assessment of coupled chemistry-climate models: evaluation ...dwaugh1/eyring_etal.pdf · 7...

EYRING ET AL.: CHEMISTRY-CLIMATE MODEL EVALUATION

1

Assessment of coupled chemistry-climate models: Evaluation of 1

dynamics, transport, and ozone 2

3

V. Eyring1, N. Butchart2, D. W. Waugh3, H. Akiyoshi4, J. Austin5, S. Bekki6, G. E. Bodeker7, B. A. 4

Boville8, C. Brühl9, M. P. Chipperfield10, E. Cordero11, M. Dameris1, M. Deushi12, V. E. Fioletov13, S. 5

M. Frith14, R. R. Garcia8, A. Gettelman8, M. A. Giorgetta15, V. Grewe1, L. Jourdain6, D. E. Kinnison8, 6

E. Mancini16, E. Manzini17, M. Marchand6, D. R. Marsh8, T. Nagashima4, E. Nielsen14, P. A. 7

Newman14, S. Pawson14, G. Pitari16, D. A. Plummer13, E. Rozanov18,19, M. Schraner19, T. G. 8

Shepherd20, K. Shibata12, R. S. Stolarski14, H. Struthers7, W. Tian10, and M. Yoshiki4 9 10 1 DLR, Institute of Atmospheric Physics, Oberpfaffenhofen, Germany. 11 2 Met Office Climate Research Division, Exeter, United Kingdom. 12 3 Johns Hopkins University, Baltimore, Maryland, USA. 13 4 National Institute for Environmental Studies, Tsukuba, Japan. 14 5 NOAA GFDL, Princeton, USA. 15 6 Service d'Aeronomie du CNRS, Paris, France. 16 7 National Institute of Water and Atmospheric Research, Lauder, New Zealand. 17 8 National Center for Atmospheric Research, Boulder, CO, USA. 18 9 Max Planck Institut für Chemie, Mainz, Germany. 19 10 Institute for Atmospheric Science, University of Leeds, United Kingdom. 20 11 San Jose State University, San Jose, USA. 21 12 Meteorological Research Institute, Tsukuba, Japan. 22 13 Environment Canada, Toronto, Ontario, Canada. 23 14 NASA Goddard Space Flight Center, Greenbelt, USA. 24 15 Max Planck Institut für Meteorologie, Hamburg, Germany. 25 16 Università L'Aquila, Dipartimento di Fisica, L'Aquila, Italy. 26 17 Istituto Nazionale di Geofisica e Vulcanologia, Bologna, Italy. 27 18 Institute for Atmospheric and Climate Science ETHZ and Physical-Meteorological Observatory. 28

Davos/World Radiation Center, Switzerland. 29 19 Institute for Atmospheric and Climate Science ETH, Zurich, Switzerland. 30 20 Department of Physics, University of Toronto, Canada. 31

32


2

Abstract. Simulations of the stratosphere from 1980 to 2000 from thirteen coupled chemistry-climate 1

models (CCMs) are compared with each other and to observations. Comparisons of the temperature 2

fields show that the models reproduce the global, annual mean temperature fairly well, but most CCMs 3

still have a cold bias in winter-spring polar regions in the southern hemisphere. Most display the 4

correct stratospheric response to wave forcing in the northern, but not in the southern hemisphere. 5

Comparisons of simulations of methane, mean age of air, and the so-called water vapor “tape recorder,” 6

show a wide spread in the results, indicating differences in transport. However, for around half the 7

models there is reasonable agreement with observations. In these models the mean age and tape 8

recorder are generally better than reported in previous model comparisons. Comparisons of the water 9

vapor and inorganic chlorine (Cly) fields also show a large inter-model spread. Differences in water 10

vapor are primarily related to biases in the simulated tropical tropopause temperatures, and not 11

transport. The spread in Cly, which is largest in the polar lower stratosphere, appears to be primarily 12

related to transport differences. In general the amplitude and phase of the annual cycle in total ozone is 13

well simulated apart from the Antarctic, where there are significant differences in the Antarctic ozone 14

hole among the models and in comparison with observations. CCMs show a large range of ozone 15

trends over the past 25 years and large differences in comparisons with observations. Global 16

temperatures trends are in reasonable agreement with satellite and radiosonde observations and 17

compared to the CCM results presented in this study most models that neglect changes in stratospheric 18

ozone generally show significantly less cooling at 50 hPa between 1980 and 2000. 19

20


3

1. Introduction 1

2

The future evolution of the stratosphere is of particular interest for the timing of ozone recovery and 3

chemistry-climate interactions. An important tool for understanding past changes and predicting the 4

future evolution of the stratosphere is the coupled chemistry-climate model (CCM). CCMs include full 5

representations of dynamical, radiative, and chemical processes in the atmosphere and their 6

interactions. In particular, CCMs include feedbacks of the chemical tendencies on the dynamics, and 7

hence the transport of chemicals. This feedback is a major difference between CCMs and chemical 8

transport models, and is required to simulate the evolution of ozone in an atmosphere with increasing 9

greenhouse gases. Ozone is a major radiative agent in the stratosphere, and is also strongly affected by 10

dynamics and transport, so the ozone radiative-dynamical feedback in CCMs is of particular 11

importance for the representation of chemistry-climate coupling [WMO, 2003]. 12

13

CCMs have been used to simulate the evolution of the stratosphere in the 21st century, and to predict 14

the recovery of the ozone layer [e.g. Austin et al., 2003; Austin and Wilson, 2006; Dameris et al., 15

2006]. To interpret and assess these predictions the capabilities and limitations of the models first need 16

to be determined by comparing their simulated meteorology and trace gas distributions with 17

meteorological analyses and observations. Such comparisons not only point to model deficiencies and 18

uncertainties, but can also help in understanding processes, mechanisms and feedbacks within the 19

atmosphere and identifying deficiencies in our understanding. It is also of interest to compare models 20

with each other, to determine the consistency between the models and to examine the sensitivity of 21

errors to model formulation and experimental set-up. 22

23

An initial inter-comparison and assessment of multiple CCMs was performed by Austin et al. [2003]. 24

Since this initial CCM assessment a number of new CCMs with a focus on the middle atmosphere have 25

been developed, and changes have also been made to the CCMs that were included in Austin et al. 26

[2003]. Some of the new and updated CCMs have been tested and compared to observations [e.g. 27

Austin et al., 2006; Dameris et al., 2005; Egorova et al., 2005; Garcia et al., 2006; Manzini et al., 28

2003; Pitari et al., 2002; Steil et al., 2003; Steinbrecht et al., 2006a,b; Struthers et al., 2004; Tian and 29

Chipperfield, 2005] but there is need for a new multi-model assessment. 30

31


4

The aim of this study is to assess dynamics, transport, and simulated ozone distributions in the current 1

generation of CCMs. This will include an update of the previous multi-CCM assessment [Austin et al., 2

2003] and a previous assessment of the middle atmospheric general circulation models that generally 3

did not have a coupling between simulated trace gases and radiative calculations [Pawson et al., 2000]. 4

However, in contrast to these previous studies, the CCM simulations used here have near-identical 5

forcings (e.g., sea surface temperatures, greenhouse gases, and halocarbons). This eliminates many of 6

the uncertainties in the conclusions of earlier assessments that resulted from the differences in 7

experimental set up of individual models. The simulations generally follow the specifications of the 8

past reference simulation of the Chemistry-Climate Model Validation Activity for SPARC (CCMVal) 9

[Eyring et al., 2005a]. Furthermore, this study is the first multi-model assessment of stratospheric 10

transport characteristics of CCMs. As the common simulations include relevant natural and 11

anthropogenic forcings (e.g., increasing greenhouse gases, major volcanic eruptions, and solar cycle) 12

we also evaluate how well the models simulate past temperature and total ozone time series. In a 13

subsequent study the same models will be used to predict the evolution on the stratosphere in the 21st 14

century. 15

16

The models and simulations as well as the observational data sets that are used in this study to evaluate 17

the CCMs are described in section 2. In section 3 the simulated stratospheric basic state in the CCMs is 18

evaluated by comparing model temperatures and stratospheric response to wave forcing with similar 19

diagnostics calculated from meteorological analyses. Section 4 focuses on the evaluation of transport 20

characteristics and long-lived tracers, by comparing methane, water vapor and mean age distributions 21

to observations. Inaccuracies in dynamics and transport affect modeled inorganic chlorine as well as 22

ozone distributions, which are discussed in sections 5 and 6. Section 7 focuses on past variability and 23

trends in temperature and total ozone time series and a concluding summary is presented in section 8. 24

25

2. Models and Observational Data Sets 26

27

2.1. Model Descriptions 28

29

Thirteen coupled chemistry-climate models have participated in this model inter-comparison. Their 30

main features are summarized in Table 1 and detailed descriptions of the CCMs are given in the cited 31

literature. 32


5

1

Some of the CCMs can be grouped together as they have been built on the same underlying GCM, 2

although important differences exist between the models in each group (see Table 1). For example, 3

UMETRAC and UMSLIMCAT are both based on the U.K. Unified Model (UM) [Pope et al., 2000]. 4

MAECHAM4CHEM and SOCOL are based on the same middle atmosphere GCM [Manzini et al., 5

1997], but each model used a different forcing for the Hines [1997] non-orographic gravity wave drag 6

(NonO-GWD) scheme [see Egorova et al., 2005]. The E39C CCM can also be grouped with 7

MAECHAM4CHEM and SOCOL, because their underlying GCMs are all derived from ECHAM4 8

[Roeckner et al., 1996]. In addition, the chemistry schemes of E39C and MAECHAM4CHEM are both 9

based on Steil et al. [1998], while those of UMETRAC and AMTRAC are based on the chemistry 10

scheme described in Austin et al. [2006]. Other groups of models are not based on the same GCM but 11

do share some similar characteristics in their dynamics and transport. For example, AMTRAC, 12

GEOSCCM, and WACCM all use finite-volume dynamical and advection schemes and GEOSCCM 13

and WACCM have similar physical parameterizations. However, in all cases there are either 14

completely separate components in the related CCMs (e.g. the base GCM is the same, but the 15

chemistry differs) or significant modifications have been made to the originally common components. 16

17

There are large variations in the characteristics of the CCMs. The horizontal resolution varies from as 18

low as 10° x 20° (ULAQ) to 2° x 2.5° (AMTRAC, GEOSCCM), and the location of the upper 19

boundary from being centered at 10 hPa (E39C) to 10-6 hPa (WACCM). Also, the number of vertical 20

levels ranges from 26 (ULAQ) to 71 (CMAM). The numerical methods used for the dynamical core 21

and advection schemes also vary (see Table 1). The dynamics in all CCMs is determined by solving the 22

‘primitive’ equations, except for ULAQ which is a quasi-geostrophic model [Pitari et al., 2002]. 23

24

A major issue with GCMs of the middle atmosphere is the treatment of gravity waves. Given the 25

limited spatial resolution of these models, gravity waves need to be parameterized, and this 26

parameterization can vary significantly among models. Orographic gravity wave drag schemes are 27

employed in all models except ULAQ. To represent the effects of gravity waves of non-orographic 28

origin, parameterizations of gravity wave spectra are used in all the other models (see Table 1), but not 29

in E39C (the upper boundary is located below the region where these effects are important) and ULAQ 30

(Rayleigh friction). Currently a limitation of most of the gravity wave schemes is that their source 31


6

spectrum is generally specified externally and specified sources do not evolve in time with a changing 1

climate. 2

3

All CCMs have a comprehensive range of chemical reactions and include a number of chemical species 4

from the Ox, HOx, ClOx, BrOx, and NOx families (apart from E39C and MAECHAM4CHEM, which do 5

not include bromine species) and source gases. All models include both gas-phase chemistry and 6

heterogeneous chemistry on liquid and solid aerosols, and on polar stratospheric clouds (PSCs), but 7

different PSC and denitrification schemes are used. The formation of NAT particles and 8

dehydration/denitrification by gravitational settling of PSC particles is not included in CMAM and 9

UMSLIMCAT. 10

11

As all models considered are coupled climate-chemistry models they all have coupling between the 12

chemical composition, the radiation, and the dynamics. However, the CCMs differ in their treatment of 13

this coupling. All models use the simulated O3 and H2O fields in their radiation calculations, but the 14

number of other simulated greenhouse gases used varies (see Table 1). 15

16

In contrast to the other models, the UMSLIMCAT model carries two separate water vapor fields, a 17

tropospheric humidity field that is used for tropospheric physics but is not coupled with the chemistry, 18

and a stratospheric H2O tracer used in the stratospheric chemistry module. This second water vapor 19

field has a constant value of 3.3 ppmv in the troposphere and a chemical source in the stratosphere 20

from CH4 oxidation. In all other models there is a single water vapor field that is used in the 21

tropospheric physics and stratospheric chemistry. 22

23

A sub-set of CCMs (CCSRNIES, CMAM, E39C, MAECHAM4CHEM, ULAQ, and UMETRAC) have 24

already contributed to an earlier assessment of CCMs [Austin et al., 2003]. However, all of these 25

models have been modified compared to the last assessment. The horizontal resolution in the 26

CCSRNIES CCM has been increased from T21 to T42, and the non-orographic gravity wave drag 27

parameterization of Hines [1997] has been included. The chemistry scheme has also been updated and 28

now includes bromine chemistry and the methane oxidation process [Akiyoshi et al., 2004], and 29

heterogeneous chemistry has been updated with the chemistry scheme of the SLIMCAT CTM [Sessler 30

et al., 1996]. The plane-parallel radiative transfer scheme has been modified with a pseudo-spherical 31

approximation that is capable of photolysis calculations at solar zenith angles larger than 90o 32


7

[Kurokawa et al., 2005]. The number of vertical levels in CMAM has been increased from 65 to 71 to 1

improve resolution through the tropopause region, and the parameterization of both orographic and 2

non-orographic gravity wave drag has also been changed (see Table 1). The underlying GCM has not 3

changed in E39C, MAECHAM4CHEM and UMETRAC since Austin et al. [2003] but changes have 4

been made to the chemistry schemes: Two more heterogeneous reactions on stratospheric aerosol have 5

been added to the chemistry module in E39C and MAECHAM4CHEM and photolysis for twilight 6

conditions with a solar zenith angle up to 93° is included [Steil et al., 2003]. The UMETRAC CCM 7

now uses the same chemistry scheme as in AMTRAC [Austin et al., 2006]. The surface drag and 8

vertical diffusion coefficients have been decreased in the ULAQ model and cirrus ice particles have 9

been included in the upper troposphere. 10

11

2.2. Model Simulations 12

13

The model simulations considered here are transient simulations of the last decades of the 20th century 14

which, in general, follow the specifications for “reference simulation 1” (REF 1) of CCMVal [Eyring et 15

al., 2005a]. The REF1 simulations include anthropogenic and natural forcings based on changes in 16

sea-surface temperatures (SSTs), trace gases, solar variability, and sulfate aerosols (from volcanic 17

eruptions). Full details of the REF1 specifications are summarized in the work of Eyring et al. [2005a] 18

and only the key aspects are given here. All forcings for the simulations can be downloaded from the 19

CCMVal Project website (http://www.pa.op.dlr.de/CCMVal/). 20

21

In REF1 simulations: 22

23

• The SSTs are prescribed as monthly means following the global sea ice and sea surface 24

temperature (HadISST1) data set provided by the UK Met Office Hadley Centre [Rayner et al., 25

2003]. This data set is based on blended satellite and in situ observations. 26

• The surface concentrations of well-mixed greenhouse gases (WMGHG) are based on IPCC [2001]. 27

• The surface halogens are based on Table 4B-2 of WMO [2003] and are extended through 2004 (S. 28

Montzka, pers. communication). 29

• Chemical kinetics is taken from JPL 2005 (updated from JPL [2003]). 30


8

• The influence of the 11-year solar cycle on photolysis and heating rates is parameterized according 1

to the intensity of the 10.7 cm radiation of the sun [Lean et al., 1997]. 2

• Both chemical and direct radiative effects of enhanced stratospheric aerosol abundance from large 3

volcanic eruptions are considered (see below). 4

• A quasi-biennial oscillation (QBO) is either internally generated or forced (see below). 5

6

The three major volcanic eruptions (Agung, 1963; El Chichón, 1982; Pinatubo, 1991) are taken into 7

account In REF1 simulations by prescribing observed sulfate aerosol densities and additional modeled 8

heating rates (see Table 2). Sulfate surface area densities (SADs) are specified from a monthly 9

climatology based on satellite data, similar to that used in the work of Jackman et al. [1996] and 10

updated by D. B. Considine (NASA Langley Research Center). The heating rates are monthly means 11

from January 1950 to December 1999 for all-sky condition, and were calculated using volcanic aerosol 12

parameters from Sato et al. [1993] and Hansen et al. [2002] and GISS ModelE radiative routines and 13

climatology [Schmidt et al., 2006] (G. Stenchikov and L. Oman, pers. communication). 14

15

The QBO is included in one of two ways in REF1 simulations. Three of the models (MRI, UMETRAC, 16

and UMSLIMCAT) are able to internally generate a QBO driven by resolved and parameterized wave 17

mean-flow interaction [Scaife et al., 2000; Shibata and Deushi, 2005]. The QBO in these models is an 18

internal mode and hence is not synchronized with the observed QBO. The other CCMs do not generate 19

an in-situ QBO, and simulate permanent tropical easterlies instead of a QBO. In these models the QBO 20

is forced externally by relaxation ('nudging') of the stratospheric equatorial zonal wind to observed 21

equatorial zonal wind profiles [Giorgetta and Bengtsson, 1999]. The externally driven QBO is 22

therefore synchronized with the observed QBO. 23

24

Not all of the model simulations examined here followed the REF1 specifications exactly. The SSTs 25

used are all based on observations, but in the AMTRAC and WACCM models the Hurrell et al. [2005] 26

dataset was used while the AMIP II SSTs [Taylor et al., 2000; Gates et al., 1999] were used in the 27

LMDZrepro and UMSLIMCAT models. Furthermore, not all models included a QBO or changes in 28

solar forcings and sulfate aerosols, and the method for accounting for the radiative effects of volcanic 29

aerosols varied and is not included in all simulations. The forcings employed in each of the 13 transient 30

CCM simulations used in this inter-comparison are summarized in Table 2. 31


9

1

2.3. Observational Data Sets 2

3

The main observational data set used to validate the meteorological fields in this study is the UK Met 4

Office stratospheric analyses [Swinbank and O’Neill, 1994]. Production of these assimilated analyses 5

started in October 1991, and therefore in order to produce a ten year climatology for the 1990s we use 6

the assimilated fields from January 1992 to December 2001. For consistency we use the same 10 years 7

of data from the NCEP/NCAR [Kalnay et al., 1996] and ERA-40 [Uppala et al., 2005] re-8

analyses. Because many of the CCM simulations stopped at the end of December 1999 the 9

climatologies for the models are based on the 10 years from January 1990 to December 1999. This 10

slight discrepancy in the dates does not affect any of the conclusions presented in this paper. 11

12

To evaluate transport as well as hydrogen chloride and ozone distributions, we compare modeled tracer 13

distributions with measurements of the Halogen Occultation Experiment (HALOE) on board the Upper 14

Atmosphere Research Satellite. HALOE has made measurements of methane, water vapor, hydrogen 15

chloride, and ozone since October 1991 [Russell et al., 1993]. Here we compare model climatologies of 16

the 1990s to a climatology derived from 1991 to 2002 HALOE data, where data since September 2002 17

have not been included because of the unusual major warming in the Antarctic in 2002 and because the 18

observations have been less frequent after 2002 [Grooß and Russell III, 2005]. This climatology 19

provides a consistent data set and is easy accessible. Uncertainties of single profile HALOE 20

measurements have been estimated in several studies [e.g., Brühl et al., 1996; Harries et al., 1996; 21

Park et al., 1996; Russell et al., 1996]. For all measured species the accuracy of the HALOE 22

observations decreases near the tropopause. In addition, sparse coverage of the polar regions increases 23

the uncertainty in the HALOE climatologies there. In all inter-comparisons the HALOE climatological 24

mean and the inter-annual standard deviation (1σ) are shown. 25

26

We examine the simulated tracer distributions in polar spring (80oN in March, and 80oS in 27

October/November). As no long-term (1990-1999) climatology of satellite-based measurements is 28

available at 80oN in March, model profiles are compared with HALOE measurements at 80oN 29

equivalent latitude. Ideally, model profiles averaged by equivalent latitude should be compared to the 30

HALOE profiles, but these are not available. The HALOE profiles at 80oN equivalent latitude are 31


10

indicative of the expected shape of the profile and should not be quantitatively compared with the 1

model profiles. 2

3

Not enough observations are available to form a global climatology for the mean age of air and 4

inorganic chlorine (Cly). However, there are more limited data that can be used for assessing the 5

simulations. These data are described in the relevant sections below. 6

7

Past trends of modeled total column ozone are compared against four observational data sets: ground-8

based, merged satellite data, NIWA and Solar Backscatter Ultraviolet (SBUV, SBUV/2). The ground-9

based zonal mean data set used in this study includes long-term measurements of Dobson, Brewer and 10

filter ozonometer and is updated from Fioletov et al. [2002]. Data are available from the World Ozone 11

and Ultraviolet Radiation Data Centre at http://www.woudc.org. The merged data set are monthly-12

mean zonal and gridded average data sets constructed by merging individual TOMS and SBUV/2 13

satellite data sets (see http://hyperion.gsfc.nasa.gov/Data_services/merged/mod_data). All data in the 14

present version of the merged data set have been derived using the TOMS Version 8 and SBUV 15

Version 8 algorithms. The NIWA data set is the same as that described by Bodeker et al. [2005] except 16

that daily total column ozone fields from the Ozone Monitoring Instrument (OMI) are used, when they 17

are available, in preference to the Earth Probe TOMS data. Differences between OMI overpass 18

measurements and ground-based Dobson spectrophotometer measurements are small (-2.8+-5DU) and 19

are corrected before the data are combined with the other data sources. The SBUV data set is based on 20

zonally averaged SBUV and SBUV/2 data (version 8) from the period from 1979 to 2004 (updated 21

from Miller et al., [2002]) . The SBUV(/2) data are available at 22

http://www.cpc.ncep.noaa.gov/products/stratosphere/sbuv2to. 23

24

Modeled Temperature trends are compared with ERA-40 reanalysis data and radiosonde data derived 25

from the Radiosonde Atmospheric Temperature Products for Assessing Climate (RATPAC) 26

climatology [Free et al., 2005]. 27

28

3. Stratospheric Climate 29

30

Stratospheric dynamics is determined by a number of processes. These include the forcing mechanisms 31

and propagation of planetary-scale Rossby and gravity waves, wave-mean-flow interaction, and the 32


11

diabatic circulation. We assess the ability of the CCMs to model these processes by examining the 1

simulated stratospheric basic state and the stratospheric response to wave forcing in the CCMs. 2

3

3.1. Temperatures 4

5

The global and annual mean stratosphere is fairly close to radiative equilibrium. Significant biases in 6

the modeled global mean temperatures can therefore be related to systematic errors, e.g., in the 7

radiative heating rates. Figure 1 shows global mean temperatures for the 1990s from the 13 8

participating CCMs compared to three different data sets: UKMO, ERA-40 and NCEP/NCAR re-9

analyses for 1992 to 2001. In the lower stratosphere where there is good agreement among the three 10

observational data sets, a large subset of models (9 out of 13) have relatively small biases and are, on 11

average, about 1 K too warm between 100 hPa and 10 hPa. AMTRAC and WACCM have slightly 12

larger biases whereas CCSRNIES and UMETRAC have problems reproducing the observed global 13

mean temperatures over the entire model domain. These two models both have significant cold biases 14

around 100 hPa. In the case of UMETRAC the bias extends well into the stratosphere. The reasons for 15

this peculiar low temperature problem are unclear but it is worth noting that the underlying GCM for 16

UMETRAC does not have the same problem. UMSLIMCAT, which is based on the same GCM as 17

UMETRAC does not have this problem either. Therefore, it is likely that the bias in UMETRAC is 18

related to the coupling of the chemistry scheme to the GCM. Improvements in the long-wave radiation 19

code have reduced the bias in the CCSRNIES model [M. Sekiguchi and S. Watanabe, private 20

communication, 2006], but these improvements are not included in the run used here. 21

22

It is important to account for uncertainty in meteorological analyses when making comparisons, 23

especially above 10hPa when the only constraint on the analyses comes from deep-layer radiances and 24

the characteristics of the underlying model may dominate the analysis. Also, radiosondes are not 25

available to provide bias correction for the satellite measurements. Above 10 hPa the differences in the 26

three assimilation data sets are of the same magnitude as differences between a subset of 11 out of 13 27

CCMs. Most models have a negative temperature bias of about 3K compared to UKMO analyses but 28

compare well with ERA-40. In contrast, CMAM is warmer than observed in the upper stratosphere, 29

though the reasons for this are not currently understood. The positive bias in CMAM temperatures 30

occurs despite also underestimating the concentration of ozone in this region (see section 6). Low water 31

vapor compared to observations (see section 4.2) could explain part of the warm bias through 32


12

underestimated long-wave cooling. Note that unlike most models CMAM includes near-infrared CO2 1

solar heating, which warms the upper stratosphere by 1-1.5 K [Fomichev et al., 2004]. 2

3

Overall the comparison shown in Figure 1 represents a significant improvement on the persistent 4

stratospheric problem of negative global, annual mean temperature biases found in most of the middle 5

atmosphere GCMs evaluated by Pawson et al. [2000, Figure 1]. This is due in part to improvements in 6

the radiation schemes and, in the upper troposphere, in dynamics and/or vertical resolution, but could 7

also be related to chemistry (see e.g. discussion on LMDZrepro and UMETRAC below). In the upper 8

stratosphere, using interactive rather than prescribed ozone as in many of the Pawson et al. [2000] 9

models, has a self-correcting effect on temperature, because of the negative ozone-temperature 10

feedback [de Grandpré et al., 2000]. Note also that some of the poorer performing models of Pawson 11

et al. [2000] are no longer included in the comparison and have been replaced by new and more 12

sophisticated models. In addition, some of the models have not been developed to include chemistry so 13

are not included in this comparison. 14

15

An accurate simulation of high latitude temperatures in winter and spring is important for correctly 16

modeling polar ozone depletion. The mean winter and spring time temperature biases poleward of 60º 17

are shown in Figure 2. Below 10 hPa the meteorological analyses from UKMO, ERA-40, and NCEP 18

agree very well and, except for UMETRAC and CCSRNIES which have both cold biases in particular 19

in winter, the majority of the models have rather small biases (1-2 K) in the northern hemisphere 20

between 100 and 10 hPa especially in spring. This contrasts with Austin et al. [2003, Figure 3] who 21

found that the CCMs they considered mostly had a warm bias in the lower stratosphere in the boreal 22

winter, though they only considered a single latitude (80ºN). Also in the high latitude northern winter 23

the differences between Figure 2a and 2b and Austin et al.’s results could quite easily be within the 24

range of uncertainty due to decadal variability [Butchart et al., 2001]. Although this can be mitigated 25

for in the model results by using multi-model means, the observational records for the stratosphere are 26

generally not long enough to eliminate uncertainty due to decadal variability in particular in the 27

northern hemisphere. For CMAM, the elimination of a significant warm bias in the Arctic is attributed 28

to a reduction in the critical inverse Froude number used to define wave breaking in the orographic 29

gravity wave drag parameterization (following independent work by S. Webster and B. Boville, private 30

communication). 31

32


13

In the upper stratospheric winter and spring there are differences of up to 9 K between the assimilation 1

data sets and most models are within this range. In the southern hemisphere the CCSRNIES, E39C, and 2

MAECHAM4CHEM models have a cold bias compared to UKMO of up to -20 K in winter, whereas 3

LMDZrepro shows a significant warm bias of up to 20 K in winter / spring that is partly related to the 4

pronounced cold bias in the lower stratosphere. In the southern hemisphere winter (June-July-August) 5

the lower stratosphere (100 to 30 hPa) biases vary between ±5 K with almost zero bias in the multi-6

model mean compared to UKMO. However, the models tend to be clustered into 2 groups with either a 7

cold bias of around 5 K or a warm bias. Low temperature biases in the lower southern hemispheric 8

stratosphere in spring are most pronounced in LMDZrepro and E39C (around 15 K), but also apparent 9

in most of the other models. This “cold-pole” problem is a long standing problem of stratospheric 10

GCMs and CCMs that can be attributed in part to missing wave drag [Garcia and Boville, 1994]. 11

Potential improvements in the “cold-pole” problem in CCMs might for example be related to 12

improvements in non-orographic gravity wave drag schemes [e.g. Austin et al., 2003] and an improved 13

simulation of the stratospheric water vapor distribution [e.g. Stenke and Grewe, 2006]. 14

15

Based on the mean seasonal cycle in the monthly and zonally averaged westerly winds at 60°S the 16

transition from westerlies to easterlies occurs much too late in nearly all of the models (Figure 3). In 17

some models (E39C, GEOSCCM, LMDZrepro, and WACCM) stratospheric westerlies persist into 18

January. The poor representation of this aspect of the seasonal cycle appears to be a persistent feature 19

of most of the models with important implications for tracking the seasonal longevity of Antarctic 20

ozone depletion. In particular the late break down of the polar vortex means that the Antarctic ozone 21

hole persists longer than in reality in most models (see section 6). 22

23

3.2. Stratospheric Response to Planetary Wave Forcing 24

25

The vertical propagation of planetary waves from the troposphere into the stratosphere plays a 26

significant role in determining the temperatures of the polar stratosphere during winter and spring and 27

their inter-annual temperature variability. These waves deposit momentum in the stratosphere and 28

mesosphere that drives a poleward and downward circulation, which adiabatically warms the polar 29

region. The wave forcing can be quantified in terms of the vertical component of the Eliassen-Palm 30

(EP) flux entering the stratosphere at 100 hPa, which is proportional to the meridional heat flux (v'T') at 31

100 hPa. Using NCEP analyses Newman et al. [2001] showed that the interannual variations in the 32


14

meridional heat flux at 100 hPa correlated well with 50 hPa polar cap average temperatures, which is 1

consistent with our theoretical understanding of the adiabatic heating processes occurring in the 2

stratosphere. The correlation calculated from ERA40 re-analyses gives almost identical linear 3

relationships, with some scatter among individual years (not shown). Austin et al. [2003] demonstrated 4

that this process-oriented diagnostic was useful for assessing models’ representation of planetary wave 5

forcing of polar stratospheric temperatures. 6

7

Figure 4 shows scatter plots of polar cap temperatures at 50 hPa versus heat fluxes at 100 hPa for 8

northern (upper panels) and southern (lower panels) hemisphere. The displacement of the model results 9

above or below the observations is a reflection of the temperature bias (see Figure 2). In the northern 10

hemisphere the models generally reproduce the slope of the linear fit between temperature and heat 11

flux, though in some models (especially ULAQ and WACCM) the correlation between the variables is 12

lower than in the NCEP analyses (see Table 3). The slope of the linear fit is quite accurate in most of 13

the models which suggests that they should be able to predict the correct temperature response to a 14

change in the wave forcing. The MRI model has a slightly steeper slope indicating greater sensitivity to 15

wave forcing, which may account for the slightly greater range and variability of the temperatures in 16

that model (Table 3 and Figure 4). However, the MRI model also has the largest inter-annual variability 17

in the heat flux and the highest correlation between the heat fluxes and temperatures (again see Table 18

3). On the other hand the model with the lowest correlation (WACCM) is the least responsive to 19

changes in wave forcing (upper right panel in Figure 4). The low correlation between heat flux and 20

temperature would suggest that other processes, for example the gravity wave drag, are contributing to 21

the temperature variability in WACCM, but this would not explain the very low correlation in the 22

ULAQ model which has no gravity wave forcing. For the other models the inter-annual variability is 23

consistent with the NCEP analyses. Compared to Austin et al. [2003], results from two of the models 24

(CCSRNIES and ULAQ) show significant changes. Higher horizontal resolution (T42 rather than T21) 25

in the CCSRNIES model has led to, on average, larger and more realistic heat fluxes but note also that 26

some of this increase can be attributed to the use of daily fields, rather than 10 day averages used in the 27

previous assessment. In ULAQ the decrease of eddy friction in the troposphere could be the main 28

reason for the improvement in the heat flux versus temperature diagnostic; previously, the slope in the 29

northern hemisphere was much too shallow, indicating too weak a sensitivity to the wave forcing. 30

31


15

In the southern hemisphere (lower panels in Figure 4) there is much greater variation among models, 1

and on average the models are in poorer agreement with the NCEP analyses than in the north. Relative 2

to the NCEP analyses several of the models (e.g. CCSRNIES, E39C, MAECHAM4CHEM, MRI, 3

LMDZrepro, and ULAQ) have temperatures that are rather unresponsive to changes in the heat fluxes 4

(i.e., the slope of the linear fit is much flatter than observed; see Table 3), with the ULAQ model even 5

having the incorrect sign. These same six models also have the lowest actual correlations between heat 6

fluxes and temperatures (see Table 3) indicating problems in their dynamical response to changes in the 7

stratospheric meteorology of the southern hemisphere in winter. The LMDZrepro model has both lower 8

magnitude heat fluxes and slightly lower temperatures than the NCEP analyses, while the AMTRAC 9

model has higher magnitude heat fluxes and higher temperatures. In general, in most of the models but 10

not in AMTRAC or LMDZrepro, the mean heat flux or wave activity entering the stratosphere appears 11

to be broadly correct (see Table 3) and therefore the deficiencies in the heat-flux temperature 12

relationship are probably due more to weaknesses in the representation of the stratosphere per se. As in 13

Austin et al. [2003] it is not possible to identify any direct link between the deficiencies in the southern 14

hemisphere and model resolution. 15

16

4. Stratospheric Transport Characteristics 17

18

Transport in the stratosphere involves both meridional overturning (the residual circulation) and 19

mixing. Horizontal mixing is highly inhomogeneous, with transport barriers in the subtropics and at the 20

edge of the wintertime polar vortices [e.g. Sankey and Shepherd, 2003]. Possible errors in transport 21

characteristics have important impacts on the distribution of chemical families and individual species 22

that affect ozone chemistry (NOy, Cly, H2O, and CH4) and consequently the ozone distribution itself. 23

24

Useful information on transport characteristics can be obtained from examining the distribution of 25

long-lived tracers, such as methane and nitrous oxide. Also, information on the tropical ascent and 26

tropical-extratropical mixing can be obtained from the vertical propagation of the annual cycle in water 27

vapor (“the tape recorder”), while the mean age of air provides information on integrated transport 28

within the stratosphere. 29

30

4.1. Methane 31

32


16

As all CCMs simulate methane and there is a long observational record from HALOE we focus on this 1

tracer. The concentration of methane entering the stratosphere is largely controlled by surface 2

emissions and hydroxyl-initiated oxidation in the troposphere. Once methane enters the stratosphere 3

the concentration is controlled by methane oxidation (primarily in the middle and upper stratosphere) 4

and transport. In the simulations the surface concentration of CH4 is prescribed with the same time 5

varying values and it is expected that the oxidation in the stratosphere is very similar in all the models, 6

so differences in the CH4 distribution can be attributed to differences in transport. 7

8

Figure 5 (a-c) shows vertical profiles at different latitudes and (d,e) latitudinal variation at 50 hPa of 9

monthly-mean zonal mean CH4 from the CCMs and HALOE. There are large variations in the CH4 10

from different CCMs, particularly in polar regions. For example, the simulated mean CH4 in the polar 11

lower stratosphere (50 hPa) in the southern hemisphere varies from around 0.6 ppmv to around 1.4 12

ppmv. These large variations in CH4 indicate large differences in the transport. However, in most 13

regions, the large range of model values is caused by a few CCMs (discussed below), and for a large 14

subset of the models (8 out of the 13 models) the spread in CH4 is much smaller and the simulated 15

mean CH4 is, in general, within 1σ of the HALOE mean. 16

17

The tropical CH4 mixing ratios in the lower stratosphere (50 hPa) from 4 CCMs (MAECHAM4CHEM, 18

MRI, SOCOL, and ULAQ) are much smaller (around 0.2 ppm) than those observed or calculated by 19

other models. The low bias in these models persists throughout the tropical stratosphere (Figure 5b), 20

except for SOCOL where CH4 in the upper stratosphere lies within 1σ of the HALOE mean. 21

MAECHAM4CHEM, MRI and ULAQ are also generally lower than other models in extra-tropical 22

(Figure 5d,e). The difference is particularly large in MAECHAM4CHEM as a consequence of the low 23

tropical value but similar than observed strong meridional tropical-extratropical gradients. On the other 24

hand SOCOL and in particular MRI and ULAQ have much weaker meridional gradients, and the low 25

bias is therefore smaller in the extratropics. 26

27

In contrast to the above models, the CH4 distribution from E39C is biased high, except in the tropics 28

(Figure 5d,e). The tropical distribution from E39C is close to the HALOE observations, but the 29

calculated values in the extratropics of both hemispheres are much higher than observed, and there are 30

weak meridional gradients. The E39C-HALOE differences are particularly large in the polar middle 31

stratosphere. 32


17

1

The reasons for the differences in simulated latitudinal and vertical CH4 profiles cannot be easily 2

generalized. Differences in the meridional overturning, latitudinal mixing and vertical diffusivity, as 3

well as differences in chemistry, could all play a role. For some models there are some indications of 4

the cause of the differences. For E39C the most likely cause of the high CH4 mixing ratios in the 5

extratropics is the low upper boundary centered at 10 hPa. Although the downward transport at high 6

latitudes is fairly well captured, the poleward transport in the model near the upper boundary is too 7

strong, which leads to too high CH4 values in the extratropics [Grewe, 2006]. Also, the sink above 10 8

hPa is missing. The CH4 distribution in MAECHAM4CHEM shows realistic horizontal structure but is 9

shifted to lower values, resulting in a negative bias everywhere in the lower stratosphere. This is due to 10

excessive transport between the troposphere and stratosphere, as discussed in detail in Steil et al. 11

[2003]. In MRI the strong vertical gradient in CH4 is due in part to smaller mass flux across the 12

tropical tropopause: the values of annual-mean and seasonal cycle are about 70% and 50% of those 13

estimated from ERA40. Biases in other models are smaller and explanations do not appear to be due to 14

a single, dominant model deficiency. 15

16

As mentioned above, if the five models that have large biases with respect to HALOE are not 17

considered, there is a much smaller spread between the other 8 models and these models are in 18

reasonable agreement with the observations, i.e. within 1σ of the HALOE mean. The one possible 19

exception is in polar regions in spring where there is larger variation between the 8 CCMs, and all are 20

higher than HALOE observations. However, there is large interannual, including decadal, variability in 21

polar transport (and tracers), and this could contribute to some of the model spread (see earlier 22

discussion with regards to polar temperatures). Also, there are limited HALOE observations in polar 23

regions, and the use of equivalent latitude for the HALOE data will necessarily lead to smaller CH4 24

values compared to the zonal means from the models. In addition, at high latitudes the use of 25

equivalent latitude for the HALOE data will necessarily lead to smaller CH4 values compared to the 26

zonal means from the models. This happens because a displaced vortex will increase the zonal-mean 27

values as a result of averaging in-vortex and out-of-vortex air, whereas equivalent latitude will tend to 28

preserve the character of in-vortex air. 29

30

4.2. Water Vapor 31 32


18

Future changes in stratospheric water vapor could have a major impact on stratospheric ozone, both 1

through changes in the radiative balance, changes in HOx, and through the formation of PSCs. It is 2

therefore important to assess how well the CCMs simulate the observed water vapor distribution. 3

Furthermore, the vertical propagation of the annual cycle in tropical water vapor (the water vapor “tape 4

recorder” [Mote et al., 1996]) can be used to assess the ascent rate and the mixing with mid-latitudes in 5

models. HALOE water vapor observations [Mote et al., 1996; Randel et al., 2004; Grooß & Russell III, 6

2005] have been used in previous model-data comparisons, including the NASA Models and 7

Measurements II ("MMII") study [Hall et al., 1999; Park et al., 1999] and NASA GMI model 8

evaluation [Douglass et al., 1999]. 9

10

Figure 6 shows latitudinal and vertical profiles of zonal mean water vapor from the CCMs and 11

HALOE, for the same regions and seasons as shown for CH4 in Figure 5. There is an extremely large 12

spread in the simulated water vapor fields and, unlike the case for CH4, this is not due to only a small 13

sub-set of the models. The water vapor at 50 hPa in the models varies from around 2 ppmv in some 14

models to over 6 ppmv in others. Most models have a minimum water vapor concentration 15

('hygropause' [Kley et al., 1979]) near the tropical cold point (see Figure 6b). However, there is a large 16

spread in the value of the hygropause, with the minimum tropical mean March water vapor 17

concentration varying between 1.5 and 4.5 ppmv. The observed minimum from HALOE is less than 3 18

ppmv and the uncertainty in the absolute value of the HALOE measurements is around 10% [SPARC, 19

2000]. 20

21

In general the water vapor value is a function of the model cold point temperatures near the tropical 22

tropopause, and model-model variations in tropical tropopause temperature can explain much of the 23

variation in lower stratospheric water vapor. This is illustrated in Figure 7 which shows the annual 24

cycle of equatorial (a) temperature and (b) water vapor at 100 hPa. Note that the cold point is not 25

necessarily at 100 hPa, and a better comparison may be between values of temperature and water vapor 26

at the cold point. However, for simplicity, and consistency with Pawson et al. [2000], we show the 100 27

hPa values. Most models reproduce the phase of the seasonal cycle of temperature and water vapor 28

reasonably well, but there are variations in amplitudes of the cycles and the annual-mean values. 29

However, in comparison with the models examined by Pawson et al. [2000], the annual cycle of 30

tropical temperature at 100 hPa has improved markedly and now the models nearly all have the correct 31

annual cycle, and the spread of values is only 10 K rather than 20 K. The differences in the annual-32


19

mean temperature at 100 hPa explain most of the variations in water vapor, and models with higher 1

temperatures (E39C, MAECHAM4CHEM, MRI, and ULAQ) generally have higher water vapor, while 2

those models with lower temperatures (UMETRAC, AMTRAC, and CCSRNIES) generally have lower 3

water vapor mixing ratios. 4

5

Dehydration can also cause differences in the simulated water vapor in polar regions (Figure 6a,c). 6

There is a wide spread in simulated stratospheric water vapor in polar lower stratosphere in spring of 7

both hemispheres (with concentrations varying between 1 and 7 ppmv), which is linked to the large 8

differences in polar temperatures (see Figure 2). The cause of the large values in southern polar regions 9

in October in the UMETRAC model is not known, and is currently under investigation. Note that 10

CMAM and UMSLIMCAT do not simulate the process of dehydration in these runs. 11

12

Reproducing the correct amplitude of the stratospheric water vapor 'tape recorder' is dependent on 13

getting the annual cycle of tropical cold point temperatures reasonably correct (Figure 7). As discussed 14

above, most models reproduce the observed phase of the water vapor cycle near the tropical 15

tropopause, but there are variations in amplitude and annual-mean value. These variations propagate 16

vertically into the tropical stratosphere. The water vapor ‘tape recorder’ is illustrated for the CCMs and 17

HALOE in Figure 8 which shows the time-height sections of water vapor mixing ratio calculated as the 18

deviation (in parts per million by volume) from the time-mean profile, averaged between 10°S and 19

10°N. It should be noted that the UMSLIMCAT chemical water vapor field in the simulation used here 20

is not coupled to the dynamical core of the model, which is why this model does not show a tape 21

recorder signal (not shown). In addition to the model-model and model-data differences in the 22

amplitude around 100 hPa there are also differences in the attenuation of the amplitude and the speed 23

of propagation (“phase propagation”) of the water vapor signal. These differences are due to 24

differences in tropical transport (ascent rates, mixing with mid-latitudes and vertical diffusion), and 25

comparison of these quantities provides information on the transport differences between the models 26

(e.g., Hall et al. [1999]). The vertical variation of the normalized amplitude and phase propagation of 27

the water vapor signals in the models and in the HALOE observations is shown in Figure 9. The phase 28

propagation of the water vapor tape recorder signal (Figure 9c) is calculated as the average propagation 29

of the maximum and minimum phases of the signal, and the phase lag is taken zero at 16 km. The 30

phase propagation in GEOSCCM is in good agreement with the HALOE observations, but the 31

propagation is too fast in the remaining models. For several models the propagation time to 26 km is 32


20

only a few months less than HALOE, but for half the models the propagation time is around or less 1

than half that of HALOE (CCSRNIES, E39C, MAECHAM4CHEM, MRI, SOCOL and UMETRAC). 2

Interestingly, differences from the observed CH4 distribution were also found in most of the models 3

with rapid phase propagation (e.g., E39C, MAECHAM4CHEM, MRI, and SOCOL). Differences from 4

observations in both CH4 and propagation of the tape recorder indicate deficiencies in the transport in 5

these models. However, it is not exactly clear which aspects of the transport are causing the differences 6

in the tracer fields, see further discussion below (section 4.4). Too rapid propagation of the phase has 7

been also reported in the models that have been assessed in the MMII study [Hall et al., 1999], where 8

nearly all the models had a propagation time to 26 km of less than 9 months. 9

10

To focus on the decay of the amplitude with height due to mixing rather than temperature-induced 11

differences in the amplitude near 100 hPa, the amplitudes shown in Figure 9a,b are normalized to unity 12

at a reference level. In Figure 9a, as in previous comparisons [e.g., Hall et al., 1999; Douglass et al., 13

1999] this reference level is 16 km. In most models the amplitude decays with height above this level, 14

but in some models the amplitude increases with height or is nearly constant between 16 and 18 km. 15

The most notable increases are in WACCM and UMETRAC, see Figure 8 where clear maximum and 16

minimum values can be seen above 16 km. These variations in the bottom 1 or 2 km are related to the 17

height of the cold point in these models, i.e., the WACCM and UMETRAC cold points are above 16 18

km. Because of this we also compare the amplitude normalized to unity at the level where the 19

maximum value occurs (between 16 and 20 km) and examine the attenuation with height above the 20

level where this maximum occurs (Figure 9b). This figure shows that the decay of amplitude with 21

height (attenuation) in most models is in reasonable agreement with HALOE. There is too rapid decay 22

in the lower stratosphere of the LMDZrepro model and in the middle stratosphere (above 20 km) of the 23

MRI, and too little decay in the CMAM and CCSRNIES models. However, in the other models the 24

agreement with HALOE is good. 25

26

The results shown here are in much better agreement with HALOE than the majority of the models in 27

the MMII study [Hall et al., 1999] where nearly all the models over-attenuated the signal. Note that 28

although the attenuation with height is reasonable in most models, the attenuation per wavelength (as 29

considered by Hall et al. [1999]) will be too strong in many of these models because of the too rapid 30

propagation of the signal. 31

32


21

4.3. Mean Age of Air 1

2

The mean age of air is defined as the mean time that a stratospheric air mass has been out of contact 3

with the well-mixed troposphere. This diagnostic can be inferred from observations of conserved 4

tracers with an approximately linear increase in concentration with time (e.g., CO2, SF6, total chlorine), 5

and has been used in previous model-data comparisons [e.g., Hall et al., 1999; Park et al., 2000]. These 6

previous comparisons showed that most models, independent of whether 2 or 3-dimensional, had mean 7

ages that were significantly lower than observations. The cause of the young bias in the models 8

considered in these studies could not in general be attributed to any single aspect of the models. 9

10

Figure 10 shows the latitudinal variation of the annual-mean mean age of air at (a) 0.5 hPa, (b) 10 hPa, 11

and (c) 50 hPa from several CCMs, together with observations (symbols). Not all CCMs included a 12

tracer that enables the mean age to be calculated, and the "age tracers" also varied between the CCMs 13

(e.g., idealized linearly increasing tracer, CO2 tracer, total chlorine, or tracer with internal source). In 14

some cases two age tracers were available, and produced almost identical results. The age of air 15

observations in Figure 10 are based on (a) ER-2 aircraft measurements of CO2 from many different 16

years [Andrews et al., 2001], (b) much more limited balloon CO2 measurements made in northern mid-17

latitudes [see Table 1 of the work by Waugh and Hall, 2002; also Aoki et al., 2003], and (c) satellite 18

measurements of HF and HCl from HALOE [Anderson et al., 2000]. Note that some age of air 19

estimates derived from SF6 observations indicate ages significantly older than 6 years, but these are 20

now thought to be biased high due to mesospheric loss processes (see Waugh and Hall [2002] and 21

references therein). 22

23

There is very good agreement between CCMs and observations at 50 hPa, with the exception of MRI 24

where the age of air is 1-2 years older than observed at all latitudes, MAECHAM4CHEM where it is 25

1-2 years older at high latitudes, and UMETRAC where it is 0.5 years younger at all latitudes. The 26

agreement is not as good for higher altitudes, where most models have younger age of air than 27

observed, and MRI and MAECHAM4CHEM have older age of air (the MRI age at 0.5 hPa is greater 28

than 8 years). The one exception is the ULAQ model where the age of air agrees with the observations 29

at 0.5 hPa, and is only slightly larger at 10 hPa. 30

31


22

As with the tape recorder comparisons, the model-observation agreement in Figure 10 is much better 1

than in the NASA MMII multi-model assessments [e.g., Hall et al., 1999]. In the Hall et al. [1999] 2

study the mean ages from all but one of the over 20 2D and 3D models examined were younger than 3

observed at 50 hPa (and most were significantly younger), and all but three had mean ages less than 4 4

years at 10 hPa in northern mid-latitudes. So, although the mean age of air in the CCMs assessed here 5

is generally younger than observed in the middle and upper stratosphere, the differences are generally 6

not as large as previously reported. 7 8 4.4. Discussion 9 10 Some general conclusions on the transport in the CCMs can be drawn from the above comparisons of 11

methane, mean age, and water vapor tape recorder with observations. 12

13

In around half of the models the simulated tracer fields are in reasonable agreement with the 14

observations, for all or the majority of the diagnostics. Furthermore, for these models the agreement for 15

mean age and the tape recorder is much better than it was in nearly all models reported in the MM2 16

study [Hall et al., 1999]. This indicates that the transport in these models is reasonable, and some of 17

the previously reported transport problems may not apply to these models. Note that the transport 18

analysis has focused on the tropics and middle latitudes, and some of these models still may have major 19

problems in polar transport. 20

21

On the other hand, for several models (E39C, MAECHAM4CHEM, MRI, and SOCOL) there are large 22

differences from observations in several (if not all) of the transport diagnostics considered here. This 23

indicates that there are problems with the transport in these models (particularly in the tropics). 24

However, the cause of the problems in these models is unclear and needs further examination. 25

26

The major issue with identifying the problems is that tracer distributions depend on balance of 27

advective and diffusive processes and it is generally not straightforward to link a difference in a tracer 28

field to a particular transport process. As an illustration consider the tropical transport in 29

MAECHAM4CHEM, MRI, and SOCOL. The too rapid phase propagation of the tape recorder (Figure 30

9c) in these models suggests too rapid vertical motion (the phase is in general only weakly influenced 31

by tropical-mid-latitude mixing [Hall et al., 1999]). However, the vertical gradients of CH4 (Figure 5b) 32

tend to be too strong in particular in MRI and MAECHAM4CHEM, but also in SOCOL, which is the 33


23

opposite of what would be expected for too rapid ascent. The vertical gradient of the age of air in MRI 1

and MAECHAM4CHEM is too large, which is also not a symptom of too rapid ascent. Therefore, 2

other transport processes must be playing a role. One possibility is that the errors in the tropical CH4 3

and mean age of air are due to overly strong tropical-extratropical mixing (this mixing has a larger 4

impact on these tracers than the phase of the tape recorder). While this could be the case in some 5

models (e.g., MRI which has weak subtropical gradients of CH4 and rapid attenuation of the tape 6

recorder above 20 km) it does not appear to be the case for MAECHAM4CHEM, which has strong 7

subtropical gradients of CH4 similar to what is observed and reasonable attenuation of the tape recorder 8

amplitude (although the attenuation per wavelength is too strong). Thus, although we have identified 9

problems in the transport in these models in most cases more detailed analysis of the models and output 10

fields is needed to identify the cause. 11

12

5. Hydrogen Chloride (HCl) and Inorganic chlorine (Cly) 13

14

The accumulation of halogenated compounds in the stratosphere over the past 40 years has been the 15

primary driver of stratospheric ozone depletion. As stratospheric chlorine and bromine levels decline as 16

a result of reductions in their source gas emissions, ozone is expected to recover. The ability of CCMs 17

to reproduce past stratospheric chlorine concentrations clearly affects the confidence we can place in 18

their predictions of future ozone changes, mainly in the Antarctic. It is therefore important to assess 19

how realistic the simulated inorganic chlorine (Cly) is in the CCMs with a particular focus on polar 20

spring Cly in the southern hemisphere. There are only limited observational data available for Cly (i.e. 21

not enough to form a climatology). We therefore compare simulated hydrogen chloride (HCl) with 22

HALOE measurements. HCl is one of the principal components of Cly, and overall the model-model 23

differences in HCl are similar to those in Cly (not shown). There are however large differences in polar 24

regions in October and March which are related to a different timing of the chlorine deactivation in the 25

models leading to large differences in HCl/Cly ratios. For example, at 50 hPa the HCl/Cly ratio varies 26

between 0.0 and 0.8 in the polar southern hemisphere in October and between 0.2 and 0.8 in the polar 27

northern hemisphere (not shown). These differences are much smaller in November and April, where 28

the HCl/Cly ratio ranges only between 0.7 and 0.95 in the southern hemisphere and between 0.3 and 0.8 29

in the northern hemisphere. We therefore show climatological Cly and HCl data in November and 30

April, rather than March and October as for other species in previous figures. The range in simulated 31

HCl/Cly ratio is small in the tropics (between 0.7 and 0.9) in all months. 32


24

1

Figure 11 shows the climatological-mean latitudinal and vertical profiles of HCl from the CCMs and 2

HALOE. As with the other trace gases considered above there is a large spread between the models. 3

All the CCMs used the same, or very similar, halogen boundary conditions so the differences are not 4

likely due to differences in tropospheric concentrations. A more likely cause are differences in 5

transport, and, consistent with this, the CCMs that have problems simulating the CH4 distribution also 6

generally have problems simulating HCl. For example, in the tropical lower stratosphere HCl in E39C 7

is biased low compared to HALOE while HCl in MRI, MAECHAM4CHEM and ULAQ is biased high 8

(Figure 11b,d,e). 9

10

However, transport differences cannot explain all of the model-model differences, and other factors 11

play a role. For example, several models carry only a few chlorine source gases and scale them to 12

represent all other components, which could introduce errors in the rate of conversion to Cly. Another 13

example of non-transport related differences are the high values of HCl in the AMTRAC and to a 14

smaller extent in the UMETRAC simulations compared to observations and all other models at 50 hPa 15

(Figure 11d,e). Simulated AMTRAC and UMETRAC Cly is also higher compared to other models (not 16

shown). The CH4 (Figure 5) and mean age (Figure 10) in these models is similar to that in several other 17

models and is also similar to observations, so transport is unlikely to cause the higher values of HCl 18

and Cly. In these two CCMs the photolysis rates of organic chlorine species are adjusted to account for 19

the low bias in the simulated mean age so that the Cly is in closer agreement with the observed Cly. 20

Hence these models have higher Cly and HCl than models with similar mean ages. However, these 21

models also have very high HCl relative to HALOE throughout the lower stratosphere. 22

23

There are some more interesting features in the vertical profiles of HCl and Cly in polar spring (Figure 24

11a,c and 12a). The extent to which high Cly and HCl values reach down to the lower altitudes reflects 25

the degree of unmixed descent from the upper stratosphere within the polar vortex. In AMTRAC the 26

high Cly and HCl values reach deeper into the lower stratosphere than in all other models, most likely 27

due to the chemical correction for transport anomalies. E39C Cly (HCl) profiles are generally shifted 28

upward which leads to too low Cly (HCl) mixing ratios at nearly all levels. A first test with a fully 29

Lagrangian advection scheme instead of the semi-Lagrangian scheme employed indicates significant 30

improvements with regard to the structure of the profiles [Stenke and Grewe, 2006]. SOCOL has 31

lowest Cly values from all models at 50 hPa in polar southern hemisphere in spring. 32


25

1

The vertical profile of Cly also indicates some serious problems in the upper stratosphere for some 2

models. Two models (CCSRNIES and SOCOL) produce values of Cly in the upper stratosphere that are 3

larger than the maximum total chlorine entering the stratosphere, which is physically impossible unless 4

there is a chlorine source in the stratosphere. In the SOCOL model there is a maximum in Cly (and total 5

chlorine) in the tropical stratosphere around 10 hPa (not shown) which might be due to the global mass 6

correction used in the semi-Lagrangian part of the transport scheme which can lead to local errors. At 7

the other extreme the MRI model simulates unrealistically low values of Cly in the upper stratosphere. 8

These low values are consistent with low CH4 (Figure 5) and very old mean age (Figure 10) in this 9

model, as an old mean age would delay the buildup of chlorine in the upper stratosphere, but even ages 10

older than 8 years cannot explain the low Cly in the upper stratosphere of the MRI model. Analysis of 11

total chlorine (CCly + Cly) from this model (not shown) indicates that the total chlorine in the upper 12

stratosphere is not only shifted in time from the troposphere but also increases at a much slower rate, so 13

that the time lag from the troposphere increases with time. However, the reasons for the very slow 14

increase in Cly and total chlorine have not yet been identified. 15

16

The above comparisons considered climatological means (1990-1999) when there were large trends in 17

Cly, and it is of interest to compare the simulated time evolution of Cly. Figure 12b shows the evolution 18

of Cly in the SH polar lower stratosphere from the different CCMs in October. All models show Cly 19

increasing rapidly during the 1980s and early 1990s, however, consistent with the above analysis, there 20

are large variations in the simulated peak Cly (from around 1.2 ppb to over 3.5 ppb). Note that the 21

UMSLIMCAT run had not fully spun-up by 1980 and did not capture the correct stratospheric lag 22

compared to the troposphere until around 1985. Estimates of Cly from measurements in southern polar 23

lower stratosphere in spring can be made from measurements of HCl in this region, as Cly ~ HCl when 24

there are very low values of O3 [e.g., Douglass et al., 1995]. Measurements of HCl by UARS HALOE 25

in 1992 yield a value around 3 ppb [Douglass et al., 1995; Santee et al., 1996], whereas measurements 26

by Aura MLS in 2005 yield a mean value around 3.3 ppb [M. Santee, pers. communication], see 27

symbols in Figure 12b. There are reasonably large uncertainties (10-15%) in these values because of 28

the limited coverage of polar region (for HALOE) and possible biases in the HCl measurements 29

[Froidevaux et al., 2005]. However, even with this uncertainty it is clear that peak Cly values around or 30

less than 2.5 ppb, as simulated in several CCMs (especially SOCOL and E39C) are too low. 31

32


26

6. Ozone 1

2

Figure 13 compares vertical profiles and latitudinal cross sections of ozone derived from the models 3

and HALOE. In the tropical lower stratosphere between 100 and 30 hPa a subset of 9 out of 13 CCMs 4

agree well with HALOE observations and lie within 1σ of the HALOE mean. MAECHAM4CHEM, 5

MRI and SOCOL are biased high, which is consistent with low CH4 (see Figure 5a-c and discussion in 6

section 4.1). At higher altitudes (above 7 hPa) these four models lie within the 1σ HALOE mean, 7

whereas peak values of ozone in the tropics are biased high in WACCM and CCSRNIES. The tropical 8

stratosphere in WACCM is dynamically isolated and the mean age of air is too young near the ozone 9

volume mixing ratio maximum (youngest age of all models at 10 hPa, see Figure 10) leading to NOx 10

mixing ratios that are too low by approximately 10 to 20%. In this region, the photochemical lifetime 11

of ozone is between 5 and 10 days [Solomon et al., 1985] and the primary odd-oxygen catalytic loss 12

cycle is NO2 + O, consistent with the positive bias. CCSRNIES overestimates peak values of ozone not 13

only in the tropics but also in both polar regions, likely due to an overestimation of O2 photolysis rates 14

at this altitude. 15

16

At 80°N in March (Figure 13a) the CCM profiles are compared to HALOE profiles at the same 17

equivalent latitude, which are indicative of the expected shape of the ozone profile. Despite the 18

uncertainty introduced by using equivalent latitude for HALOE, it seems that at lower altitudes (below 19

20 hPa) most models are biased high, consistent with the biases seen in total ozone climatologies (for 20

fixed northern hemispheric polar latitudes), which are shown in Figure 14. The latitudinal profile at 50 21

hPa in March (Figure 13d) shows that the high bias is a feature of high latitudes only. Most of the 22

models are close to observations equatorward of 60°N. The profiles derived from all models indicate 23

that the altitude of maximum ozone mixing ratio is generally below the observed peak level, i.e. 7 to 10 24

hPa in the models instead of 3 hPa in the HALOE data. However, as with all Arctic springtime 25

diagnostics, one needs to be aware of the possible sampling uncertainty due to variability, with only a 26

10-year data set. Another uncertainty lies in the sampling of HALOE observations at polar latitudes 27

[Grooß and Russell III, 2005]. 28

29

In the southern hemispheric polar spring at 80°S (Figure 13c) most models are biased high in ozone 30

compared to HALOE observations below 50 hPa and are biased low between 50 and 20 hPa. The most 31

pronounced ozone bias is with LMDZrepro, which simulates very low ozone mixing ratios at 80°S in 32


27

October even though Cly, HCl, CH4, and H2O are not anomalous. This bias originates from the 20 K 1

bias in temperature. The LMDZrepro temperatures in October and November stay well below the 2

threshold temperature for PSC formation in the simulation whereas in reality the temperature should 3

increase above the PSC formation temperature. Interestingly, this large negative ozone bias reinforces 4

the temperature bias, as the temperature bias is less pronounced in the pure GCM simulation [Lott et 5

al., 2005]. The two models that show lowest Cly values (E39C and SOCOL) in the lower stratosphere 6

(Figures 11 and 12) are also the ones that simulate highest ozone values in the southern hemisphere 7

polar regions in October at 50 hPa (Figure 13e). 8

9

Figure 14 compares the 10-year mean climatological total column ozone for the 1990s from models, 10

merged satellite data and NIWA assimilated data. Detrended mean annual cycles for the 1980s are also 11

shown for the polar regions and globally in Figure 15. The well-known features of highest ozone values 12

in northern spring, low ozone values in the tropics with a small seasonal cycle, a relative ozone 13

maximum in the mid-latitudes of the southern hemisphere in late winter / early spring, and a minimum 14

ozone column above the Antarctic are represented in all models. With the exception of the southern 15

hemisphere high latitudes, the amplitude and phase of the annual cycle in individual latitude bands is 16

generally well simulated by the models, reflecting the strong linkage between the springtime maximum 17

and the summertime minimum [Fioletov and Shepherd, 2003; Weber et al., 2003]. However, most 18

models exhibit significant annual-mean biases which depend on latitude. The northern hemisphere high 19

latitude spring values are too high in a number of models (E39C, GEOSCCM, LMDZrepro, 20

MAECHAM4CHEM, MRI, and WACCM) and in these models the mid-latitude southern hemisphere 21

maxima in late winter / early spring are also too high. For some of these models this may be an 22

indication of a generally too strong Brewer-Dobson circulation (see also the meridional ozone gradients 23

in Figure 13d,e), although it should be noted that some of these models exhibit a high bias in global-24

mean total ozone (see Figure 15, right panels), especially MAECHAM4CHEM and MRI. Moreover, 25

the high southern hemisphere mid-latitude maximum is also consistent with the general southern 26

hemisphere cold pole bias, which implies an overly strong transport barrier at the edge of the vortex. In 27

contrast to many models, SOCOL and ULAQ have relatively low southern hemisphere mid-latitude 28

maxima, which are likely associated with too rapid horizontal mixing with the tropics (in particular for 29

ULAQ), as reflected in the weak latitudinal gradients of ozone and other tracers (see section 4). 30

31


28

There are also significant differences in the evolution and the size of the Antarctic ozone hole between 1

the models themselves and in comparison with observations. Several ozone depletion indices have been 2

used in the past to assess the severity of polar ozone depletion [e.g., Austin et al., 2003; Bodeker et al. 3

2005; WMO, 2003], and these indices have been calculated from the CCMs here. In particular, the 4

minimum total ozone column poleward of 40°S, the maximum Antarctic ozone hole area (OHA) for 5

the period 1 September to 30 November defined by the 220 DU contour, and the ozone mass deficit 6

calculated as the mean of daily values between 1 September and 30 November based on a 220 DU 7

threshold (OMD) have been calculated for each year. Table 4 compares the 1995 to 1999 average of 8

these ozone depletion indices. A 5-year period is used as there are large trends in the southern 9

hemisphere indices from 1980 to 1995. 10

11

The modeled minimum Antarctic ozone values are in good agreement with observations and lie within 12

20 DU of the observed minimum. However, in most models the simulated ozone hole is too small, both 13

in terms of ozone hole area and ozone mass deficit, and the date of maximum ozone hole area occurs 14

later than observed by 10 to 20 (up to around 40) days (not shown). The later occurrence of ozone hole 15

maximum area is likely related to the cold bias and late vortex breakup (see section 3 and Figure 3). 16

The reasons for the smaller ozone hole are in general not known, but it is important to note that indices 17

calculated from a fixed threshold (e.g. 220 DU) do not account for the general high bias in the total 18

ozone fields that occurs in several models (see Figure 14 and right panels in Figure 15). This high bias 19

however often originates from layers that are not interesting for chemical ozone depletion and in most 20

models this is a global and not purely southern hemispheric bias. For example, the seasonal cycle in the 21

Antarctic could be well represented in a model indicating a correct simulation of the processes of ozone 22

depletion, whereas the indices could be underestimated due to a persistent global high bias (e.g. in 23

MAECHAM4CHEM). This points to a weakness in the definition of the ozone indices based on a fixed 24

threshold. In addition, the diagnostics are very sensitive to the detailed structure of ozone at the edge of 25

the vortex, which depends on both transport and chemistry and varies considerably between models. In 26

this respect it is also noteworthy that the annual cycle of total ozone over 60°–90°S varies considerably 27

between models (middle panel of Figure 15). Tegtmeier and Shepherd [2006] have shown how the 28

annual cycle of total ozone in this region can vary tremendously between different versions of the 29

CMAM, even while the basic photochemical relationship between springtime and summertime ozone 30

anomalies is maintained. 31

32


29

Further evidence that the ozone hole area, in particular, is not a good diagnostic of the ozone hole can 1

be seen in the performance of the LMDZrepro model: While it is obvious from Figure 14 that the 2

model severely underestimates column ozone throughout southern hemispheric spring and even winter 3

due to a cold pole problem (see Figure 2), the maximum ozone hole area from this model agrees best 4

with observations among the participating models. On the other hand, the ozone mass deficit is 5

significantly overestimated in LMDZrepro, which more accurately reflects what is seen in Figure 14. 6

Therefore, the ozone mass deficit should be the preferred diagnostic over the ozone hole area in 7

agreement with the findings of Bodeker et al. [2005], but the weakness that also this diagnostic is based 8

on a fixed threshold remains (see discussion above). 9

10

7. Past variability and trends in temperature and total column ozone 11

12

Having assessed the simulated climatological temperature and tracer structure in the CCMs we now 13

examine their ability to track past temporal changes in ozone and temperature. 14

15

7.1. Total Ozone Variations and Trends 16

17

Temporal changes in ozone are evaluated by calculating temporal anomalies in total column ozone. To 18

calculate the total column ozone anomalies a regression model is used to recreate the ‘detrended mean 19

annual cycle’, i.e., the average annual cycle over the period 1980 to 1989 that would occur in the 20

absence of trends. This detrended mean annual cycle is subtracted from total ozone time series (see 21

Appendix A). 22

23

Figure 15 shows the area weighted global annual (90°S to 90°N) as well as spring time polar (60° to 24

90°) monthly mean total column ozone anomalies from the CCMs and observations (left panels) and 25

the detrended mean annual cycles calculated using the method described in Appendix A (right panels). 26

27

The essential characteristics of the global mean annual cycles are also listed in Table 5. The 28

measurement precision for the individual total column ozone measurements making up the 29

observational data sets is typically a few percent, and spatial and temporal averaging reduces this 30

further. This high precision, and the good agreement between the four observational data sets, suggests 31

that the values spanned by the measurement data sets are likely to be representative of the true state of 32


30

the atmosphere and its variability. The observed differences between the data sets most likely result 1

from different spatial coverage and uncertainties introduced through spatial interpolation over the polar 2

caps in winter where measurements are not generally available. The features already discussed for the 3

1990s climatology in section 6 are also valid for the 1980s: Apart from differences in the mean value 4

between the models and the observations, most models reproduce the observed amplitude and phasing 5

of the observed mean annual cycle. The MRI model shows the greatest difference in global mean 6

annual cycle in comparison to observations (mean of 366 DU compared to observed mean of 299 DU, 7

see bottom right panel of Figure 15). This bias is consistent with low CH4 (section 4.1), older mean age 8

of air (section 4.3) and very low Cly and HCl concentrations in the upper stratosphere (section 5). The 9

high global mean annual cycle for MAECHAM4CHEM is also consistent with low CH4. 10

11

Observations indicate no significant change in global total column ozone prior to 1980, overall 12

decreasing values between the late 1970s to early 1990s, a relative minimum during 1992 and 1993, 13

increase between 1994 and 1998 and no significant changes between 1998 and 2004. The ozone time 14

series is modulated by the 11-year solar cycle, with maxima at about 1970, 1980, 1990, and 2000 [e.g., 15

WMO, 2003]. The relative minimum during 1992 and 1993 can likely be related to Mt. Pinatubo 16

eruption effects [Gleason et al., 1993]. 17

18

The CCMs show a wide range of behavior with significant differences in global-mean total ozone 19

trends and variability over the period 1980 to 2000, which is partly expected as not all models include 20

the parameterization of the solar cycle and volcanic eruption effects (see Table 2). In the six model 21

simulations that extend back to 1960 (AMTRAC, CMAM, E39C, GEOSCCM, ULAQ, and WACCM), 22

two models (E39C and WACCM) show little or no changes prior to 1980. Little or no changes are also 23

indicated by the observations and are to be expected based on the chlorine loading. However, two 24

models (AMTRAC and ULAQ) show positive ozone anomalies prior to 1980. 25

26

The majority of models represent the observed time evolution between 1980 and 2000 very well, 27

especially those that include both the solar cycle and the volcanic eruption effects. Although the 28

magnitude differs, most of the models that include volcanic eruption effects capture the early 1990s 29

minimum seen in the observations (CCSRNIES, E39C, LMDZ-REPRO, MRI, MAECHAM4CHEM, 30

SOCOL, UMSLIMCAT, ULAQ, and WACCM), while others (AMTRAC and UMETRAC) do not. 31

The differences in the ability of the models to capture the mid-1990s minimum are likely dependent on 32


31

their ability to capture the response in global ozone to the Mt. Pinatubo volcanic eruption. Only 1

GEOSCCM and CMAM do not consider effects of volcanic eruptions (see Table 2). AMTRAC and 2

UMETRAC show a much larger than observed trend after 1980 (see discussion below). 3

4

Over the Arctic, the February to April means of total column ozone anomalies show a wide range of 5

variability with some models (SOCOL and WACCM) showing little or no change in Arctic ozone since 6

1980, while other models (e.g. in particular UMETRAC, and to a lesser extent AMTRAC and MRI) 7

show large negative trends. Only one of the models (CCSRNIES) tracks the observed ozone minimum 8

in the mid-1990s and returns close to 1980 ozone levels in recent years. Understanding the observed 9

record of Arctic springtime ozone over decadal timescales requires accounting for the combined effects 10

of chlorine loading and dynamical variability [WMO, 2003], so differences in model behavior should 11

reflect differences in one or both of these processes. Of the models shown in Figure 15, SOCOL has the 12

smallest trend in Cly while AMTRAC and UMETRAC have the largest trends (Figure 12b), and 13

accordingly these models have Arctic ozone trends that are respectively too small and too large. The 14

larger negative ozone trend in UMETRAC compared with AMTRAC, despite very similar Cly trends, is 15

likely the result of UMETRAC’s large cold bias in the lower stratosphere spring (Figure 2) which 16

would amplify the chemical ozone loss. Consideration of Arctic late-winter/spring temperature trends 17

(Figure 16, discussed further below) indicates that MRI exhibits the most excessive cooling of all 18

models, and accordingly this model has an Arctic ozone trend that is too large. 19

20

Over the Antarctic all models show negative trends in ozone but with some models (UMETRAC and 21

AMTRAC) showing negative anomalies higher than observed. Others (e.g. CCSRNIES, E39C, and 22

SOCOL) show about half those observed. Current understanding of Antarctic ozone loss suggests that 23

the observed record over decadal timescales is mainly driven by chlorine loading [WMO, 2003]. 24

Accordingly, the models with the largest (UMETRAC, AMTRAC) and smallest (E39C, SOCOL) Cly 25

trends exhibit respectively too-strong and too-weak ozone trends. 26

27

7.2. Temperature Variations and Trends 28

29

Figure 16 shows time series of zonally averaged temperature anomalies at 50 hPa between 1960 and 30

2005 from the REF1 simulations as well as ERA-40 and radiosonde data. The RATPAC data are used 31

for comparisons with the global anomalies, and the ERA-40 data are only included since 1980, when 32


32

satellite observations became available. The temperature anomalies are calculated with respect to a 1

mean reference period between 1980 and 1989 using three month averages for February to April in the 2

polar northern hemisphere (60-90°N), September to November in the polar southern hemisphere (60-3

90°S) and annual averages for the global anomalies. A linear temperature trend in K/decade is 4

calculated for each model using results between 1980 and 1999, and the statistical significance of that 5

trend being non-zero is evaluated using a Student t-test at the 95% level. 6

7

In the northern hemisphere (60-90°N) winter (Figure 16, top panel) the year-to-year temperature 8

anomalies range between ±5K. Overall, 10 out of 13 models show a cooling trend between 1980 and 9

2000, although in only 6 of those models the trend is statistically significant (assuming each year is 10

independent). The ERA-40 data indicate a statistically significant cooling trend of -2.1 K/decade, a 11

value larger than all but three of the models (GEOSCCM -2.4 K/decade; LMDZ-REPRO -3.0 12

K/decade; MRI -3.5 K/decade). The relatively large interannual variability of the high latitude winter 13

hemisphere and small magnitude of the trend makes it difficult to determine accurate trends over only 14

20 years of data. 15

16

In the high latitude southern hemisphere (60-90°S) winter (Figure 16, middle panel) a stronger cooling 17

trend is observed in the models and observations. In all model simulations, a cooling trend is found 18

which varies between -1.1 K/decade and -4.5 K/decade, while the ERA-40 analysis indicates a cooling 19

of -2.4 K/decade; the trends in the observations and in 9 out of 13 of the models show statistically 20

significant trends. The stronger cooling trend in the southern hemisphere high latitudes is expected due 21

to the larger southern hemisphere polar ozone loss. Over the length of the model simulations, the 22

interannual variability in the southern hemisphere appears larger than in the northern hemisphere, 23

although at higher altitudes, this relationship does not hold (10 hPa) or reverses (1 hPa). Of the six 24

model simulations that extend back to 1960 (AMTRAC, CMAM, E39C, GEOSCCM, ULAQ, and 25

WACCM all show statistically significant cooling between 1960 and 1980 although weaker in 26

magnitude compared to 1980-1999. 27

28

The annual global temperature anomalies from observations and the CCMs (Figure 16, lower panel) all 29

show cooling between 1980 and 1999. The model temperature trends range from -0.29 K/decade to -30

0.97 K/decade, with 7 out of 13 models showing statistically significant trends. The year to year 31

variability between the ERA-40 analysis and the RATPAC climatology is relatively small until around 32


33

1995, the magnitude of the cooling trends in the ERA-40 (-0.77 K/decade; not significant) and 1

RATPAC (-1.0 K/decade significant) data do differ. The six models that extend back to 1960 all show 2

a statistically significant 1960 to 1980 cooling trend between -0.13 to -0.44 K/decade compared to 3

RATPAC at -0.45 K/decade. 4

5

Superimposed on the overall observed cooling trend are large stratospheric warming signals due to the 6

three major volcanic eruptions (Agung, 1963; El Chichón, 1982; Pinatubo, 1991) which occur because 7

volcanic aerosols absorb both incoming solar radiation and outgoing thermal radiation [Ramaswamy et 8

al., 2001]. The magnitude of the observed temperature anomaly from ERA-40 and RATPAC is around 9

0.5 K for El Chichón and around 0.75 K for Mt. Pinatubo. From the eleven models that include 10

volcanic aerosols a subset of eight models (see Table 2) considered both chemical and direct radiative 11

effects of enhanced stratospheric aerosol abundance. The volcanic eruptions of El Chichón and Mt. 12

Pinatubo are clearly visible in these models. In the models that include direct radiative effects of 13

enhanced stratospheric aerosol abundance there is a tendency to over predict the magnitude of the 14

temperature response. The heating rates have not been uniformly specified in different simulations 15

which is responsible for some of the inter-model differences in the warming signals. The observed 16

‘step-like’ behavior of the long-term evolution of 50 hPa temperature, i.e. obviously lower temperature 17

after the disappearance of the aerosol signal and more or less no trend in the following years, is only 18

reproduced by those models which include the solar cycle effects as well as chemical and direct 19

radiative effects from volcanic eruptions (AMTRAC, CCSRNIES, E39C, MAECHAM4CHEM, MRI, 20

SOCOL, and UMETRAC). It appears that the long-term behavior of the lower stratospheric 21

temperatures has a linear trend caused by the greenhouse gas effect (stratospheric cooling) and 22

stratospheric ozone depletion which is modulated by the solar cycle, that causes higher temperatures in 23

solar maximum than in solar minimum [Santer et al., 2006]. 24

25

The computed global trends are in good agreement with the model intercomparison of Shine et al. 26

[2003] who compare various model simulated trends (with imposed ozone changes) to observations. 27

At 50 hPa, the CCM average model trends (using only statistically significant models) between 1980-28

1999 are -0.6 K/decade, while the similar average shown in Figure 5 of Shine et al. [2003] is -0.55 29

K/decade. Both sets of models underpredict the RATPAC observations and are closer to the satellite 30

(MSU) observation which show a cooling of -0.8 K/decade. It is noted that the RATPAC climatology 31

may poses a cold biases compared to the MSU observations in the tropical stratosphere [Randel and 32


34

Wu, 2006]. At higher altitudes the average global temperature trends (not shown) indicate cooling of -1

0.6 K/decade at 10 hPa and -1.6 K/decade at 1 hPa, compared to Shine et al. [2003] which show -0.6 2

K/decade at 10 hPa and -1.9 K/decade at 1 hPa. In the southern hemisphere polar region, the cooling is 3

considerably smaller at both 10 hPa (-0.8 K/decade) and 1hPa (-0.95 K/decade) compared to 50 hPa (-4

2.67 K/decade). In the northern hemisphere polar region, there is slightly less cooling at 10 hPa (–1.2 5

K/decade) and at 1 hPa (-1.6 K/decade) compared to 50 hPa (-2.0 K/decade). 6

7

In the coupled atmosphere ocean models used for the 2007 IPCC, statistically significant global mean 8

cooling trends (-0.4 to -0.8 K/decade between 1980 and 2000) at 50hPa are found in models that 9

include changes in stratospheric ozone [Cordero and Forster, 2006]. However, for the models that 10

neglect changes in stratospheric ozone, significantly less cooling at 50hPa was found. 11

12

8. Concluding Summary 13

14

In this paper a detailed assessment of the current generation of coupled chemistry-climate models 15

(CCMs) has been performed using simulations of the stratosphere from 1980 to 2000. The simulated 16

climatological thermal structure and distribution of tracers has been compared with meteorological 17

analyses and trace gas observations. 18

19

All the models used in this study reproduce the vertical profile of global and annual mean temperature 20

in the stratosphere fairly accurately, which implies that the models accurately represent the radiative 21

heating and cooling, even though they all use their own internally generated ozone fields in the 22

calculations. However, temperature biases still occur in the high latitudes in winter and spring. For the 23

northern hemisphere the errors are rather small and on average the temperature bias in the lower 24

stratosphere is small, though given the variability of the northern winter one needs to be cautious in 25

estimating biases from just 10 years of observations. In addition most of the models exhibit the correct 26

sensitivity to variations in wave fluxes from the troposphere. These are both notable improvements 27

relative to the results shown in Austin et al. [2003]. In the southern hemisphere most models still have 28

cold biases with the polar vortex breaking down much later than observed, and on average there has 29

been little improvement compared to the results reported in Austin et al. [2003]. The main consequence 30

of the southern hemisphere temperatures biases is that the ozone hole season lasts longer than observed. 31

32


35

The transport in the CCMs has been assessed by examining the distributions of methane and mean age, 1

as well as the propagation of tropical water vapor “tape recorder”. There is in general a large spread in 2

the simulated tracer fields (methane, tropical water vapor, mean age) within the models, indicating 3

large variations in the transport. However, for each diagnostic a large fraction of the spread is due to 4

only a few models. For around half the models there is a much smaller spread for all tracers and the 5

tracer fields are in general in good agreement with observations. Previous model-data comparisons 6

have identified serious model deficiencies in simulating the mean age of air and propagation of the 7

water vapor tape-recorder [Hall et al., 1999], but better agreement is found for many of the CCMs 8

examined here. For the models that do have significant biases in the tracer fields the cause is generally 9

not known, but in some cases can be attributed to specific model features, e.g., very low resolution or 10

an upper boundary located at relatively low altitude. 11

12

The spread in the modeled water vapor distribution from the CCMs is extremely large, with lower 13

stratospheric water vapor varying between 2 and 7 ppm. In contrast to the transport diagnostics, this is 14

not due to only a small subset of models. This spread between the models is primarily due to inter-15

model variability in the equatorial tropopause temperatures, rather than differences in transport. Most 16

models reproduce the correct phase of the annual cycle of equatorial temperatures at 100 hPa, but there 17

are variations in the amplitude and, more importantly, the annual mean value. In general, models with a 18

warm bias at 100 hPa tend to be too moist throughout the stratosphere, and those with a cold bias tend 19

to be too dry. Although the temperatures near the tropical tropopause are still an important issue, the 20

simulations shown here are a significant improvement on those shown in Pawson et al. [2000], where 21

the annual cycles of many of the models were unrealistic, and the differences between models spanned 22

about 20K. 23

24

There are also substantial quantitative differences between the simulated HCl and Cly. For example, the 25

simulated peak value of Cly in polar spring at 50 hPa varies from around 1.2 ppb to over 3.5 ppb. 26

Observations of 3 ppb in 1992 and 3.3 ppb in 2005 indicate that the higher values are more realistic, 27

and simulated peaks around or less than 2.5 ppb, as simulated in several CCMs, are far too low. A 28

likely cause of the differences is transport, and, consistent with this, the CCMs that have problems 29

simulating the CH4 distribution also generally have problems simulating Cly and HCl. However, 30

transport differences cannot explain all the model-model differences, and other factors play a role in 31

determining the model-model differences in Cly. 32


36

1

The CCMs are generally able to reproduce the observed amplitude and phasing of the mean annual 2

cycle in total column ozone in different latitude bands, except in southern high latitudes. However, 3

most models exhibit large annual-mean biases, with the majority overestimating total ozone. 4

Furthermore, in most models the simulated ozone hole is too small in terms of both ozone hole area and 5

ozone mass deficit. CCMs show a large range of ozone trends over the past 25 years and large 6

differences in comparisons with observations. 7

8

Modeled global temperature trends are in reasonable agreement with satellite and radiosonde 9

observations and show a significant cooling trend from 1980 to present-day. Perturbations by volcanic 10

eruptions are well captured in many models, but generally the temperature response is over-predicted. 11

In the northern hemisphere polar region the polar temperatures are dominated by interannual variability 12

and models as well as observations only show a small cooling trend. The southern hemisphere polar 13

region shows a significant cooling trend both in observations and in models, but the temperature 14

variability is overestimated by many models. The observed ‘step-like’ behavior of the long-term 15

evolution of 50 hPa temperature, i.e. obviously lower temperature after the disappearance of the aerosol 16

signal and more or less no trend in the following years, can be reproduced by most models that include 17

the solar cycle effects and volcanic eruptions. Most models that neglect changes in stratospheric ozone 18

generally show significantly less cooling at 50 hPa between 1980 and 2000 [Cordero and Forster, 19

2006]. 20

21

In this study we have focused on the representation of some of the main dynamical and transport 22

processes in CCMs. An assessment of stratospheric chemistry and microphysics as well as the radiation 23

in CCMs is still outstanding. Also, sensitivity studies with individual models are needed to provide 24

more information on model qualities and more research is needed that focuses on specific aspects in 25

more detail. This includes further examination of some issues identified in this study, e.g., examination 26

of differences in the chlorine budget/partitioning between models, and more detailed examination of 27

the impact of the solar cycle and QBO on ozone. Analysis is also needed of the representation of the 28

tropical tropopause layer, stratosphere-troposphere exchange, as well as coupled modes of troposphere-29

stratosphere variability in the CCMs [Eyring et al., 2005b]. 30

31


37

The CCMs examined here have performed, or are currently performing, predictions of the future 1

evolution of the ozone layer. The results shown here provide insight into the ability of the different 2

CCMs to model key dynamical and transport processes and the observed state of the stratosphere. This 3

in turn provides guidance on how to interpret any differences in the CCM predictions. The predictions 4

from the suite of CCMs are currently being examined and will be described in a follow up study. 5

6

APPENDIX A. 7

8

The following approach has been adopted to calculate total column ozone anomalies. A regression 9

model of the form: 10

O3 = A1 +B1× t + 11

A2×sin(2π(t-0.5)/12) + A3×cos(2π(t-0.5)/12) + 12

A4×sin(4π(t-0.5)/12) + A5×cos(4π(t-0.5)/12) + 13

A6×sin(6π(t-0.5)/12) + A7×cos(6π(t-0.5)/12) + 14

A8×sin(8π(t-0.5)/12) + A9×cos(8π(t-0.5)/12) + 15

B2×t×sin(2π(t-0.5)/12) + B3×t×cos(2π(t-0.5)/12) + 16

B4×t×sin(4π(t-0.5)/12) + B5×t×cos(4π(t-0.5)/12) 17

where t is the number of months after the start of 1980 (January 1980 = 1, December 1989 = 120), is fit 18

to the modeled monthly mean total column ozone using the standard linear least squares regression 19

method described in Press et al. [1989]. The fit generates the coefficients A1 to A9 and B1 to B5. The 20

A1 term accounts for the baseline offset as at 1 January 1980 while the B1×t term accounts for the 21

subsequent trend on that offset. The terms with coefficients A2 to A9 account for the stationary part of 22

annual cycle, while the terms with coefficients B2 to B5 account for seasonally dependent trends in that 23

stationary annual cycle. The coefficients A1 to A9 can then be used to recreate the ‘detrended mean 24

annual cycle’, i.e., the average annual cycle over the period 1980-1989 that would occur in the absence 25

of trends. These functional fits, evaluated monthly, are subtracted from the raw data to produce 26

anomaly time series, which are divided by A1/100 to produce percentage anomalies. 27

28

Acknowledgements. 29

Co-ordination of this study was supported by the Chemistry-Climate Model Validation Activity 30

(CCMVal) for WCRP's (World Climate Research Programme) SPARC (Stratospheric Processes and 31


38

their Role in Climate). We thank the British Atmospheric Data Centre, the SPARC data center and the 1

UK Met Office for providing the facilities for central data archives. Special thanks go to Georgiy 2

Stenchikov for providing the heating rates, David Considine for providing sulfate aerosol densities, 3

Steven Montzka for his help extending WMGHG and halogen time series, and A.R. Ravishankara and 4

Stanley Sander for providing chemical kinetics from JPL 2005 for the CCMVal reference simulations. 5

We also wish to thank Eric Nash for calculating ERA-40 heat flux data and highly appreciate 6

comments on a draft manuscript by Ulrich Schumann. The European groups acknowledge support of 7

the SCOUT-O3 Integrated Project which is funded by the European Commission. CMAM is supported 8

by CFCAS and NSERC. 9

10


39

References 1

Akiyoshi, H., T. Sugita, H. Kanzawa, and N. Kawamoto (2004), Ozone perturbations in the Arctic 2

summer lower stratosphere as a reflection of NOx chemistry and planetary scale wave activity, J. 3

Geophys. Res., 109, D03304, doi:10.1029/2003JD003632. 4

Anderson, J., J. M. Russell III, S. Solomon, and L. E. Deaver (2000), Halogen Occultation Experiment 5

confirmation of stratospheric chlorine decreases in accordance with the Montreal Protocol, J. 6

Geophys. Res., 105, 4483-4490. 7

Anderson, J. L. et al. (2004), The new GFDL global atmosphere and land model AM2/LM2: 8

Evaluation with prescribed SST simulations, J. Climate, 17, 4641-4673. 9

Andrews, A. E., K. A. Boering, B. C. Daube, S. C. Wofsy, M. Loewenstein, H. Jost, J. R. Podolske, C. 10

R. Webster, R. L. Herman, D. C. Scott, G. J. Flesch, E. J. Moyer, J. W. Elkins, G. S. Dutton, D. F. 11

Hurst, F. L. Moore, E. A. Ray, P. A. Romashkin, and S. E. Strahan (2001), Mean ages of 12

stratospheric air derived from in situ observations of CO2, CH4, and N2O, J. Geophys. Res., 106, 13

32295-32314, 10.1029/2001JD000465. 14

Aoki, S., T. Nakazawa, T. Machida, S. Sugawara, S. Morimoto, G. Hashida, T. Yamanouchi, K. 15

Kawamura, and H. Honda (2003), Carbon dioxide variations in the stratosphere over Japan, 16

Scandinavia and Antarctica, Tellus B, 55, 178-186. 17

Austin, J. (2002), A three-dimensional coupled chemistry-climate model simulation of past 18

stratospheric trends, J. Atmos. Sci., 59, 218-232. 19

Austin, J., D. Shindell, S. R. Beagley, C. Brühl, M. Dameris, E. Manzini, T. Nagashima, P. Newman, 20

S. Pawson, G. Pitari, E. Rozanov, C. Schnadt, and T. G. Shepherd (2003), Uncertainties and 21

assessments of chemistry-climate models of the stratosphere, Atmos. Chem. Phys., 3, 1-27. 22

Austin, J., and R. J. Wilson (2006), Ensemble simulations of the decline and recovery of stratospheric 23

ozone, J. Geophys. Res., submitted. 24

Austin, J., R. J.Wilson, F. Li, and H. Vömel (2006), Evolution of water vapor concentrations and 25

stratospheric age of air in coupled chemistry-climate model simulations, J. Atmos. Sci., submitted. 26

Beagley, S. R., J. de Grandpré, J. N. Koshyk, N. A. McFarlane, and T.G. Shepherd (1997), Radiative-27

dynamical climatology of the first-generation Canadian Middle Atmosphere Model, Atmos.-Ocean, 28

35, 293–331. 29


40

Bloom, S., A. da Silva, D. Dee, M. Bosilovich, J.-D. Chern, S. Pawson, S. Schubert, M. Sienkiewicz, I. 1

Stajner, W.-W. Tan, and M.-L. Wu (2005), Documentation and Validation of the Goddard Earth 2

Observing System (GEOS) Data Assimilation System - Version 4. Technical Report Series on 3

Global Modeling and Data Assimilation 104606 , 26. 4

Bodeker, G. E., H. Shiona, and H. Eskes (2005), Indicators of Antarctic ozone depletion, Atmos. Chem. 5

Phys., 5, 2603-2615. 6

Brühl, C., S. R. Drayson, J. M. Russell III, P. J. Crutzen, J. McInerney, P. N. Purcell, H. Claude, H. 7

Gernand, T. McGee, I. McDermid, and M. R. Gunson (1996), HALOE ozone channel validation, J. 8

Geophys. Res., 101, 10,217-10,240. 9

Butchart, N., J. Austin, J. Knight, A. A. Scaife, and M. L. Gallani (2000), Response of the stratospheric 10

climate to projected changes in the concentrations of well mixed greenhouse-gases from 1992 to 11

2051, J. Climate, 13, 2142-2159. 12

Collins, W. D., P. J. Rasch, B. A. Boville, J. J. Hack, J. R. McCaa, D. L. Williams, J. T. Kiehl, B. P. 13

Briegleb, C. Bitz, S.–J Lin, M. Zhang, and Y. Dai (2004), Description of the NCAR Community 14

Atmosphere Model (CAM3), Nat. Cent. For Atmos. Res., Boulder, Colorado. 15

Cordero, E. C., and P. M. de F. Forster (2006), Stratospheric Variability and Trends in IPCC Model 16

Simulations, J. Clim., submitted. 17

Dameris, M., V. Grewe, M. Ponater, R. Deckert, V. Eyring, F. Mager, S. Matthes, C. Schnadt, A. 18

Stenke, B. Steil, C. Brühl, and M. Giorgetta (2005), Long-term changes and variability in a 19

transient simulation with a chemistry-climate model employing realistic forcings, Atmos. Chem. 20

Phys., 5, 2121-2145. 21

Dameris, M., S. Matthes, R. Deckert, V. Grewe, and M. Ponater (2006), Solar cycle effect delays onset 22

of ozone recovery, Geophys. Res. Lett., 33, L03806, doi:10.1029/2005GL024741. 23

de Grandpré, J., S. R. Beagley, V. I. Fomichev, E. Griffioen, J. C. McConnell, A. S. Medvedev, and T. 24

G. Shepherd (2000), Ozone climatology using interactive chemistry: Results from the Canadian 25

Middle Atmosphere Model, J. Geophys. Res., 105, 26,475-26,492. 26

Douglass, A. R., M. R. Schoeberl, R. S. Stolarski, J. W. Waters, J. M. Russell III, and A. E. Roche 27

(1995), Interhemispheric differences in springtime production of HCl and ClONO2 in the polar 28

vortices, J. Geophys. Res., 100, 13,967-13,978. 29


41

Douglass, A. R., M. J. Prather, T. M. Hall, S. E. Strahan, P. J. Rasch, L. C. Sparling, L. Coy, and J. M. 1

Rodriguez (1999), Choosing meteorological input for the global modeling initiative assessment of 2

high-speed aircraft, J. Geophys. Res., 104, 27,545-27,564. 3

Egorova, T., E. Rozanov, V. Zubov, E. Manzini, W. Schmutz, and T. Peter (2005), Chemistry-climate 4

model SOCOL: a validation of the present-day climatology, Atmos. Chem. Phys., 5, 1557-1576. 5

Eyring, V., D. E. Kinnison, and T. G. Shepherd (2005a), Overview of planned coupled chemistry-6

climate simulations to support upcoming ozone and climate assessments, SPARC Newsletter No. 7

25, 11-17. 8

Eyring, V., N. R. P. Harris, M. Rex, T. G. Shepherd, D. W. Fahey, G. T. Amanatidis, J. Austin, M. P. 9

Chipperfield, M. Dameris, P. M. de F. Forster, A. Gettelman, H. F. Graf, T. Nagashima, P. A. 10

Newman, S. Pawson, M. J. Prather, J. A. Pyle, R. J. Salawitch, B. D. Santer, and D. W. Waugh 11

(2005b), A strategy for process-oriented validation of coupled chemistry-climate models, Bull. Am. 12

Meteorol. Soc., 86, 1117–1133. 13

Fioletov, V. E., and T. G. Shepherd (2003), Seasonal persistence of midlatitude total ozone anomalies, 14

Geophys. Res. Lett., 30, 10.1029/2002GL016739. 15

Fioletov, V. E., G. E. Bodeker, A. J. Miller, R. D. McPeters, and R. Stolarski (2002), Global ozone and 16

zonal total ozone variations estimated from ground-based and satellite measurements: 1964-2000, 17

J. Geophys. Res., 107, 4647, doi:10.1029/2001JD001350. 18

Fomichev, V. I., C. Fu, J. de Grandpré, S. R. Beagley, V. P. Ogibalov, and J.C. McConnell (2004), 19

Model thermal response to minor radiative energy sources and sinks in the middle atmosphere, J. 20

Geophys. Res., 109, doi:10.1029/2004JD004892. 21

Free, M., D. J. Seidel, J. K. Angell, J. Lanzante, I. Durre, and T. C. Peterson (2005), Radiosonde 22

Atmospheric Temperature Products for Assessing Climate (RATPAC): A new data set of large-area 23

anomaly time series, J. Geophys. Res., 110, D22101, doi:10.1029/2005JD006169. 24

Garcia, R. R., and S. Solomon (1985), The effect of breaking gravity waves on the dynamics and 25

chemical composition of the mesosphere and lower thermosphere, J. Geophys. Res., 90, 3850-3868. 26

Garcia, R. R., and B. A. Boville (1994), “Downward control” of the mean meridional circulation and 27

temperature distribution of the polar winter stratosphere, J. Atmos. Sci., 51, 2238–2245. 28


42

Garcia, R. R., D. Marsh, D. Kinnison, B. Boville, and F. Sassi (2006), Simulations of secular trends in 1

the middle atmosphere, 1950-2003, J. Geophys. Res., submitted. 2

Gates, W. L., J. S. Boyle, C. Covey, C. G. Dease, C. M. Doutriaux, R. S. Drach, M. Fiorino, P. J. 3

Gleckler, S. Marlais, T. J. Phillips, G. L. Potter, B. D. Santer, K. R. Sperber, K. E. Taylor, and D. 4

N. Williams (1999), An overview of the results of the Atmospheric Model Intercomparison Projects 5

(AMIP), Bull. Am. Meteorol. Soc., 80, 29-56. 6

Giorgetta M. A., and L. Bengtsson (1999), The potential role of the quasi-biennial oscillation in the 7

stratosphere-troposphere exchange as found in water vapor in general circulation model 8

experiments, J. Geophys. Res., 104, 6003-6020. 9

Gleason, J. F., P. K. Bhartia, J. R. Herman, R. McPeters, P. Newman, R. S. Stolarski, L. Flynn, G. 10

Labow, D. Larko, C. Seftor, C. Wellemeyer, W. D. Komhyr, A. J. Miller, and W. Planet (1993), 11

Record low global ozone in 1992, Science, 260, 523-526. 12

Gregory, D. G., G. J. Shutts, and J. R. Mitchell (1998), A new gravity wave drag scheme incorporating 13

anisotropic orography and low-level wave breaking: impact upon the climate of the UK 14

Meteorological Office unified model, Q.J.R. Meteorol. Soc., 124, 463-494. 15

Gregory, A. R., and V. West (2002), The sensitivity of a model's stratospheric tape recorder to the 16

choice of advection schemes, Q.J.R. Meteorol. Soc., 128, 1827-1846. 17

Grewe, V. (2006), The origin of ozone, Atmos. Chem. Phys., accepted. 18

Grooß, J. -U., and J. M. Russell III (2005), Technical note: A stratospheric climatology for O3, H2O, 19

CH4, NOx, HCl and HF derived from HALOE measurements, Atmos. Chem. Phys., 5, 2797-2807. 20

Hall, T. M., D. W. Waugh, K. A. Boering, and R. A. Plumb (1999), Evaluation of transport in 21

stratospheric models, J. Geophys. Res., 104, 18,815-18,840. 22

Harries, J. E., J. M. Russell III, A. F. Tuck, L. L. Gordley, P. Purcell, K. Stone, R. Bevilacqua, M. R. 23

Gunson, G. Nedoluha, and W. A Traub (1996), Validation of measurements of water vapor from 24

the Halogen Occultation Experiment, HALOE, J. Geophys. Res., 101, 10,205-10,216. 25

Hines, C. O. (1997), Doppler-spread parameterization of gravity-wave momentum deposition in the 26

middle atmosphere. Part2: Broad and quasi monochromatic spectra, and implementation, J. Atmos. 27

and Solar-Terr. Phys., 59, 387-400. 28


43

Holton, J. R. (1982), The role of gravity wave induced drag and diffusion in the momentum budget of 1

the mesosphere, J. Atmos. Sci., 39, 791-799. 2

Hurrell, J. W., J. Hack, J. Caron, J. Rosinski, and D. Shea (2005), A new sea surface temperature and 3

sea ice boundary data set for the NCAR Community Atmospheric Model, J. Clim., submitted. 4

IPCC (Intergovernmental Panel on Climate Change) (2001), Climate change 2001: The scientific basis. 5

Contribution of Working Group 1 to the Third Assessment Report, J.T. Houghton (Eds.), 6

Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. 7

Iwasaki, T., S. Yamada, and K. Tada (1989), A parameterization scheme of orographic gravity wave 8

drag with the different vertical partitioning, part 1: Impact on medium range forecasts, J. Meteor. 9

Soc. Japan, 67, 11-41. 10

Jackman, C. H., E. L. Fleming, S. Chandra, D. B. Considine, and J. E. Rosenfield (1996), 11

Past, present, and future modeled ozone trends with comparisons to observed trends, 12

J. Geophys. Res., 101, 28,753-28,768. 13

JPL (Jet Propulsion Laboratory) (2003), Chemical kinetics and photochemical data for use in 14

stratospheric modeling: Evaluation 14. JPL-02-25, NASA Panel for Data Evaluation, NASA Aero- 15

nautics and Space Administration, Pasadena, CA, USA, 73. 16

Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G. White, 17

J. Woollen Y. Zhu, M. Chelliah, W. Ebisuzaki, W. Higgins, J. Janowiak, K. C. Mo, C. Ropelewski, 18

J. Wang, A. Leetmaa, R. Reynolds, R. Jenne, and D. Joseph (1996), The NCEP/NCAR 40-year 19

reanalysis project, Bull. Am. Meteorol. Soc., 77, 437-471. 20

Kiehl, J. T., J. J. Hack, G. B. Bonan, B. A. Boville, D. L. Williamson, and P. J. Rasch (1998) The 21

National Center for Atmospheric Research Community Climate Model: CCM3, J. Clim., 11, 1131-22

1150. 23

Kley, D., E. J. Stone, W. R. Henderson, J. W. Drummond, W. J. Harrop, A. L. Schmeltekopf, T. L. 24

Thompson, and R. H. Winkler (1979), In situ measurements of the mixing ratio of water vapor in 25

the stratosphere, J. Atmos. Sci., 36, 2513-2524. 26

Kurokawa, J., H. Akiyoshi, T. Nagashima, H. Masunaga, T. Nakajima, M. Takahashi, and H. Nakane 27

(2005), Effects of atmospheric sphericity on stratospheric chemistry and dynamics over Antarctica, 28

J. Geophys. Res., 110, D21305, doi:10.1029/2005JD005798. 29


44

Lean, J. L. (2000), Evolution of the Sun’s spectral irradiance since the maunder minimum, Geophys. 1

Res. Lett., 27, 2425–2428. 2

Lean, J. L., G. J. Rottman, H. L. Kyle, T. N. Woods, J. R. Hickey, and L. C. Puga (1997), Detection 3

and parameterization of variations in solar mid- and near-ultraviolet radiation (200 to 400 nm), J. 4

Geophys. Res., 102, 29,939-29,956. 5

Leonard, B. P., A. P. Lock, and M. K. MacVean (1995), The NIRVANA scheme applied to one-6

dimensional advection, Methods Heat Fluid Flow, 5, 341–377. 7

Lin, S.-J. (2004), A 'vertically Lagrangian' finite volume dynamical core for global models, Mon. Wea. 8

Rev., 132, 2293-2307. 9

Lindzen, R. S. (1981), Turbulence and stress owing to gravity wave and tidal breakdown, J. Geophys. 10

Res., 86, 9707-9714. 11

Lott, F., and M. Miller (1997), A new subgrid scale orographic drag parameterization; its testing in the 12

ECMWF model, Quart. J. Roy. Meteor. Soc., 123, 101-127. 13

Lott, F., L. Fairhead, F. Hourdin, and P. Levan (2005), The stratospheric version of LMDz: Dynamical 14

Climatologies, Arctic Oscillation, and Impact on the Surface Climate, Climate Dynamics, 25, doi: 15

10.1007/s00382-005-0064. 16

Manzini, E., N. A. McFarlane, and C. McLandress (1997), Impact of the Doppler spread 17

parameterization on the simulation of the middle atmosphere circulation using the MA/ECHAM4 18

general circulation model, J. Geophys. Res., 102, 25,751-25,762. 19

Manzini, E., B. Steil, C. Brühl, M. A. Giorgetta, and K. Krüger (2003), A new interactive chemistry 20

climate model. 2: Sensitivity of the middle atmosphere to ozone depletion and increase in 21

greenhouse gases: implications for recent stratospheric cooling, J. Geophys. Res., 108, 4429, 22

doi:10.1029/2002JD002977. 23

McFarlane, N. A. (1987), The effect of orographically excited gravity wave drag on the general 24

circulation of the lower stratosphere and troposphere, J. Atmos. Sci., 44, 1775-1800. 25

Miller, M. J., T. N. Palmer, and R. Swinbank (1989), Parameterization and influence subgrid-scale 26

orography in general circulation and numerical weather prediction models, Meteorol. Atmos. Phys., 27

40, 84-109. 28


45

Miller, A. J., R. M. Nagatani, L. E. Flynn, S. Kondragunta, E. Beach, R. Stolarski, R. D. McPeters, P. 1

K. Bhartia, M. T. Deland, C. H. Jackman, D. J. Wuebbles, K. O. Putten, and R. P. Cebula (2002), A 2

cohesive total ozone data set from SBUV(/2) satellite system, J. Geophys. Res., 107, 3

doi:10.1029/2001JD000853. 4

Mote, P. W., K. H. Rosenlof, M. E. McIntyre, E. S. Carr, J. C. Gille, J. R. Holton, J. S. Kinnersley, H. 5

C. Pumphrey, J. M. Russell III, and J. W. Waters (1996), An atmospheric tape recorder: The imprint 6

of tropical tropopause temperatures on stratospheric water vapor, J. Geophys. Res., 101, 3989-4006, 7

10.1029/95JD03422. 8

Newman, P. A., E. R. Nash, and J. E. Rosenfield (2001), What controls the temperature of the Arctic 9

stratosphere during spring?, J. Geophys. Res., 106, 19,999-20,010. 10

Numaguti, A. (1993), Dynamics and energy balance of the Hadley circulation and the tropical 11

precipitation zones: Significance of the distribution of evaporation, J. Atmos. Sci., 50, 1874-1887. 12

Park, J. H. , J. M. Russell III, L. L. Gordley, S. R. Drayson, D. C. Benner, J. M. McInerney, M. R. 13

Gunson, G. G. Toon, B. Sen, J.-F. Blavier, C. R. Webster, E. C. Zipf, P. Erdman, U. Schmidt, and 14

C. Schiller (1996), Validation of Halogen Occultation Experiment CH4 measurements from the 15

UARS, J. Geophys. Res., 101, 10,183-10,204. 16

Park, J. H., M. K. W. Ko, C.H. Jackman, R. A. Plumb, J. A. Kaye, and K. H. Sage (1999), Models and 17

Measurements Intercomparison II. NASA/TM-1999-209554. 18

Pawson, S., K. Kodera, K. Hamilton, T. G. Shepherd, S. R. Beagley, B. A. Boville, J. D. Farrara, T. D. 19

A. Fairlie, A. Kitoh, W. A. Lahoz, U. Langematz, E. Manzini, D. H. Rind, A. A. Scaife, K. Shibata, 20

P. Simon, R. Swinbank, L. Takacs, R. J. Wilson, J. A. Al-Saadi, M. Amodei, M. Chiba, L. Coy, J. de 21

Grandpré, R. S. Eckman, M. Fiorino, W. L. Grose, H. Koide, J. N. Koshyk, D. Li, J. Lerner, J. D. 22

Mahlman, N. A. McFarlane, C. R. Mechoso, A. Molod, A. O'Neill, R. B. Pierce, W. J. Randel, R. B. 23

Rood, and F. Wu (2000), The GCM-Reality Intercomparison Project for SPARC: Scientific Issues 24

and Initial Results, Bull. Am. Meteorol. Soc., 81, 781-796. 25

Pitari, G., S. Palermi, G. Visconti, and R. G. Prinn (1992), Ozone response to a CO2 doubling: results 26

from a stratospheric circulation model with heterogeneous chemistry, J. Geophys. Res., 97, 5953-27

5962. 28

Pitari, G. (1993), A numerical study of the possible perturbation of stratospheric dynamics due to 29

Pinatubo aerosols: Implications for tracer transport, J. Atmos. Sci., 50, 2443–2461. 30


46

Pitari, G., E. Mancini, V. Rizi, and D. Shindell (2002), Feedback of future climate and sulfur emission 1

changes an stratospheric aerosols and ozone, J. Atmos. Sci., 59, 414–440. 2

Press, W. H., B. R. Flannery, S. A. Teukolsky, and W. T. Vettering (1989), Numerical Recipes in Pascal, 3

759 pp., Cambridge Univ. Press, New York. 4

Pope, V., M. Gallani, P. Rowntree, and R. Stratton (2000), The impact of new physical 5

parametrizations in the Hadley Centre climate model - HadCM3, Climate Dynamics, 16, 123-146. 6

Ramaswamy, V., M.-L. Chanin, J. Angell, J. Barnett, D. Gaffen, M. Gelman, P. Keckhut, Y. 7

Koshelkov, K. Labitzke, J.-J. R. Lin, A. O'Neill, J. Nash, W. Randel, R. Rood, K. Shine, M. 8

Shiotani, and R. Swinbank (2001), Stratospheric temperature trends: observations and model 9

simulations, Reviews of Geophysics, 39, 71-122. 10

Randel, W. J., F. Wu, S. J. Oltmans, K. Rosenlof, and G. E. Nedoluha (2004), Interannual Changes of 11

Stratospheric Water Vapor and Correlations with Tropical Tropopause Temperatures, J. Atmos. 12

Sci., 61, 2133-2148. 13

Rozanov, E., M. Schraner, C. Schnadt, T. Egorova, M. Wild, A. Ohmura, V. Zubov, W. Schmutz, and 14

Th. Peter (2005), Assessment of the ozone and temperature variability during 1979–1993 with the 15

chemistry-climate model SOCOL, Adv. Space. Res., 35, 1375-1384. 16

Roeckner, E., K. Arpe, L. Bengtsson, M. Christoph, M., Claussen, L. Dümenil, M. Esch, M. Giorgetta, 17

U. Schlese, and U. Schulzweida (1996), The atmospheric general circulation model ECHAM-4: 18

Model description and simulation of present-day climate, Report No. 218, Max-Planck-Institut für 19

Meteorologie, Hamburg. 20

Russell, J. M. III, L. L. Gordley, J. H. Park, S. R. Drayson, W. D. Hesketh, R. J. Cicerone, A. F. Tuck, 21

J. E. Frederick, J. E. Harries, and P. J. Crutzen (1993), The Halogen Occultation Experiment, J. 22

Geophys. Res., 98, 10,777–10,797. 23

Russell, J. M. III, L. E. Deaver, M. Luo, J. H. Park, L. L. Gordley, A. F. Tuck, G. C. Toon, M. R. 24

Gunson, W. A. Traub, D. G. Johnson, K. W. Jucks, D. G. Murcray, R. Zander, I. G. Nolt, and C. R. 25

Webster (1996), Validation of hydrogen chloride measurements made by HALOE from the UARS 26

platform, J. Geophys. Res., 101, 10,151-10,162. 27

Sankey, D., and T. G. Shepherd (2003), Correlations of long-lived chemical species in a middle 28

atmosphere general circulation model, J. Geophys. Res., 108, doi:10.1029/2002JD002799. 29


47

Santee, M. L., L. Froidevaux, G. L. Manney, W. G. Read, J. W. Waters, M. P. Chipperfield, A. E. 1

Roche, J. B. Kumer, J. L. Mergenthaler, and J. M. Russell III (1996), Chlorine deactivation in the 2

lower stratospheric polar regions during late winter: Results from UARS, J. Geophys. Res., 101, 3

18,835-18,860. 4

Santer, B. D., J. E. Penner, and P. W. Thorne (2006), How well can the observed vertical temperature 5

changes be reconciled with our understanding of the causes of these changes? In: Temperature 6

Trends in the Lower Atmosphere: Steps for Understanding and Reconciling Differences. A Report 7

by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research 8

[Karl, T.R., S.J. Hassol, C.D. Miller, and W.L. Murray (eds.)], National Oceanic and Atmospheric 9

Administration, National Climatic Data Center, Asheville, NC, USA, in press. 10

Sato, M., J. E. Hansen, M. P. McCormick, and J. B. Pollack (1993), Stratospheric aerosol optical 11

depths, 1850-1990, J. Geophys. Res., 98, 22987-22994. 12

Scaife, A. A., N. Butchart, C. D. Warner, and R. Swinbank (2002), Impact of a spectral gravity wave 13

parameterization on the stratosphere in the Met Office Unified Model, J. Atmos. Sci., 59, 1473-14

1489. 15

Schmidt, G.A., R. Ruedy, J. E. Hansen, I. Aleinov, N. Bell, M. Bauer, S. Bauer, B. Cairns, V. Canuto, 16

Y. Cheng, A. Del Genio, G. Faluvegi, A. D. Friend, T. M. Hall, Y. Hu, M. Kelley, N. Y. Kiang, D. 17

Koch, A. A. Lacis, J. Lerner, K. K. Lo, R. L. Miller, L. Nazarenko, V. Oinas, Ja. Perlwitz, Ju. 18

Perlwitz, D. Rind, A. Romanou, G. L. Russell, Mki. Sato, D. T. Shindell, P. H. Stone, S. Sun, N. 19

Tausnev, D. Thresher, and M.-S. Yao (2006), Present-Day Atmospheric Simulations Using GISS 20

ModelE: Comparison to In Situ, Satellite, and Reanalysis Data, J. Atmos. Sci., 19, 153-192. 21

Scinocca, J.F. (2003), An Accurate Spectral Non-Orographic Gravity Wave Parameterization for 22

General Circulation Models, J. Atmos. Sci., 60, 667-682. 23

Scinocca, J. F., and N. A. McFarlane (2000), The parameterization of drag induced by stratified flow 24

over anisotropic orography, Quart. J. Roy. Meteor. Soc., 126, 2353-2393. 25

Scinocca, J. F., and N. A. McFarlane (2004), The Variability of Modeled Tropical Precipitation, J. 26

Atmos. Sci., 61, 1993–2015. 27

Sessler, J., P. Good, A. R. MacKenzie, and J. A. Pyle (1996), What role do type I polar stratospheric 28

cloud and aerosol parameterizations play in modeled lower stratospheric chlorine activation and 29

ozone loss?, J. Geophys. Res., 101, 28817-28835. 30


48

Shibata, K., and M. Deushi (2005), Partitioning between resolved wave forcing and unresolved gravity 1

wave forcing to the quasi-biennial oscillation as revealed with a coupled chemistry-climate model, 2

Geophys. Res. Lett., L12820, doi:10.1029/2005GL022885. 3

Shibata, K., H. Yoshimura, M. Ohizumi, M. Hosaka, and M. Sugi (1999), A simulation of troposphere, 4

stratosphere and mesosphere with an MRI/JMA98 GCM, Meteorology and Geophysics, 50, 15-53. 5

Shibata, K., M. Deushi, T. T. Sekiyama, and H. Yoshimura (2005), Development of an MRI chemical 6

transport model for the study of stratospheric chemistry, Papers in Meteorology and Geophysics, 7

55, 75-119. 8

Shine, K. P., M. S. Bourqui, P. M. de F. Forster, S. H. E. Hare, U. Langematz, P. Braesicke, V. Grewe, 9

M. Ponater, C. Schnadt, C. A. Smiths, J. D. Haighs, J. Austin, N. Butchart, D. T. Shindell, W. J. 10

Randels, T. Nagashima, R. W. Portmann, S. Solomon, D. J. Seidel, J. Lanzante, S. Klein, V. 11

Ramaswamy, and M. D. Schwarzkopf (2003), A comparison of model-simulated trends in 12

stratospheric temperatures, Q. J. Royal Meteor. Soc., 129, 1565-1588. 13

Shutts, G. J. (1990), A new gravity wave drag parameterization scheme for the unified model. 14

Turbulence and diffusion note 204, available from National Meteorological archive, Met Office, 15

Exeter. 16

Smith, A. K., and L. V. Lyjak (1985), An observational estimate of gravity wave drag from the 17

momentum balance in the middle atmosphere, J. Geophys. Res., 90, 2233-2241. 18

Solomon, S., R. R. Garcia, and F. Stordal (1985), Transport Processes and Ozone Perturbations, J. 19

Geophys. Res., 90, D7, 12,981-12,989. 20

SPARC (2000), SPARC assessment of upper tropospheric and stratospheric water vapour, Tech. Rep. 21

WMO-TD No. 1043, WCRP Series Report No. 113, SPARC Report No. 2, Berrieres le Buisson 22

Cedex. 23

SPARC (2006), Assessment of Stratospheric Aerosol Properties (ASAP), Tech Rep. WCRP-124, 24

WMO/TD- No. 1295, SPARC Report No. 4. 25

Steil, B., M. Dameris, C. Brühl, P. J. Crutzen, V. Grewe, M. Ponater, and R. Sausen (1998), 26

Development of a chemistry module for GCMs: First results of a multi-annual integration, Ann. 27

Geophys., 16, 205– 228. 28


49

Steil, B., C. Brühl, E. Manzini, P. J. Crutzen, J. Lelieveld, P. J. Rasch, E. Roeckner, and K. Krüger 1

(2003), A new interactive chemistry climate model. 1: Present day climatology and interannual 2

variability of the middle atmosphere using the model and 9 years of HALOE/UARS data, J. 3

Geophys. Res., 108, 4290, doi:10.1029/2002JD002971. 4

Steinbrecht, W., B. Haßler, C. Brühl, M. Dameris, M. A. Giorgetta, V. Grewe, E. Manzini, S. Matthes, 5

C. Schnadt, B. Steil, and P. Winkler (2006a), Interannual variation patterns of total ozone and 6

temperature in observations and model simulations, Atmos. Chem. Phys., 6, 349-374. 7

Steinbrecht, W., H. Claude, F. Schoenenborn, S. McDermid, T. Leblanc, S. Godin, T. Song, D. P. J. 8

Swart, Y. J. Meijer, G. E. Bodeker, B. J. Connor, N. Kaempfer, K. Hocke, Y. Calisesi, N. 9

Schneider, J. de la Noe, A. D. Parrish, I. S. Boyd, C. Brühl, B. Steil, M. A. Giorgetta, E. Manzini, 10

L. W. Thomason, J. M. Zawodny, M. P. McCormick, J. M. Russell III, P. K. Bhartia, R. S. 11

Stolarski, and S. M. Hollandsworth-Frith (2006b), Long-Term Evolution of Upper Stratospheric 12

Ozone at Selected Stations of the Network for the Detection of Stratospheric Change (NDSC), J. 13

Geophys. Res., in press. 14

Stenke, A., and V. Grewe (2006), Lagrangian transport of water vapor and cloud water in the 15

ECHAM4 GCM and its impact on the cold bias, Climate Dynamics, submitted. 16

Stolarski, R. S., A. R. Douglass, S. Steenrod, and S. Pawson (2006): Trends in Stratospheric Ozone: 17

Lessons Learned from a 3D Chemical Transport Model, J. Atmos. Sci., 63, 1028-1041. 18

Struthers, H., K. Kreher, J. Austin, R. Schofield, G. E. Bodeker, P. V. Johnston, H. Shiona, and A. 19

Thomas (2004), Past and future simulations of NO2 from a coupled chemistry-climate model in 20

comparison with observations, Atmos. Chem. Phys., 4, 2227–2239. 21

Swinbank R., and A. O' Neill (1994), A stratosphere-troposphere data assimilation system, Monthly 22

Weather Review, 122, 686-702. 23

Takigawa, M., M. Takahashi, and H. Akiyoshi (1999), Simulation of ozone and other chemical species 24

using a Center for Climate Systems Research/National Institute for Environmental Studies 25

atmospheric GCM with coupled stratospheric chemistry, J. Geophys. Res., 104, 14,003-14,018. 26

Taylor, K. E., D. Williamson, and F. Zwiers (2000), The Sea surface Temperature and sea-ice 27

concentration boundary concentrations for AMIP II simulations, PMCDI Report 60, Program for 28

Climate Model Diagnosis and Intercomparison, Lawrence Livermore National Laboratory. 29


50

Tegtmeier, S., and T. G. Shepherd (2006), Persistence and photochemical decay of springtime total 1

ozone anomalies in the Canadian Middle Atmosphere Model, Atmos. Chem. Phys., submitted. 2

Tian, W., and M.P. Chipperfield (2005), A new coupled chemistry–climate model for the stratosphere: 3

The importance of coupling for future O3-climate predictions, Quart. J. Roy. Meteor. Soc., 131, 4

281–304. 5

Uppala, S. M., P. W. Kållberg, A. J. Simmons, U. Andrae, V. da Costa Bechtold, M. Fiorino, K. K. 6

Gibson, J. Haseler, A. Hernandez, G. A. Kelly, X. Li, K. Onogi, S. Saarinen, N. Sokka, R. P. Allan, 7

E. Andersson, K. Arpe, M. A. Balmaseda, A. C. M. Beljaars, L. van de Berg, J. Bidlot, N. 8

Bormann, S. Caires, F. Chevallier, A. Dethof, M. Dragosavac, M. Fisher, M. Fuentes, S. 9

Hagemann, E. Hólm, B.J. Hoskins, L. Isaksen, P. A. E. M. Janssen, R. Jenne, A. P. McNally, J.-F. 10

Mahfouf, J.-J. Morcrette, N. A. Rayner, R.W. Saunders, P. Simon, A. Sterl, K. E. Trenberth, A. 11

Untch, D. Vasiljevic, P. Viterbo, and J. Woollen (2005), The ERA-40 re-analysis, Quart. J. R. 12

Meteorol. Soc., 131, 2961-3012.doi:10.1256/qj.04.176. 13

Waugh, D. W., and T. M. Hall (2002), Age of stratospheric air: theory, observations, and models, Rev. 14

Geophys., 40, 10.1029/2000RG000101. 15

Warner, C. D., and M. E. McIntyre (1996), On the propagation and dissipation of gravity wave spectra 16

through a realistic middle atmosphere, J. Atmos. Sci., 53, 3213-3235. 17

Weber, M., S. Dhomse, F. Wittrock, A. Richter, B.-M. Sinnhuber, and J. P. Burrows (2003), Dynamical 18

control of NH and SH winter/spring total ozone from GOME observations in 1995 - 2002?, 19

Geophys. Res. Lett., 30, doi:10.1029/2002GL016799. 20

World Meteorological Organization (WMO)/United Nations Environment Programme (UNEP) (2003), 21

Scientific Assessment of Ozone Depletion: 2002, World Meteorological Organization, Global 22

Ozone Research and Monitoring Project, Report No. 47, Geneva, Switzerland. 23

Zubov, V., E. Rozanov, and M. Schlesinger (1999), Hybrid scheme for tree-dimensional advective 24

transport, Mon. Wea. Rev., 127, 1335-1346. 25

26


51

Table 1: CCMs used in this study. The models are listed alphabetically by name. The horizontal 1 resolution is given in either degrees latitude × longitude (grid point models), or as T21, T30, etc., 2 which are the resolutions in spectral models corresponding to triangular truncation of the spectral 3 domain with 21, 30, etc., wave numbers, respectively. All CCMs have a comprehensive range of 4 chemical reactions. The time varying concentrations of CO2 as well as other WMGHGs are considered 5 in the radiation schemes of all models. 6


52

Model Investigators Underlying GCM

Domain / resolution

Radiative Feedbacks

Advection Scheme

O-GWD

NonO-GWD

Reference

AMTRAC J. Austin AM2 [Anderson et al., 2004]

2° x 2.5° 48 L 0.0017 hPa

O3, H2O Lin [2004] Anderson et al. [2004]

Anderson et al. [2004]

Austin et al. [2006]; Austin and Wilson [2006]

CCSR NIES

H. Akiyoshi, T. Nagashima, M. Yoshiki

CCSRNIES [Numaguti, 1993]

T42 34 L 0.01 hPa

O3, H2O, CH4, N2O, CFCs

Spectral in the horizontal, finite difference for the vertical

McFarlane [1987]

Hines [1997]

Akiyoshi et al. [2004]; Kurokawa et al. [2005]

CMAM D. Plummer, T.G. Shepherd

CCCma AGCM3 [Scinocca and McFarlane, 2004]

T32 71L 0.0006 hPa

O3, H2O Spectral in the horizontal, finite elements in the vertical

Scinocca and McFarlane [2000]

Scinocca et al. [2003]

Beagley et al. [1997]; de Grandpré et al. [2000]

E39C M. Dameris, V. Eyring , V. Grewe, M. Ponater

ECHAM4 [Roeckner et al., 1996]

T30 39L 10 hPa

O3, CH4, N2O, H2O, CFCs

Semi- Lagrangian [Williamson & Rasch, 1994]

Miller et al. [1989]

None Dameris et al. [2005, 2006]

GEOS CCM

A. Douglass, P. Newman, S. Pawson, R. Stolarski

GEOS-4 [Bloom et al., 2005]

2° x 2.5° 55 L 0.01hP

O3, H2O, CH4, N2O, CFC-11, CFC-12

Finite-Volume [Lin, 2004]

Kiehl et al. [1998]

Adapted from Garcia and Boville [1994]

Bloom et al. [2005]; Stolarski et al. [2006]

LMDZ-REPRO

S. Bekki, M. Marchand

LMDz4 [Lott et al., 2005]

2.5° x 3.75° 50 L 0.07 hPa

O3, CH4, N2O, H2O, CFC-11, CFC-12

First-order volume finite scheme [Hourdin and Armengaud, 1999]

Lott and Miller [1997]

Hines et al. [1997]

In preparation

MA ECHAM4 CHEM

C. Brühl, M. Giorgetta, E. Manzini

MA ECHAM4 [Manzini et al., 1997]

T30 39 L 0.01 hPa

O3, H2O, CH4

Fluxform Semi-Lagrange SPITFIRE [Steil et al., 2003]

McFarlane [1987]

Hines [1997]

Manzini et al. [2003]; Steil et al. [2003]

MRI K. Shibata, M. Deushi

MRI/JMA98 [Shibata et al., 1999]

T42 68 L 0.01 hPa

O3, CH4, N2O

Hybrid semi-Lagrangian [Shibata et al., 2005]

Iwasaki et al. [1989], only short-wavelength GW

Hines [1997]

Shibata and Deushi [2005]; Shibata et al. [2005]

SOCOL E. Rozanov MA ECHAM4 [Manzini et al., 1997]

T30 39 L 0.1 hPa

O3, CH4, N2O, H2O, CFCs

Hybrid advection scheme [Zubov et al. 1999]

McFarlane [1987]

Hines [1997]

Egorova et al. [2005]; Rozanov et al. [2005]


53

ULAQ E. Mancini, G. Pitari

ULAQ-GCM [Pitari et al., 1992]

10° x 22.5 26 L 0.04 hPa

O3, H2O, CH4, N2O, CFCs, HCFCs, aerosols.

Flux-form Eulerian fully explicit scheme [e.g Pitari et al. 2002]

Rayleigh friction [Smith and Lyjak, 1985]

Pitari et al. [2002]

UMETRAC G. Bodeker, N. Butchart, H. Struthers

UM [Pope et al., 2000]

2.5° x 3.75° 64 L 0.01 hPa

O3, H2O Extended non-oscillatory integrally reconstructed volume-averaged numerical advection [Leonard et al., 1995]

Shutts [1990]

Warner and McIntyre [1996]

Austin et al. [2002]; Struthers et al. [2004]

UM SLIMCAT

M. Chipperfield, W. Tian

UM [Pope et al., 2000]

2.5° x 3.75° 64 L 0.01 hPa

O3, CH4, N2O, H2O

Qunitic-mono Gregory and West [2002]

Gregory et al. [1998]

Warner and McIntyre (1996)

Tian and Chipperfield [2005]

WACCM B. Boville, R Garcia, A. Gettelman, D. Kinnison, D. Marsh

CAM [Collins et al., 2004]

4° x 5° 66 L 10-6 hPa

O3, H2O, N2O, CH4, CFC-11, CFC-12 NO, CO2, O2

Finite Volume, [Lin, 2004]

McFarlane [1987]

Based on Lindzen [1981], Holton [1982], Garcia and Solomon [1985]

Garcia et al. [2006]

1 2


54

Table 2: External forcings and representation of QBO used in the model simulations. Changes in 1 halocarbons are prescribed as monthly means based on Table 4B-2 of WMO [2003]. WMGHGs are 2 based on IPCC [2001] in all simulations. Both data sets are extended through 2004 based on 3 observations. 4

External Forcings Volcanoes

Model Length of Simulation

SSTs Solar Variability Chemical

effects Direct radiative effects

QBO

AMTRAC 1960-2004

Hurrell et al. [2005]

Intensity of the 10.7 cm radiation of the sun from

Lean [2000]

SADs from G. Stenchikov

(pers. comm., 2005)

Calculated on line with climate model radiation

scheme

NO

CCSRNIES 1980-2004

HadISST1, Rayner et al.

[2003]


Lean et al. [1997]

SADs from D. Considine

(pers. comm., 2005)

Calculated on line using aerosol data from [Sato et

al., 1993]

Forced

CMAM 1960-2004

Modeled NO NO NO NO

E39C 1960-1999


[2003]


Lean et al. [1997]

SADs from Jackman et al.

[1996]

Pre-calculated net heating rates [Kirchner et al.,

1999]

Forced

GEOSCCM 1960-2003


[2003]

NO NO NO NO

LMDZrepro 1979-1999

AMIP II, Taylor et al.

[2000]

NO Modeled SADs NO NO

MAECHAM4 CHEM 1980-1999


[2003]


Lean et al. [1997]

SADs from Jackman et al.

[1996]

Pre-calculated net heating rates [Kirchner et al.,

1999]

Forced

MRI 1980-2004


[2003]


Lean et al. [1997]

SADs from D. Considine

(pers. comm., 2005)

Calculated on line from the optical depth and effective

radius based on the updated dataset of Sato et

al. [1993]

Internally generated

SOCOL 1980-2004


[2003]

Lean [2000] SADs from SPARC [2006]

Calculated on line using aerosol data from SPARC

[2006]

Forced

ULAQ 1970-2004


[2003]

NO SADs from D. Considine (pers comm., 2005)

Calculated on line [Pitari, 1993]

NO

UMETRAC 1980-1999


[2003]


Lean et al. [1997]

SADs from D. Considine (pers comm., 2005)

Pre-calculated heating rates (G. Stenchikov, pers.

comm., 2005)

Internally generated

UMSLIMCAT 1980-1999

AMIP II Gates et al.

[1999]

NO SADs from D. Considine (pers comm., 2005)

NO Internally generated

WACCM 1950-2003

Hurrell et al. [2005]

Daily values of 10.7 cm radio flux from

www.sec.noaa.gov

SADs from D. Considine (pers comm., 2005)

NO NO

5


55

Table 3. Mean and standard deviation (SD) of 100 hPa heat fluxes and 50 hPa temperatures; 1 correlation coefficient, R, between the temperature and heat flux; zero-intercept, T0, and slope, β, of the 2 line. The temperature is the area weighted temperature at 50 hPa poleward of 60° for February and 3 March in the northern hemisphere and August and September in the southern hemisphere, and the heat 4 flux is averaged between 40° and 80° for January and February in the northern hemisphere and July 5 and August in the southern hemisphere. 6 7

Northern Hemisphere Heat flux Temperature

Met. Analysis / Model

Mean SD Mean SDR T0 β

NCEP 14.2 2.51 214.9 2.98 0.79 201.6 0.94AMTRAC 12.6 2.20 214.7 2.80 0.63 204.6 0.80CCSRNIES 11.2 2.35 210.6 3.51 0.60 200.5 0.90CMAM 13.8 2.17 214.1 3.51 0.71 199.4 1.09E39C 19.6 4.41 214.3 4.70 0.87 196.1 0.93GEOSCCM 12.4 2.45 213.7 3.71 0.76 199.4 1.16LMDZrepro 12.7 3.01 214.0 3.33 0.87 201.8 0.96MAECHAM4CHEM 14.6 2.66 212.8 3.44 0.75 198.6 0.97MRI 16.5 4.46 212.5 7.13 0.86 189.7 1.38SOCOL 15.2 2.59 215.8 3.82 0.80 197.8 1.18ULAQ 14.7 2.19 214.3 2.75 0.49 205.3 0.61UMSLIMCAT 15.4 2.37 213.8 3.40 0.70 198.4 1.00WACCM 11.4 2.23 216.2 2.21 0.56 209.8 0.56 Southern Hemisphere NCEP 7.8 1.59 196.7 2.43 0.74 187.8 1.14AMTRAC 10.5 3.07 203.3 4.75 0.89 188.8 1.38CCSRNIES 6.5 1.35 189.4 1.39 0.53 185.8 0.54CMAM 7.9 1.67 191.0 2.31 0.80 181.8 1.21E39C 7.7 1.18 189.8 0.99 0.48 186.7 0.40GEOSCCM 8.8 2.29 197.7 4.34 0.90 182.8 1.71LMDZrepro 4.9 1.61 193.1 0.71 0.63 191.8 0.28MAECHAM4CHEM 6.3 1.51 192.5 1.15 0.57 189.7 0.44MRI 7.1 1.37 192.2 1.19 0.62 188.4 0.54SOCOL 7.5 1.61 196.2 1.98 0.82 188.6 1.01ULAQ 5.7 2.18 196.5 2.40 -0.11 197.1 -0.12UMSLIMCAT 7.1 1.76 197.0 2.23 0.65 191.2 0.82WACCM 10.1 2.29 198.5 3.93 0.88 183.3 1.51 8

9


56

Table 4: Mean Antarctic ozone depletion indices for observations and CCMs: the minimum total ozone 1 column poleward of 40°S, the maximum Antarctic ozone hole area (OHA) for the period 1 September 2 to 30 November defined by the 220 DU contour, and the ozone mass deficit calculated as the mean of 3 daily values between 1 September and 30 November based on a 220 DU threshold (OMD) have been 4 calculated for each year. The values given in the table are 1995 to 1999 averages for the three ozone 5 depletion indices (a 5-year period is used as there are large trends in the southern hemisphere indices 6 from 1980 to 1995). 7

8 Model Minimum SH OHA OMDNIWA data base 91 27 23AMTRAC 66 23 25CCSRNIES 122 22 11CMAM 86 21 19E39C 114 17 10GEOSCCM 114 17 11LMDZrepro 48 27 41MAECHAM4CHEM 130 14 6MRI 98 17 11SOCOL 65 36 35ULAQ 79 27 19UMETRAC 86 21 20UMSLIMCAT 97 17 17WACCM 106 15 12

9


57

Table 5: A summary of the main characteristics of the global detrended mean annual cycles plotted in 1 Figure 16. The amplitude is the mean annual cycle peak-to-peak amplitude. Data based on observations 2 are shown in the upper part of the table. 3 4 Model Maximum Day of maximum Minimum Day of minimum AmplitudeGround-based 304 DU 16 April 291 DU 16 December 13.1 DU Merged 305 DU 18 April 293 DU 12 December 12.0 DU NIWA 302 DU 19 April 291 DU 12 December 11.4 DU SBUV 308 DU 12 April 296 DU 11 December 12.0 DU AMTRAC 288 DU 10 April 279 DU 4 January 9.4 DU CCSRNIES 318 DU 23 March 303 DU 30 October 14.5 DU CMAM 295 DU 3 April 285 DU 16 December 10.3 DU E39C 337 DU 31 March 320 DU 9 December 16.6 DU GEOSCCM 317 DU 20 April 305 DU 24 December 11.3 DU LMDZrepro 334 DU 31 March 320 DU 2 December 13.9 DU MAECHAM4CHEM 355 DU 18 March 345 DU 28 June 9.2 DU MRI 375 DU 25 March 357 DU 16 November 18.8 DU SOCOL 293 DU 27 March 281 DU 5 December 11.2 DU ULAQ 307 DU 14 August 297 DU 27 December 10.6 DU UMETRAC 311 DU 11 April 301 DU 5 January 10.2 DU UMSLIMCAT 298 DU 12 April 286 DU 30 December 12.8 DU WACCM 318 DU 2 April 306 DU 22 December 11.2 DU 5

6


58

Annual− and Global−Mean T[K]

200 210 220 230 240 250 260 270 280 290 300T[K]

1000700

500

300

200

100

70

50

30

20

10

7

5

3

2

1

pre

ssu

re [

hP

a]

−10 −5 0 5 100Model Temperature bias (K)

1000700

500

300

200

100

70

50

30

20

10

7

5

3

2

1

Pre

ssu

re (

hP

a)

−100 −50 0 50 1000.0

0.2

0.4

0.6

0.8

1.0 NCEPERA40 (1992−2001)

AMTRACCCSRNIES

CMAME39C

GEOSCCMLMDZrepro

−100 −50 0 50 1000.0

0.2

0.4

0.6

0.8

1.0MAECHAM4CHEMMRI

SOCOLULAQ

UMETRACUMSLIMCAT

WACCMUKMO 1

2 Figure 1. Annual global-mean temperatures for the 1990s for the thirteen participating CCMs 3 compared to UKMO, NCEP and ERA-40 meteorological analyses (left), and (right) biases relative to 4 UKMO analyses. Grey area shows UKMO plus and minus 1 standard deviation (σ) about the 5 climatological mean. 6 7

8


59

Dec/Jan/Feb 60−90N

−30 −20 −10 0 100Model Temperature bias (K)

1000

300

100

30

10

3

1

Pre

ssu

re (

hP

a)

0

Mar/Apr/May 60−90N


1000

300

100

30

10

3

1

Pre

ssu

re (

hP

a)

0

Jun/Jul/Aug 60−90S


1000

300

100

30

10

3

1

Pre

ssu

re (

hP

a)

0

Sep/Oct/Nov 60−90S


1000

300

100

30

10

3

1

Pre

ssu

re (

hP

a)

0

____ MULTI−MODEL MEAN ____ AMTRAC ____ CMAM ____ CCSRNIES ____ E39C _ _ GEOSCCM ____ LMDZrepro

_ _ MAECHAM4chem _ _ MRI ____ SOCOL ____ ULAQ _ _ UMETRAC ____ UMSLIMCAT_ _ WACCM 1

2 Figure 2. Temperature biases for 60°-90°N (upper panels) and 60°-90°S (lower panels) for the winter 3 (right) and spring seasons (left). The biases are calculated as in Figure 1. Grey area shows UKMO plus 4 and minus 1 standard deviation (σ) about the climatological mean. 5 6


60

transition to easterlies at 60S

Date

100

30

10

3

1

.3

.1

Pre

ssu

re (

hP

a)

November December January0.0

AMTRAC

CMAMCCSRNIES

GEOSCCMLMDZMAECHAM4chemMRISOCOLULAQUMETRACUMSLIMCATWACCM

UKMO

1 2 3 Figure 3. Climatological mean (1990-1999) for the models and 1992-2002 for the observations) 4 seasonal decent of the zero zonal mean wind lines at 60°S. Grey area shows UKMO plus and minus 1 5 standard deviation (σ) about the climatological mean. 6 7

8


61

0 5 10 15 20 25Heat flux at 100 hPa, Jan and Feb mean

200

205

210

215

220

225

Mea

n T

emp

erat

ure

at

50 h

Pa,

Feb

an

d M

arch

0 5 10 15 20 25Heat flux at 100 hPa, Jan and Feb mean

200

205

210

215

220

225

Mea

n T

emp

erat

ure

at

50 h

Pa,

Feb

an

d M

arch

0 2 4 6 8 10 12Heat flux at 100 hPa, Jul and Aug mean

180

185

190

195

200

205

Mea

n T

emp

erat

ure

at

50 h

Pa,

Au

g a

nd

Sep

0 2 4 6 8 10 12Heat flux at 100 hPa, Jul and Aug mean

180

185

190

195

200

205

Mea

n T

emp

erat

ure

at

50 h

Pa,

Au

g a

nd

Sep

0 5 10 15 20 25200

205

210

215

220

225 AMTRAC

CCSRNIES

CMAM

E39C

GEOSCCM

LMDZrepro

NCEP

0 5 10 15 20 25200

205

210

215

220

225 MAECHAM4chem

MRI

SOCOL

ULAQ

UMSLIMCAT

WACCM

NCEP 1 2

Figure 4. Upper panels: Heat fluxes (v'T') at 100 hPa (averaged over 40°N to 80°N for January and 3 February) versus temperatures at 50 hPa (averaged over 60°N to 90°N for February and March). 4 Shown are all REF1 years for each simulation compared to observations from NCEP assimilations. 5 Lower panels: Same for southern hemisphere, but heat fluxes (v'T') at 100 hPa averaged over 40°S to 6 80°S for July and August versus temperatures at 50 hPa averaged over 60°S to 90°S for August and 7 September. 8 9 10 11

12


62

1 2

3 4

Figure 5. Climatological zonal-mean methane from models and HALOE in ppmv. Upper panels (a-c): 5 Vertical profiles at 80°N in March (left), 0° in March (middle) and 80°S in October for (right). Lower 6 panels (d,e): Latitudinal profiles at 50hPa in March (left) and October (right). Grey dots indicate 7 HALOE observations averaged in equivalent latitudes, black dots are values averaged in latitudes 8 (regular zonal mean), and the grey area shows plus and minus 1 standard deviation (σ) about the 9 climatological zonal mean HALOE. 10

11

12

13 14


63

1 2

3 4 Figure 6. As in Figure 5, but for H2O in ppmv. 5 6 7

8


64

100 hPa Temperature at Equator

J F M A M J J A S O N DMonth

185

190

195

200

Tem

per

atu

re (

K)

UKMO

AMTRAC

CMAM

CCSRNIES

E39C

GEOSCCM

LMDZrepro

MAECHAM4CHEMMRI

SOCOL

ULAQ

UMETRAC

UMSLIMCAT

WACCM

ERA40

1 2

100 hPa Water Vapor at Equator

J F M A M J J A S O N DMonth

1

2

3

4

5

6

7

8

9

Wat

er V

apo

r (p

pm

v)

HALOE

AMTRAC

CMAM

CCSRNIES

E39C

GEOSCCM

LMDZrepro

MAECHAM4CHEM

MRI

SOCOL

ULAQ

UMETRAC

UMSLIMCATWACCM

3 4 Figure 7. Seasonal variation of climatological means at 100 hPa at the equator (a) temperature (upper 5 panel) and (b) water vapor (lower panel). 6 7 8 9 10 11 12

13


65

1 2 Figure 8. Time-height sections of water vapor mixing ratio shown as the deviation (in parts per 3 million by volume) from the time-mean profile, averaged between 10°S and 10°N (‘tape recorder’) for 4 all CCMs and HALOE data. Two consecutive cycles are shown. 5

6


66

1 2 Figure 9. Vertical variation of (a) amplitude, (b) amplitude relative to amplitude at 100 hPa, and (c) 3 phase of minimum (relative to 100hPa) in annual cycle of water vapor averaged between 10°S and 4 10°N. Solid circles are HALOE observations. 5 6 7

8


67

1 2 Figure 10. Mean age of air at (a) 0.5, (b) 10, and (c) 50 hPa. Symbols correspond to observations, see 3 text for details. 4 5 6 7 8

9


68

1 2

3 4 Figure 11. As in Figure 5, but for HCl in ppbv and vertical profiles are shown at (a) 80°N in April (b) 5 0° in April and (c) 80°S in November. Lower panels show latitudinal profiles at 50hPa in (d) April and 6 (e) November. 7 8 9 10 11 12

13


69

1 2 Figure 12. Left: Climatological mean vertical profiles (1990 to 1999) at 80°S in November for Cly in 3 ppbv. Right: Time series of October mean Antarctic Cly at 80°S from CCM model simulations. 4 Estimates of Cly from HALOE HCl measurements in 1992 [Douglass et al., 1995; Santee et al., 1996] 5 and Aura MLS HCl in 2005 [Santee et al., 2006] are shown in addition. 6 7 8

9


70

1 2

3 4 Figure 13. As in Figure 5, but for O3 in ppmv. 5

6


71

1 2 Figure 14. Modeled total column ozone climatologies (1990 to 1999) compared to merged satellite and 3 NIWA assimilated data. 4

5


72

1

1960 1965 1970 1975 1980 1985 1990 1995 2000Year

-50

-40

-30

-20

-10

0

10

20

Ano

mal

y (%

)

200

250

300

350

400

Det

rend

ed m

ean

annu

al c

ycle

198

0-19

89 (D

U)

-30

-25

-20

-15

-10

-5

0

5

10Ano

mal

y (%

)

250

300

350

400

450

500

-15

-10

-5

0

5

Ano

mal

y (%

)

260

280

300

320

340

360

60oN-90oN (Feb-Apr)

Global (annual)

60oS-90oS (Sep-Nov)

J F M A M J J A S O N D

Ground-basedObservations

MergedNIWA SBUV

AMTRACE39CCCSRNIES

ULAQSOCOL

LMDZreproMAECHAM4CHEM

UMETRACUMSLIMCATWACCMMRI

CMAM GEOSCCM

2 3 Figure 15. The panels on the right show the detrended mean annual cycles for each model (1980 to 4 1989). The panels on the left show monthly mean total column ozone anomalies. The anomalies are 5 calculated with the method described in Appendix A between February and April in the polar northern 6 hemisphere (60-90°N, upper panel), between September and November in the polar southern 7 hemisphere (60-90°S, middle panel) and annual averages for the global anomalies (lower panel). The 8 anomaly time series have been smoothed by running a 5 year running mean over the data twice, using 9 the output from the first smoothing as input for the second. Model results are compared to four 10 different global datasets (ground based data, SBUV-SBUV/2, merged satellite data, NIWA assimilated 11 data, adapted from Fioletov et al. [2002]). 12 13


73

1 Figure 16. Modeled and observed trends in time series of global-mean monthly-mean temperature 2 anomalies at 50 hPa from the REF1 simulations. The temperature anomalies are calculated with respect 3 to a mean reference period between 1980 and 1989 using three month averages for February to April in 4 the polar northern hemisphere (60-90°N, upper panel), September to November in the polar southern 5 hemisphere (60-90°S, middle panel) and annual averages for the global anomalies (lower panel). 6 MAECHAM4CHEM, MRI, UMETRAC, and WACCM are shown with dashed lines, all others CCMs 7 with solid lines. A linear temperature trend in K/decade is calculated for each model using data 8 between 1980 and 1999. The temperature trend is given next to the name of each participating model. 9 10 11

assessment of coupled chemistry-climate models: evaluation ...dwaugh1/eyring_etal.pdf · 7...

Documents