalman filter is to evolve the estimation error covariance using a dynamical model, namely, the transport model of Lin and Rood (1996). Some deficiencies of the original algorithm were related to discretization of the covariance field. The evolution of error covariance in the continuum was studied first by Cohn (1993) and extended later to include the observation counterpart in Cohn (1997). The statistical validation conducted during this year has shown that very large discretization error arises in the prediction of the error covariance, resulting in a significant loss of skill in the estimated analysis error variance. On the basis of Cohn (1993), a variance preserving scheme has been developed and successfully implemented. In addition, several other alternative schemes which have conservative properties of the correlation field are under development. The mismatch between the horizontal sampling of the continuum inherent in satellite observations and in the representation of the continuum by the transport model results in "error of representativeness", that needs to be accounted for during the assimilation. It was found that the representativeness error standard deviation for the CLAES methane with a 4x5 degree finite volume grid point model is about 10% of the signal, with a slightly lower value for the HALOE methane. The represenativiness error is larger than the instrument error for the gas and altitude considered (3.5% for the CLAES and 0.7% for the HALOE). Although, it is not yet clear how the representativeness error can be significantly reduced in practice so as to maximize the use of satellite observations in data assimilation, a preliminary study by Peter Lyster (personal communication) has indicated that for point measurements, the representativeness error only slowly decrease with increasing resolution. This, in turn, demands that very high resolution models be used in data assimilation schemes in order to take full advantage of point observations of constituents.

Development of an operational three-dimensional ozone assimilation system is in progress at the DAO. The initial goal is to enable the DAO to provide high quality near-real time three-dimensional ozone fields as input to the EOS instrument teams. In the longer term we expect to use the data generated by the system also for process and trend studies. The assimilating model is the CTM developed jointly by the Atmospheric Chemistry and Dynamics Branch and the DAO at Goddard (Douglass et al., 1996; Lin and Rood, 1996). The model is driven by GEOS-DAS assimilation wind fields, and the input observations are TOMS total ozone and SBUV ozone profiles.

The analysis system itself is mainly designed along the lines of PSAS (Physical-space Statistical Analysis System; Cohn et al., 1997), the DAO meteorological analysis system. However, there are a couple of important differences with respect to the current implementation of that system. The ozone assimilation is being designed to work as a filter with continuous updates at every time step, rather than using a six-hourly forecast/analysis cycle as is common in assimilation for numerical weather prediction. The reason for this is that the ozone system is driven by satellite data obtained on a continuing basis, as opposed to meteorological systems that still rely extensively on conventional observations taken at the synoptic times. The method of continuous updates is computationally more efficient than the intermittent cycling, and it will also minimize phase errors due to the difference between the analysis time and the actual time of the observation. However, the ozone assimilations system is expected to be more demanding than an intermittent system, as far as the validity of the forecast error covariance model is concerned. The forecast error covariance matrix formally contains N*(N+1)/2 independent elements, where N is the number of predicted variables - on the order of 10⁶ for the ozone system, depending on resolution. This is both computationally and conceptually an unreasonable amount of information to generate and manipulate, and in current meteorological analysis systems the covariance is therefore modeled assuming isotropy, and often also homogeneity, of the forecast error. This greatly simplifies the problem, but the shape of the analysis increment then no longer depends on the actual state of the atmosphere. In particular, the increment added to the background field in order to obtain the analysis based on an isolated observation will be isotropic, a feature that is clearly unphysical, for instance, near a front.

2.4.1 Fixed Two Dimensional Model Studies

An important goal of our IDS investigation is to understand the changes which have occurred in the concentration of stratospheric ozone. To do this we use a variety of 2D, 3D, and trajectory models. Our understanding of the processes which control ozone is incorporated into models which can be used to make predictions of the changes of ozone which might be expected in the future. These models are usually termed "assessment" models. They always contain simplifying assumptions and are currently mostly 2D (latitude and altitude).

The GSFC fixed climatological 2D model is primarily funded through ACMAP, although there is partial EOS IDS support. The Goddard 2D assessment model has recently been reformulated with a new derivation of the meridional circulation and eddy coefficients (Fleming et al., in preparation, 1997). The new formulation is similar to that which has been used in the 2D coupled model developed under this proposal. The model dynamics are derived from an elliptic streamfunction which is based on a solution of the zonal momentum and energy equations along with the thermal wind equation. The temperatures, zonal-mean winds, and EP flux divergences are derived from data (17-year average of NMC data plus zonal mean winds in the equatorial region from HRDI on UARS). Thus, although the dynamical formulation is the same as the coupled model, dynamical feedbacks are not included and the model is constrained to the present atmosphere.

The 2D model develops a number of features which are qualitatively consistent with the behavior of stratospheric transport as derived from data analyses. In the southern hemisphere winter and early spring, the model develops a clearly isolated vortex and midlatitude surf zone as exhibited by the methane mixing ratios shown in the accompanying figure. The equivalent behavior in the northern hemisphere does not occur. The model develops only the barest hint of the development of an isolated vortex. This is presumably because the derivation of the EP flux divergences has been done using the wave amplitudes derived from variations about a zonal mean. In the northern hemisphere, this means that near the vortex edge data from within and without the vortex is mixed together leading to enhanced mixing in the derivation.

Another feature which arises naturally in the formulation of this model is the partial isolation of the tropical region from the subtropics. A preliminary examination shows that the air rising in the tropics in the model has about 20-25% incorporated into the flow from the midlatitudes. This makes the model tropics somewhat more isolated than has been derived from data analyses. Analysis of the age of air derived from tracers such as CO2 in the model indicate that the oldest air in the model (age = time since entering through the tropical tropopause) is about one year younger than derivations from ER-2 CO2 data. This is possibly because the model does not recirculate enough of the midlatitude air through the tropics where it can be relofted to high altitudes. We are now at the point of examining specific deficiencies in the assessment models.

Improvements to the interactive 2d model in 1997 include the attachment of the new chemistry package (used in the fixed 2D model) and revision of the tropospheric parameterization. We continue to collaborate with NRL on the development of the model.

Research with the interactive 2D model developed under this IDS has focused on studying the radiative effects of tropical subvisible cirrus clouds. These thin clouds have been observed at and just below the tropical tropopause by the DIAL lidar during the TOTE mission, as well as by in situ [Heymsfield, 1986] and satellite observations [Wang et al., 1994]. Our coupled model forms ice clouds not only in the polar stratosphere but also in the tropical troposphere just below the tropopause. In order to study the potential radiative impact of these clouds we have performed model runs in which the radiative heating due to these clouds is included. The ice cloud heating has been calculated using the method described in Rosenfield [1992].

During the previous reporting period, high-resolution (1 degree by 1 degree) global fields of potential vorticity were generated on the 560 K isentropic surface. Created using the Reverse Domain Filling (RDF) technique, these fields exhibit realistic fluid-dynamical features. They were produced for every six hours covering a span of 30 days.

These RDF data have now been used to estimate the kind of improvements HIRDLS high resolution sampling would achieve. Thirty days of satellite sounding locations were calculated, and the RDF fields were interpolated to those locations. The simulated measurements were then used to construct regularly gridded fields, using Delaunay triangulation and subsequent bilinear interpolation to the grid points. These data, gridded from the simulated observations, were then compared to the original RDF fields.

Four modes of satellite sampling were compared: The regular HIRDLS scanning pattern, the high-resolution HIRDLS scanning pattern, a hypothetical single-scan mode, and a hypothetical ultra-high-resolution scanning mode.

2.7.1 Errors in ozone depletion due to small scale structure in chlorine

Aircraft observations of active chlorine show intense spatial variability on scales well below the resolution of global model grids. Edouard et al. [1996] proposed that the lack of resolution of these small scale features in numerical models produced an underestimate in ozone loss. In the work described below, the statistics of subgrid scale structure in active chlorine were computed from high resolution aircraft observations and were used to estimate the sensitivity of modeled ozone loss to the spatial resolution of the model grid.

A large fraction of the total ozone loss in the winter lower stratosphere occurs via the chlorine catalytic cycle, a process that is limited by the rate of ClO dimer formation. The nonlinearity of this process, together with the existence of small scale structure well below the spatial resolution of most model grids, mean that a model calculation of ozone loss will underestimate, to some extent, the actual ozone loss in the atmosphere. High resolution aircraft observations of ClO taken during the AASE II aircraft campaign were used to estimate the error in the ozone loss rate due to subgrid scale structure in the ClO field. Because the ozone loss rate depends quadratically on the concentration of ClO, which we denote below by , the low resolution loss rate for a model gridbox is proportional to <>², while the actual loss rate is proportional to <²>. Here, <...> means an average over an LxL gridbox, which we approximate from the data by an average over the observations within data segments of varying lengths L. The difference between the high and low resolution loss rates is the local variance of , and we define a relative resolution error = var()/<>².

With co-funding from ACMAP, a modified-residual technique has been developed by Hudson and Thompson [1997] to retrieve half-monthly averaged tropical tropospheric column ozone from the TOMS radiances. This technique involves the extraction of a persistent wave-one pattern, with its maximum over the tropical Atlantic, as in Kim et al [1996]. Several improvements to the Kim et al. [1996] method have been incorporated in the modified-residual method. These include: (1) Fourier analysis of the wave pattern to set latitude limits for where the method is valid; (2) assumption that the wave-like pattern is in the troposphere; (3) normalization to the ozonesondes for 1991-92, the years for which the new method is complete. Because of the ozonesonde constraint, the derived troposphere ozone column amount can be compared to observations and imprecision assigned to the maps. The modified-residual method is now being applied to the 14-year Nimbus 7/TOMS record.

Chipperfield, M. P., A. M. Lee, and J. A. Pyle, Model calculations of ozone depletion in the Arctic polar vortex for 1991/92 to 1994/95, Geophys. Res. Lett., 23, 559-562, 1996.

Cohn, S. E., Dynamics of short-term forecast error covariances. Mon. Wea. Rev., 121, 3123-3149, 1993

Cohn, S. E.: An introduction to estimation theory. J. Met. Soc. Japan, 75, 257-288, 1997.

Cohn, S. E., A. da Silva, J. Guo, M. Sienkiewicz, and D. Lamich Assessing the Effects of Data Selection with the DAO Physical-space Statistical Analysis System; (submitted to Mon. Wea. Rev.), 1997.

Douglass, A. R., C. J. Weaver, R. B. Rood, and L. Coy, A three-dimensional simulation of the ozone annual cycle using winds from a data assimilation system. J. Geophys. Res., 101, 1463-1474, 1996.

Douglass, A. R., R. B. Rood, S. R. Kawa, and D. J. Allen, A three dimensional simulation of the evolution of the middle latitude winter ozone in the middle stratosphere, J. Geophys. Res., 102, 19,217-19,232 ,1997.

Edouard, S., B. Legras, F. Lefevre and R. Eymard, The effect of small-scale inhomogeneities on ozone depletion in the Arctic., Nature, 384, 444-447, 1996.

Heymsfield, A.J., Ice particles observed in a cirriform cloud at -83 C and implications for polar stratospheric clouds, J. Atmos. Sci., 43, 851-855, 1986.

Jensen, E.J., et al., Dehydration of the upper troposphere and lower stratosphere by subvisible cirrus clouds near the tropical tropopause, Geophys. Res. Lett., 23, 825-828, 1996.

Hudson, R. D., and A. M. Thompson, Studies of the wave-one pattern in tropical total ozone: Implications for derivation of tropospheric ozone, J. Geophys. Res., manuscript in preparation, 1997.

Kim, J-H., R. D. Hudson and A. M. Thompson, A new method of deriving time-averaged tropospheric column ozone over the tropics using total ozone mapping spectrometer (TOMS) radiances: Inter-comparison and analysis using TRACE-A data, J. Geophys. Res., 101, 24317-24330, 1996.

Lin, S.-J., and R. Rood, Multidimensional flux-form semi-Lagrangian transport schemes. Mon. Wea. Rev., 124, 2046-2070, 1996.

Lyster, P.M., S.E. Cohn, R. Menard, L.-P. Chang, S.-J. Lin, and R.G. Olsen, Parallel implementation of a Kalman filter for constituent data assimilation. Mon. Wea. Rev., 125, 1674-1686, 1977

Manney, G. L., et al., Arctic ozone depletion observed by UARS MLS during the 1994-1995 winter, Geophys. Res. Lett., 23, 85-88, 1996.

Manney, G. L., et al., 1994: Chemical depletion of ozone in the Arctic lower stratosphere during winter 1992-93, Nature, 370, 429-434, 1994.

Menard, R., L.-P. Chang, P.M. Lyster, and S.E. Cohn, 1997: Stratospheric assimilation of chemical tracer observations using a Kalman filter. Submitted to J. Geophys. Res., 1997.

Moody, J. et al. Observations of stratospheric ozone under a persistent cut-off low: AEROCE Spring 1996, Eos Trans. AGU, 78(17), Spring Meet. Suppl., S98, 1997.

Mote, P.W., An atmospheric tape recorder: The imprint of tropical tropopause temperatures on stratospheric water vapor, J. Geophys. Res., 101, 3989-4006, 1996.

Newman, P., J. Gleason, R. McPeters, R. Stolarski, Anomalously low ozone over the arctic, GRL, in press, 1997.

Riishojgaard, L. P., A direct way of specifying flow-dependent background error correlations for meteorological analysis systems, Tellus (in press), 1997.

Rosenfield, J.E., Radiative effects of polar stratospheric clouds during the Airborne Antarctic Ozone Experiment and the Airborne Arctic Stratospheric Expedition, J. Geophys. Res., 97, 7841-7858, 1992.

Sparling, L. C., A. R. Douglass and M. R. Schoeberl, An estimate of the effect of unresolved structure on modeled ozone loss from aircraft observations of ClO, Geophys. Res. Lett., 1997, submitted.

Wang, P.-H., et al., Tropical high cloud characteristics derived from SAGE II extinction measurements, Atmos. Res., 34, 58-83, 1994.

Bevilacqua, R.M., C.P. Aellig, D. J. Debrestian, M. D. Fromm, K. Hoppel, J. D. Lumpe, E. P. Shettle, J. S. Hornstein, C. E. Randall, D. Rusch, and J. Rosenfield, POAM II observations of the Antarctic ozone hole in 1994 and 1995, J. Geophys. Res., in press, 1997.

Considine, D.B., A.E. Dessler, J.E. Rosenfield, P.E. Meade, M.R. Schoeberl, A.E. Roche, and J.W. Waters, Interhemispheric asymmetry in the 1 mbar O3 trend: An analysis using an interactive zonal mean model and UARS data, J. Geophys. Res., in press, 1997.

Menard, R., L.-P. Chang, P.M. Lyster, and S.E. Cohn, 1997: Stratospheric assimilation of chemical tracer observations using a Kalman filter., submitted to J. Geophys. Res, 1997.

Newman, P.A. and J.E. Rosenfield, Stratospheric thermal damping times, Geophys. Res. Lett., 24, 433-436, 1997.

Newman, P., J. Gleason, R. McPeters, R. Stolarski, Anomalously low ozone over the arctic, GRL, in press, 1997.

Rosenfield, J.E., D.B. Considine, P.E. Meade, J.T. Bacmeister, C.H. Jackman, and M.R. Schoeberl, Stratospheric effects of the Mt. Pinatubo aerosol studied with a coupled two-dimensional model, J. Geophys. Res., 102, 3649-3670, 1997.

Rosenfield, J.E., et al., The impact of subvisible cirrus clouds near the tropical tropopause on stratospheric water vapor, submitted to JGR, 1997

Schoeberl, M. R., C. H. Jackman, and J. E. Rosenfield, A Lagrangian estimate of aircraft effluent lifetime, submitted to JGR, 1997.

The automailer is a free service to atmospheric scientists. The automailer allows the downloading of stratospheric meteorological products such as winds, temperatures, potential vorticity, and the generation of maps as post script files. The automailer can also be used to run the GSFC trajectory model (developed and maintained under this IDS). Calculations are run on a dedicated Silicon Graphics workstation supported by this IDS. The usage statistics for 1996 are given below. There were about 2000 more accesses in this period than in the previous annual period.

Automailer usage covering 1996-01-01 to 1996-12-31

USER ACCESSES

TOTALS: 103 users, 7083 accesses

   FUNCTION   ACCESSES 
   curtain   	219
   metmaps   	143
   metprofile  	2582
   plot_met  	1965
   trajectory  	2174