Comprehensive Summary on the Workshop on Process-Oriented Validation of Coupled Chemistry-Climate Models
Veronika Eyring, DLR-Institut für Physik der Atmosphäre, Germany (Veronika.Eyring@dlr.de)
Neil R.P. Harris, European Ozone Research Coordinating Unit, UK (Neil.Harris@ozone-sec.ch.cam.ac.uk)
Markus Rex, Alfred-Wegener Institut-Potsdam, Germany (email@example.com)
Theodore G. Shepherd, University of Toronto, Canada (firstname.lastname@example.org)
David W. Fahey, NOAA Aeronomy Laboratory, Boulder, USA (email@example.com).
G. Amanatidis, J. Austin, M.P. Chipperfield, M. Dameris, P. Forster, A. Gettelman, H.F. Graf, T. Nagashima, P.A. Newman, M.J. Prather, J.A. Pyle, R.J. Salawitch, B.D. Santer, and D.W. Waugh
A number of Coupled Chemistry-Climate Models (CCMs) with detailed descriptions of the stratosphere have been developed over the last 5-10 years. As they can address how climate change, stratospheric ozone and UV radiation interact, now and in the future, a prime use of these models is to provide O3 and UV predictions for the WMO/UNEP and IPCC assessments. Simulating the interaction between chemistry and climate is of particular importance, because continued increases in greenhouse gases and a slow decrease in halogen loading are expected, which both influence the abundance of stratospheric ozone. Because CCMs have been developed with different levels of complexity, they produce a wide range of results concerning the timing and extent of ozone layer recovery (WMO, 2003). The models are required to simulate extremely complex processes that include quite subtle effects amid significant natural variability. In order for their results to be credible and treated with confidence, models must be carefully validated against measurements and other models. Recent validation work has shown both the benefits that can be gained and the problems that can be encountered.
CCMs simulate a climate that at best bears only a statistical relationship to the real atmosphere, and so a comparison of model results with measurements must be performed in a statistical manner in order to see how well natural variability is simulated. This is problematic, because it appears to take many decades of observations to define a robust stratospheric climatology, especially in the Arctic winter. While tropospheric climate models can be validated, in part, by their ability to reproduce the climate record over the 20th century, the paucity of stratospheric climate data prior to the satellite era (post-1979) severely restricts such possibilities for model validation of stratospheric ozone.
For these reasons, validation of CCMs needs a process-oriented basis to complement the standard comparisons of model and observed climatologies. By focussing on processes, models can be more directly compared with measurements. Furthermore, natural variability becomes an aid rather than an obstacle because it allows one to explore parameter space and, thereby, more readily identify cause and effect relationships within a model. An important example of this approach is the validation of chemistry and transport processes in both 2D and 3D models that is documented in the NASA Measurements and Models II Intercomparison (Park et al., 1999). In the context of stratospheric General Circulation Models (GCMs) (i.e., those without chemistry), process-oriented validation represents the level II tasks within the GCM-Reality Intercomparison Project for SPARC (GRIPS) (Pawson et al., 2000). A first attempt at process-oriented validation of stratospheric CCMs is summarised in the 2002 WMO/UNEP Assessment (WMO, 2003) and discussed in detail in Austin et al. (2003).
The development of a more comprehensive approach to CCM validation was the goal of a workshop held in November 2003 in Grainau, Germany. The workshop, titled "Process-oriented validation of coupled chemistry-climate models," attracted approximately 80 participants from Europe, the USA, Canada, Japan and New Zealand. A primary goal of the workshop was to build upon the existing foundation of validation efforts to achieve a more systematic, long-term approach to CCM validation needs. A brief workshop report was published in SPARC Newsletter N° 22 (Eyring et al., 2004).
Following this introduction, a more comprehensive summary of the workshop is presented. The summary includes a Table of Processes (Table 1), which was a primary objective of the workshop. This table lists the core processes for stratospheric CCMs within four main categories: dynamics, stratospheric transport, radiation, and stratospheric chemistry and microphysics. Processes associated with the Upper Troposphere/Lower Stratosphere (UTLS) are included under these categories. For each process, the table includes model diagnostics, variables relevant for validation and sources of observational or other data that can be used for validation. The accompanying text discusses the importance of the selected processes to CCM validation and the utility of the selected diagnostics in a validation study.
Several of the diagnostics have been applied before to a range of models, but many have not. Various criteria were used in selecting the primary diagnostics. The chosen diagnostics are associated with a well-understood model process and have reliable measurements available for validation.
Table 1. List of Core Processes to validate CCMs with a focus on their ability to model future ozone
Overall Coordination Veronika Eyring, Neil Harris and Ted Shepherd Process Diagnostic* Variables Data Dynamics Coordination: Martin Dameris and Paul Newman Forcing and propagation
of planetary waves
Wave Frequency Analysis (WFA)
Planetary Wave (PW) spectrum
(variances & co-variances)
Temperature, Geopotential Height,
High-frequency (daily) data
Met. Analyses ** Hemispheric Ozone Variability Indices Total column ozone over several years Satellite measurements of total ozone Stratospheric response to wave drag Annual cycle of temperatures in tropics and extra-tropics Zonal monthly mean temperature, residual stream function Met. Analyses **,
in situ and space-based observations, profile data
PW flux vs. polar temperature, lagged in time Heat flux (vT) at 100 hPa (Jan/Feb)
Temperature at 50 hPa (March)
Zonal monthly means
Vortex definition, structure & occurrence of sudden/final warmings Potential Vorticity, horizontal winds, Temperature, Area colder than PSC T, Vortex area/equiv. latitude
High-frequency (daily) 3D fields
Downward control integral, also scatter plot of PWD versus GWD w* from model
PWD, GWD, other drag
zonal and monthly means
Met. analyses **
total drag inferred from diabatic heating calculation
Persistence (e.g., leading EOFs), including Holton-Tan Geopotential Height, Temperature
Multi-year time series (means, frequency spectra)
QBO, SAO *** Amplitude and phase (SAO) of horizontal winds and temperature U and T, zonal and monthly means Met. analyses ** Stratospheric Transport Coordination: Markus Rex and Darryn Waugh Subtropical and polar mixing barriers PDFs of long-lived tracers N2O, CH4, F11, etc.; PV Satellite and in situ (aircraft, balloons) chemical measurements and Met. analyses Latitudinal gradients of long-lived tracers Correlations of long-lived tracers Phase and amplitude of subtropical CO2 or H2O annual cycle in lower stratosphere (tape recorder) CO2 or H2O Satellite and in situ measurements Annual cycle of streamer frequency Daily PV (maybe long-lived tracers) Met analysis, satellite measurements Meridional circulation Mean age Conserved tracer with linearly increasing concentration, SF6 or CO2 in situ measurements Correlation of interannual anomalies of total O3 and PW flux Total O3 and heat flux at 100 hPa, zonal and monthly means Satellite measurements,
Vertical propagation of tracer isopleths H2O or CO2 or idealised annually repeating tracer (tropics), CH4 or N2O (polar) in situ and ground-based (polar only) and satellite data Diabatic velocity, TEM streamfunction Diabatic velocity, residual velocities Diabatic velocity inferred from radiative calculation UTLS transport Vertical gradients of, and correlations between, chemical species in the extratropical UTLS CO2, SF6, H2O, CO, O3, HCl Balloon, aircraft Relation between meteorological indices (e.g. tropopause height) and total ozone Daily winds, temperature, Z, total O3 Met. Analyses**
Satellite measurements, ozonesondes
Diabatic velocity, vertical O3 profiles in tropical tropopause layer (TTL) Diabatic velocity, vertical O3 profiles Diabatic velocity inferred from radiative calculation, ozonesondes Radiation Coordination: Piers Forster and Steven Pawson Solar UV-vis photolysis in stratosphere Radiative Transfer of 260-800 nm solar flux;
Photolysis rates comparison up to 95° solar zenith angle including clouds
Actinic flux (direct & scatter)
Photolysis rates of O3 and NO2 at local noon
Pressure, Ozone, stratospheric aerosols
Tropospheric clouds, aerosols and ozone
Direct flux measurements (balloon, ER-2)
Inferred photolysis rates (ER-2)
Heating rates Comparison of thermal and solar heating rates in offline runs employing column version of CCM radiation codes Heating rates and irradiances from CCM radiation code, with a prescribed and standardised set of input atmospheric profiles Use sophisticated reference radiation models for comparison (Line by line) NLTE, Discrete-Ordinate scattering etc. Radiative heating Global average of temperature profile Annually averaged global trace-gas and clouds fields, temperature Assimilated fields derived from satellite and sonde data,
Transient response of global average temperature Long-term globally averaged transient temperature changes Changes in Ozone, water vapour & high clouds, greenhouse gases, Hydrofluorocarbons, aerosols etc SSU/MSU satellite time series Stratospheric Chemistry & Microphysics Coordination: Martyn Chipperfield and Ross Salawitch Photochemical mechanisms and short timescale chemical processes Offline box model comparisons of fast chemistry (of order one week or less) Full chemical constituents
(O3 loss due to Ox, HOx, NOx, ClOx, BrOx, J values)
HOx: balloon, shuttle, A/C
NOx: satellite, shuttle, balloon, A/C
ClOx: satellite, shuttle, balloon, A/C
Long timescale chemical processes Comparison of abundance of reservoirs and radical precursors Instantaneous output of all chemical constituents and temperature
(one per month)
Satellite measurements of reservoirs and precursors Tracer-tracer relations O3, NOy, CH4, H2O, N2O Summer processes Ozone changes in polar regions Total ozone, full chemical constituents, temperature Satellite measurements of total ozone Ozone changes in mid-latitude regions Polar processes in winter /spring Partitioning of species within the families Species from families (ClOx, NOx, HOx, BrOx, Cly, NOy, BrOy) temperature, PV from wind fields Satellite and aircraft measurements Chemical Ozone Loss versus PSC activity O3, passive O3 tracer, O3 prod./loss rate, PV from wind fields, temperature Chemical ozone loss diagnosed from frequent ozone profiles in the vortex over several years
Met. Analyses **
Denitrification & Dehydration NOy vs. Tracer NOy, HNO3, N2O, CH4, etc Satellite measurements of HNO3, H2O, CH4
A/C obs. of NOy, H2O, CH4, N2O.
PSC size distributions
H2O +2 CH4 H2O particle-flux rates added to daily polar chem. Instantaneous output, CH4 Stratospheric Aerosols Sulfuric acid size distribution; aerosol optical extinction Sulfuric acid mass, particle number conc., water vapour, Temperature Satellite and in situ measurements of aerosols; aerosol climatologies Aerosols & Cloud Microphysics Cirrus cloud frequency of occurrence; H2O distribution Ice water content, water vapour, Temperature, aerosol size distribution Aircraft and satellite measurements; process/cloud-resolving models Advisory Group John Austin, David Fahey, Andrew Gettelman, Tatsuya Nagashima and Ben Santer * in addition to traditional model validation (climatological means, inter-annual variations)
** due to uncertainties use several analyses, not one
*** inter-comparison currently not possible because process not included in most CCMs
The stratosphere is strongly influenced by dynamical processes that CCMs must be able to reproduce correctly. Important examples are the forcing mechanisms and propagation of planetary-scale Rossby and (parameterized) gravity waves, wave-mean-flow interaction (transfer of energy and momentum), and the diabatic circulation. It is necessary that CCMs are not only able to simulate the climatological mean state of the stratosphere, including inter-hemispheric differences, inter-annual and intra-seasonal variability. As a first step it must be shown that the basic dynamical properties of the underlying GCMs on which the CCMs are based are reproduced. The analyses carried out during the first phase of the SPARC-GRIPS project (Pawson et al., 2000) provide a solid basis for the evaluation of CCMs. The analyses compared the vertical and latitudinal structures of the long-term zonal-mean temperature derived from observations and CCM simulations. Additionally, time series of monthly mean temperatures at distinct altitudes and latitudes help to identify and quantify overall model uncertainties.
Forcing and propagation of planetary waves. To determine the properties of the generation of planetary waves, their propagation through the stratosphere and their role in the momentum budget of the stratosphere, i.e. the stratospheric response to planetary wave drag (PWD), an analysis of stationary planetary wave patterns (up to zonal wavenumber 8) at different altitudes between the free troposphere and the upper model layers is required. This diagnostic can be augmented by calculations of Empirical Orthogonal Functions (EOFs) and of refractive index. Supplementary to the standard energy spectrum analysis, investigation of transient wave behaviour is necessary. Here, a Wavenumber-Frequency Analysis (WFA) can help to resolve transient waves at distinct wavenumbers into standing and eastward and westward travelling waves at different frequencies (Hayashi, 1982). The WFA can be performed by using power-, co-, and quadrature spectra of the time spectral analysis methods, such as the maximum entropy method, the direct Fourier transform method or the lag correlation method. An example is displayed in Figure 1. In order to determine the amplitudes and phases of the zonal quasi stationary planetary waves in the lower stratosphere, total ozone fields can be analysed by means of spectral statistical methods. Here, the total ozone column is considered as a conservative tracer to illuminate the variability of wave structures in the lower stratosphere. To derive the wave parameters from the ozone distribution the spectral statistical technique Harmonic Analysis can be applied to each latitude which corresponds to an approximate deconvolution of the power spectrum. The spectral properties can further be used to gain two hemispheric Ozone Variability Indices which are defined as the hemispheric mean of the zonal amplitude of the planetary waves number 1 and 2.
Figure 1. Wavenumber-frequency analysis. DJF transient wave variance per day at 300 hPa in gpm/d for wavenumber 1 as computed by the wavenumber-frequency analysis for westward (left) and eastward (right) travelling waves. The upper panel shows the 10-year mean of ECMWF ERA-15 Re-analysis variances (1984-93, Gibson et al., 1997), the lower panel shows the 20-year mean of the CCM E39/C timeslice simulation "1990" (Hein et al., 2001).
Stratospheric response to wave drag. Correlations of Eliassen-Palm fluxes (i.e., vertical and meridional heat and momentum fluxes) with dynamical and chemical fields (e.g., temperature, wind speed, ozone) and parameters (e.g., size and persistence of the polar vortex, PSC potential) are necessary to investigate the stratospheric response to wave drag and its consequences for chemical and physical processes in CCMs (Newman et al., 2001; Austin et al., 2003).
Moreover, a check of the ability of CCMs to reproduce correctly the seasonality of the Brewer-Dobson circulation is needed. This can be done by calculations of cross sections of the residual circulation mass streamfunction (latitude vs. height), which are based on re-analyses (e.g., NCEP, ERA-40) and corresponding results derived from CCMs. Derived diagnostic properties, such as the relative roles of PWD and gravity wave drag (GWD) in polar downwelling, and seasonally dependent changes of low frequency behaviour of stratospheric chemistry (e.g., ozone loss in spring, absorbing aerosols) in coupled vs. uncoupled models must be checked.
Quasi-Biennual Oscillation (QBO), Semi-Annual Oscillation (SAO). It is also important to validate the ability of CCMs to reproduce key oscillations in the stratosphere. One such oscillation is the semi-annual oscillation (SAO) of equatorial zonal winds at the stratopause. All CCMs simulate this to some extent, but the realism of the models SAOs varies considerably. CCMs are now just beginning to simulate the quasi-biennial oscillation (QBO), usually through the inclusion of enhanced GWD. It will be important to confirm that the models are obtaining a QBO for the right reasons, and that the extratropics responds in the correct manner.
Transport in the stratosphere involves both meridional overturning (the residual circulation) and mixing, which together represent the Brewer-Dobson circulation. The most important aspects are the vertical mean motion (diabatic velocity) and the horizontal mixing. The horizontal mixing is highly inhomogeneous, with transport barriers in the subtropics and at the edge of the wintertime polar vortex; mixing is most intense in the wintertime surf zone and is extremely weak in the summertime extratropics. Accurate representation of this structure in CCMs is important for the ozone distribution itself, as well as for the distribution of chemical families that affect ozone chemistry (NOy, Cly, H2O, CH4). Within both the tropics and the polar vortex, the key physical quantities to represent are the degree of isolation and the diabatic ascent or descent, respectively.
It is useful to distinguish between transport in the stratospheric overworld and in the UTLS. In the stratospheric overworld, there is a reasonably good understanding of the relevant processes and of how to quantify them. In contrast, the theoretical understanding of transport in the UTLS is relatively poor. This presents a challenge to determine appropriate diagnostics for model-measurement comparison.
Subtropical and polar mixing barriers. With respect to the degree of isolation, useful information can be obtained from instantaneous snapshots of tracer fields, which makes the model-measurement comparison straightforward. For this purpose there is a wealth of high-quality observational data available. A simple check on the degree of isolation is provided by the sharpness of latitudinal gradients of long-lived species (CH4, N2O, CFC11). However, since these gradients can be smeared out in zonal means, it is important to look at slices perpendicular to the mixing barrier (approximately, but not necessarily, at a single longitude). Equivalent latitude is an effective tool to create composites in the polar regions, but it is probably not viable in the tropics. A way to avoid latitudinal smearing without relying on equivalent latitude is to look at tracer Probability Distribution Functions (PDFs) (see Figure 2), which allow a direct model-measurement comparison. The degree of isolation can be diagnosed in more detail from the structure of chemical correlations, though their interpretation is not always straightforward. Within the very lowest part of the overworld in the tropics, just above the tropopause, where the tropical mixing barrier appears to be fairly leaky, horizontal transport into midlatitudes can also be quantified by the propagation of the annual cycle in CO2 and H2O, which has been well observed in aircraft measurements. Finally, transport out of the tropics can also be quantified in terms of streamers, but the quantification depends on how the streamers are defined.
Figure 2: Vortex isolation inferred from tracer probability distribution functions (PDFs). Time evolution of PDFs of CH4 distributions on the 450 K isentropic surface in the SH extratropics from September through November in both HALOE data (from 1992-1999) and from three consecutive years of simulation from the Goddard 3D CTM using winds from the Goddard Finite-Volume GCM (FVGCM). The latitude range of the vortex PDFs is 60-80S while the latitude range of the surf zone PDFs is 40-60S. Dashed lines reference the peak mixing ratio of the September PDFs in order to help judge which way the PDF is shifting as the vortex erodes. Courtesy of Susan Strahan, NASA Goddard Space Flight Center.
Diabatic ascent or descent has two aspects. First, a model must have the correct vertical residual velocity (or, equivalently, diabatic heating or cooling rates). This is controlled by the wave drag in the stratosphere and above. There are no direct measurements of these quantities, and, hence, they must be inferred from radiative calculations based on observed or assimilated temperatures and radiatively active species. This introduces some uncertainties in the comparison. The second aspect is the impact of the vertical residual motion on the actual vertical motion of chemical species. This depends on the degree of isolation. For example, if a model has spurious mixing across the vortex edge, then the descent of chemical species will reflect the diabatic descent in a broad region including the surf zone, rather than within the vortex. Assuming that the degree of isolation is correct, then it is possible to make a direct comparison between models and measurements by examining the ascent or descent rate of tracer isopleths. A well known example is the ascent rates of tropical H2O mixing ratios which create the tape recorder phenomenon in mixing ratio time series plots.
Meridional circulation. The combined effect of the above processes determines the Brewer-Dobson circulation. Both horizontal mixing and the residual circulation are driven in large measure by the momentum deposition (wave drag) from planetary waves propagating from the troposphere into the stratosphere, with more wave drag leading to a stronger Brewer-Dobson circulation in both respects. Because planetary waves can only propagate into the stratosphere when the winds are westerly, the Brewer-Dobson circulation is restricted to the winter hemisphere. The wave drag is easily quantified from the net planetary wave flux into the stratosphere, nominally taken to be v'T' (vertical EP flux) at 100 hPa. The relationship between this wave flux and the residual circulation is quantified, through temperature, in the dynamics diagnostics (Table 1). With regard to chemical transport, the seasonal cycle of O3 in the extratropics exhibits a marked build-up during the winter-spring period due to the Brewer-Dobson circulation. Years with greater planetary wave flux also have a greater ozone build-up, a relationship that is well established from observations and provides a good diagnostic for CCM validation.
The Brewer-Dobson circulation also determines the mean age of air. Unfortunately, the possibilities for direct comparison with data are more limited than for the processes described above, because the measurement precision requirements are so stringent that, at present, only in situ data can be used. This particularly limits comparisons in the upper stratosphere. Nevertheless, in NASAs Models and Measurements Intercomparison II, mean age of air was found to be a very powerful diagnostic for identifying model deficiencies. Mean age can be validated from measurements of long-lived species that have linearly increasing concentrations (e.g., SF6, CO2). Propagation of the annual cycle of mean age can be validated from CO2 measurements in the overworld (and H2O in the tropics). However other components of the age spectrum (e.g., semi-annual, biennial) are very difficult to validate.
UTLS transport. In contrast to the stratospheric overworld discussed above, transport in the UTLS region is far more complex. Yet many of the same concepts appear to be useful for validation. The extratropical tropopause is a barrier to quasi-horizontal mixing, causing a significant contrast in many chemical species between the lowermost stratosphere and the troposphere. The degree of isolation can be assessed by the sharpness of vertical gradients at the tropopause (vertical gradients because tropopause height changes with latitude), and with chemical correlations (e.g., O3 vs. CO). For the former there is plentiful ozonesonde data, and for the latter there is a wealth of aircraft data. These data are not sufficient to establish climatologies, but are nevertheless useful for process-based validation. However, it is important to compare models and measurements at similar longitudes, because there is significant longitudinal variation of the dynamical features in the UTLS (especially the tropopause). Unlike in the stratospheric overworld, UTLS transport is not quasi-zonal, and many chemical species are not sufficiently long-lived to be well-mixed longitudinally. There is a well-established relationship between variations in total O3 and in various tropospheric meteorological indicators, most notably tropopause height. While the precise mechanism for this relationship is not well understood - most likely, the various meteorological indicators are all just proxies for the same process - the relationship is robust and, therefore, also provides a potentially important diagnostic for CCM validation. Ozonesonde observations show that tropopause height variations affect O3 profiles through the depth of the lowermost stratosphere, up to about 20 km. The Tropical Tropopause Layer (TTL) is a critical part of the atmosphere in the UTLS to resolve properly in CCMs. Processes in this layer are important for setting chemical boundary conditions for the stratosphere and for understanding upper tropospheric chemistry and climate. The TTL region features large horizontal inhomogeneities, localised rapid vertical transport by convection, and many scales of dynamic variation due to waves. Many of these processes cannot be explicitly resolved by CCMs, but their effects must be treated reasonably to appropriately simulate the UTLS and to simulate climatic changes in the middle atmosphere. Validation can be accomplished by comparing the horizontal and vertical structure of the TTL to observations (e.g. the SHADOZ network of ozonesondes, GPS observations of temperatures and wave induced variability in the TTL or satellite observations from instruments, such as AIRS and MODIS).
The representation of the radiation field is a crucial aspect in CCMs if ozone abundances and temperature changes are to be accurately calculated in the present and future atmosphere. Radiation affects CCMs through photolysis rate and heating rate calculations. Chemically active constituents, such as ozone, are strongly affected by photolysis rates, which are derived from the radiation field. At the same time these trace gases feed back on temperature and, thus, circulation through the radiative heating rates. At present, most models calculate radiative heating rates and photolysis rates in an inconsistent manner. For example, the spherical geometry of the Earth might be included in the photolysis rate calculation, but not in the heating rate calculation. Also different radiation schemes are usually employed for the two calculations. Ideally, such inconsistencies would be avoided. However, here we evaluate these two calculations separately.
Solar UV-visible photolysis in the stratosphere. Photolysis rates in the stratosphere control the abundance of many chemical constituents that in turn control the production and loss of ozone. A photolysis rate generally requires knowledge of the actinic fluxes at solar and UV-visible wavelengths (190-800 nm) as a function of altitude and solar zenith angle. Accurate calculations of these fluxes require accurate representation of scattering, albedo and refraction. Particular concerns in photolysis rate calculations for the lower stratosphere are the effect of tropospheric cloudiness, which can significantly increase the rates for certain gases and photolysis at solar zenith angles greater than 90°. Diagnostic parameters for photolysis rates in CCM model comparisons include the radiative transfer of UV-visible wavelengths and calculated rates for individual gases. Key variables in such model comparisons are the distributions of pressure, ozone, stratospheric aerosols and tropospheric clouds. As a minimum test, the photolysis rates of O3 and NO2 should be stored as three-dimensional fields at local noon and compared to observations. In addition, actinic fluxes at the ground in different wavelength intervals should be compared.
Radiative heating rates. The radiative heating rate calculation is the fundamental link between ozone and climate. As this calculation plays the central part in CCM feedbacks, it is extremely difficult to separate cause and effect in a fully coupled model. Radiative heating rate calculations can only be truly evaluated in an offline comparison of radiation schemes. Currently, the lack of this comparison is one of the most important limitations in understanding CCM differences and we strongly advocate such a comparison be initiated. A set of standardised background atmospheres and radiation scheme inputs should be compiled, along with a reference set of calculations from several state of the art line-by-line and scattering (e.g., Discrete-Ordinate) models. These should then be made available to the community to evaluate their own CCM radiation scheme. Differences in radiative heating rates and trace gas fields can then be used to evaluate differences between the globally averaged climatological temperature of CCMs and their temperature response to changes in greenhouse gases loadings and other perturbations.
Radiative heating within an online framework. To evaluate radiative heating within an online framework the long-term global-mean temperature climatology of CCMs can be compared to observations (see Figure 3). An online framework allows a combined test of the models background atmosphere and radiative heating profile. Also, the globally averaged transient temperature changes over both a single year and the past ~25 years can be compared to Stratospheric Sounding Unit and Microwave Sounding Unit satellite observations. This tests both the evolution of forcing agents, as well as the radiative heating and the radiative relaxation time in the model.
Figure 3. Long-term global-mean temperature climatology. Vertical structure of the long-term, annual global-mean temperature (K) from observations (thick black line) and 13 models (thin coloured lines). Observations are a 17-yr-mean (Pawson et al., 2000).
Stratospheric Chemistry and Microphysics
Chemistry is clearly a natural process controlling the distribution of ozone in the atmosphere. Virtually all reaction rates are to a varying extent temperature dependent, providing one of the ways in which chemistry and dynamics are coupled. The importance of chemistry relative to other processes, such as transport, varies substantially depending on the local solar conditions as well as altitude. In the upper stratosphere transport plays a role by controlling the concentrations of the long-lived tracers such as active chlorine, but photochemical timescales are so short that transport has a minimal direct impact on ozone. However, in the lower stratosphere, the photochemical timescales are rather longer (typically of the order of months) and interactions with dynamics are complex and difficult to model accurately. Aerosols also may have an important role to play in the lower stratosphere since, in addition to their radiative impact, chemical reactions can take place within or on the particles and these reactions may lead to additional ozone depletion. Solar conditions are also important: for example, in polar night the distribution of chemical species is quite different to that in mid-latitudes where a clear diurnal variation in solar insolation occurs. Also, photochemical conditions are different in polar summer when the impact of the continuous daylight may be to photolyse the reservoir species entirely, depending on altitude. The different timescale of the processes in different parts of the atmosphere implies that a variety of modelling techniques can be effective.
Photochemical mechanisms and short timescale chemical processes. In the list of processes for stratospheric chemistry and microphysics, one of the most important tasks is to verify the performance of the underlying photochemical mechanisms, including the computation of photolysis rates. Model comparisons of this sort need to be completed using box model versions of the code used in the CCM, looking at timescales up to one week or so. Future studies can follow the example of the 'model and measurement tests' of Park et al. (1999). Very few measurements exist for direct comparison of photolysis rates (e.g., Gao et al., 2001), but there have been some attempts at inferring photolysis rates from chemical measurements. The comparisons could be made using the different model calculations for ozone loss and production in each of the catalytic cycles supported by Lagrangian studies using observations from a wide range of sources, both in situ and remote. Model diurnal variations could also be compared and verified with a limited range of observations.
Long timescale chemical processes. The investigation of long timescale photochemical processes needs to be completed within the CCM itself as tracer transport has a significant impact. All the model chemical constituents need to be output three-dimensionally, as well as the appropriate dynamical variables, such as temperatures. One instantaneous snapshot per month should be sufficient for the purpose of comparing the abundances of model reservoirs and precursors to the radicals, which directly affect ozone. The inter-relations between long-lived tracers also need to be compared in detail with similar results determined from space-based or in situ observations.
Summer processes and polar processes in winter/spring. In the summer, the polar regions are a special case of atmospheric chemistry because of the continuous or near continuous daylight. These conditions have revealed some possible discrepancies in NOx chemistry. This has an impact on ozone amounts directly in the polar regions and also in mid-latitudes via transport from the polar regions. In the winter/spring period, low temperatures lead to the formation of condensed matter and heterogeneous chemistry becomes important. Some aspects of heterogeneous chemistry can be investigated in box model simulations, but because of the possible importance of denitrification and dehydration, as well as transport, a full three-dimensional model is required for a complete analysis. Polar processes require an extensive set of chemical and particle concentration values within the polar regions with daily frequency. One particular diagnostic, designed to address overall model ozone depletion in polar regions, requires the addition of a passive tracer to the CCM. The tracer should be initialised on a specific date in the beginning of the winter identically to the ozone on that day. Thereafter, assuming that model transport errors are negligible, the difference between the photochemically computed ozone and the passive tracer provides an indication of the chemical ozone loss. Observations (Rex et al., 2003) indicate that chemical ozone loss and Polar Stratospheric Cloud (PSC) volume are linearly correlated (see Figure 4). Comparisons with this correlation would be a useful test of the ability of a model to simulate accurately the polar chemical ozone loss in the presence of PSCs.
Denitrification & Dehydration. Large polar ozone losses in both hemispheres occur in winters that are sufficiently cold for denitrification and dehydration. However, the current representations of these processes in CCMs are simplistic, leading to large uncertainties in polar ozone loss and in the impact on mid-latitudes. This is further complicated by (a) the poor understanding of the mechanism by which denitrification occurs and (b) CCM temperature biases in the polar vortex. The CCM representation of denitrification can be investigated by analysing the key nitrogen containing species, NOy and HNO3, as a function of the well-conserved tracers N2O and CH4. Remote and in situ data can be used to clarify these relationships and indicate any local loss in NOy or HNO3. Similarly, the sum H2O + 2 CH4 is approximately conserved in the stratosphere, so significant departures would indicate dehydration or possibly settling from above.
Figure 4. Polar chemical ozone loss. Variation of the overall chemical ozone loss in ten Arctic winters versus the winter-average of the volume of air sufficiently cold for PSC existence (VPSC). Measurements are shown by coloured squares. Black points are results from a chemical transport model. The slope of a fit through the points is a measure for the sensitivity of chemical ozone loss on changes in polar stratospheric temperatures and can be used to validate the representation of chemical ozone loss in CCM calculations (Rex et al., 2003).
Aerosol processes. Reactions involving sulphate aerosol are known to affect the production and loss balance of stratospheric ozone. Not all CCMs are in a position to investigate these processes in detail, as in some instances a complete sulphur reaction set is needed. Nonetheless, even for those models with a passive sulphate amount, it would be of interest to complete simulations describing the impact of a major volcanic eruption, such as that of Mt. Pinatubo.
Aerosols & Cloud Microphysics. Aerosol and cloud related processes affect the whole UTLS region. There is a need to investigate these processes in CCMs and validate them using the available satellite and aircraft data. The required model variables are liquid water and ice, temperature and aerosols, and they will be required at a relatively high spatial and temporal frequency, i.e., at least every three days and for every model grid point in the UTLS region. Further output of chemical constituents and potential vorticity would be useful to examine heterogeneous chemistry and the dynamical structure of the tropopause.
The Way Ahead
Of the comprehensive suite of diagnostics for stratospheric CCMs listed in Table 1, several have been applied before to a range of models (Austin et al., 2003; Pawson et al., 2000; Park et al., 1999), but many have not. Some models need further development before the diagnostics can be applied. Thus, while clearly desirable, it is a major task to perform all these diagnoses given the complexity of the CCMs and the often subtle changes under consideration. A step-wise approach is required for the use of the Table. In practice, modelling groups need to develop their own priorities among these diagnostics. The choices will depend on the known strengths and weaknesses of each model, the processes and constituents already included, and the existing output from runs already performed. It will also depend on the scientific focus of each modelling group and the issue being addressed. For example, predictions of polar ozone loss will have more credibility if a model has been shown to compare well with diagnostics such as ozone loss versus PSC volume, v'T', and ClOx, NOy, etc. In this case, good performance against TTL diagnostics is less relevant. Over time each model will gradually increase the number of tests applied and overall confidence will increase.
The lasting impact and the full benefit from the workshop will come from concerted validation activities based on the Table of Processes. In order for these activities to succeed over the next several years, broad support is needed from the atmospheric sciences community and its managers. It is important that the validation procedures and goals defined for these activities are accepted at the start and valued by all participants in this joint exercise.
SPARC working groups are being set up so that real progress can be made in the next couple of years in time for the next WMO/UNEP and IPCC assessments. The SPARC GRIPS group is continuing the work on the comparisons for the dynamics issues. SPARC groups have been formed on CCM chemistry and radiation comparisons and they are defining plans for their issues. Updated information is available at http://www.pa.op.dlr.de/workshops/ccm2003/ together with the names of people coordinating the various activities. All scientists interested in participating should contact the appropriate coordinating scientist.
To facilitate this process-oriented validation of CCMs, we intend to provide participants with access to diagnostic software packages. These routines will be archived in a central location. The goal in supplying such software is to simplify such activities as quality control of model output, calculation of more complex model diagnostics, statistical evaluation of model/data differences and graphical display of results. Use of this software is not mandatory. Rather, the intent is to make it easier for groups to compute a broad range of calculations in a reasonably consistent way. Centralised software repositories have been of great benefit in other Model Intercomparison Programs (MIPs), such as the Atmospheric Model Intercomparison Project (AMIP) and the Coupled Model Intercomparison Project (CMIP). These have freely supplied software for quality control of model output, data visualisation and interpolation of boundary condition datasets to a specific model grid. The CCM community can benefit from the experiences gained during previous model intercomparison exercises, particularly in terms of experimental design, definition of standard model output and statistical aspects of model-data comparisons. Software developed in the course of previous MIPs, such as performance portraits and Taylor diagrams, provide useful means of summarising many different aspects of climate model performance. In collaboration with groups, such as the Program for Climate Model Diagnosis and Intercomparison (PCMDI), we intend to modify these diagnostic tools in order to suit the specific needs of the CCM community.
This suite of processes and diagnostics should become a benchmark for validation. Confidence in the performance of CCMs will increase as more model attributes become validated against the whole suite of diagnostics. Further, new models can be evaluated against an acknowledged, benchmark set of diagnostics as the models are developed. At the same time, the diagnostics themselves should develop as experience is gained and as new measurements become available allowing more processes to be diagnosed. It is hoped that this workshop has laid the groundwork to a more comprehensive approach to CCM validation which will be developed by all scientists who become involved, irrespective of whether they attended the workshop or not.
We wish to thank all the agencies that supported this workshop. The workshop was held under the auspices of the Institute for Atmospheric Physics of the German Aerospace Center (DLR), the EU research cluster OCLI (Ozone CLimate Interactions), and SPARC.
Austin, J., et al. Uncertainties and assessments of chemistry-climate models of the stratosphere, Atmos. Chem. Phys., 3, 1-27, 2003.
Eyring V., et al. Brief report on the Workshop on Process-Oriented Validation of Coupled Chemistry-Climate Models, SPARC Newsletter N° 22, 2004.
Gao, R.S., et al. Observational evidence for the role of denitrification in Arctic stratospheric ozone loss, Geophys. Res. Let., 28, 15, 2879-2882, 2001
Gibson, J.K., et al. ERA description, ECMWF Re-Analysis Project Report Series, 1, 1-72, 1997.
Hayashi, Y., Space-time spectral analysis and its applications to atmospheric waves, J. Meteor. Soc. Japan , 60, 156-171, 1982.
Hein, R., et al. Results of an interactively coupled atmospheric chemistry-general circulation model: Comparison with observations, Ann. Geophysicae, 19, 435-457, 2001.
Newman, P.A., et al. What controls the temperature of the Arctic stratosphere during spring?, J. Geophys. Res., 106, 19999-20010, 2001.
Park, J.H., et al. Models and Measurements Intercomparison II, NASA/TM-1999-209554, 1999.
Pawson, S., et al. The GCM-Reality Intercomparison Project for SPARC: Scientific Issues and Initial Results, Bull. Am. Meteorol. Soc., 81, 781-796, 2000.
Rex, M., et al. Arctic ozone loss and climate change, Geophys. Res. Let., 31, doi:10.1029/2003GL018844, 2004.
WMO, Scientific Assessment of Ozone Depletion: 2002, Global Ozone Research and Monitoring Project - Report No. 47, 498 pp, Geneva, 2003.