Tropospheric adjustment to increasing CO2: its timescale and the role of land–sea contrast
- First Online:
- Cite this article as:
- Kamae, Y. & Watanabe, M. Clim Dyn (2013) 41: 3007. doi:10.1007/s00382-012-1555-1
- 971 Views
Physical processes responsible for tropospheric adjustment to increasing carbon dioxide concentration are investigated using abrupt CO2 quadrupling experiments of a general circulation model (GCM) called the model for interdisciplinary research on climate version 5 with several configurations including a coupled atmosphere–ocean GCM, atmospheric GCM, and aqua-planet model. A similar experiment was performed in weather forecast mode to explore timescales of the tropospheric adjustment. We found that the shortwave component of the cloud radiative effect (SWcld) reaches its equilibrium within 2 days of the abrupt CO2 increase. The change in SWcld is positive, associated with reduced clouds in the lower troposphere due to warming and drying by instantaneous radiative forcing. A reduction in surface turbulent heat fluxes and increase of the near-surface stability result in shoaling of the marine boundary layer, which shifts the cloud layer downward. These changes are common to all experiments regardless of model configuration, indicating that the cloud adjustment is primarily independent of air–sea coupling and land–sea thermal contrast. The role of land in cloud adjustment is further examined by a series of idealized aqua-planet experiments, with a rectangular continent of varying width. Land surface warming from quadrupled CO2 induces anomalous upward motion, which increases high cloud and associated negative SWcld over land. The geographic distribution of continents regulates the spatial pattern of the cloud adjustment. A larger continent produces more negative SWcld, which partly compensates for a positive SWcld over the ocean. The land-induced negative adjustment is a factor but not necessary requirement for the tropospheric adjustment.
KeywordsTropospheric adjustmentRadiative forcingBoundary layerCloud radiative effectTranspose-AMIPAqua-planet experiment
During recent decades, general circulation models (GCMs) have played a major role in studies on past, present and future climate change. Recent works reveal that it is important to investigate changes in radiative forcing and feedback to perturbation of external forcing for understanding climate sensitivity, defined by the global mean surface air temperature (SAT) change in response to doubling of atmospheric CO2 concentration. It is well known that instantaneous CO2 forcing can lead to rapid responses in the atmosphere (e.g., vertical profile of temperature, humidity, cloud and circulation) without changes in global mean SAT, and therefore they induce adjustments in top of the atmosphere (TOA) radiative balance. The processes in the troposphere are called tropospheric adjustment (Gregory and Webb 2008, hereafter GW08; Andrews and Forster 2008; Dong et al. 2009; Colman and McAvaney 2011). Because of the fast timescale, these processes are often included in CO2 radiative forcing and are referred to as effective radiative forcing (Knutti and Hegerl 2008; Webb et al. 2012, hereafter WLG12). Slowdown of the hydrological cycle, i.e., precipitation and evaporation, is also an essential aspect of tropospheric adjustment (Mitchell et al. 1987; Allen and Ingram 2002; Lambert and Webb 2008; Andrews et al. 2009, 2010; Bala et al. 2010).
Hansen et al. (2005) summarized methods that can be used to diagnose the CO2 radiative forcing. One is to run atmospheric GCMs (AGCMs) with fixed sea surface temperature (SST) and sea ice but with different CO2 concentrations (Hansen et al. 2002). The effective radiative forcing is estimated as the radiative perturbation at the TOA, which includes the effects of stratospheric and tropospheric adjustments. This “fixed-SST method” has been widely used because of its convenience and lesser computational burden (Hansen et al. 2005). Another method that has been widely used is a regression method proposed by Gregory et al. (2004). In this approach, a coupled atmosphere–ocean GCM (AOGCM) or AGCM coupled with a slab ocean is integrated with control and abruptly increased CO2 settings. A linear regression of global mean anomalies (changes between the two experiments) of net TOA radiative fluxes on global mean SAT anomaly is computed. The y-axis intercept of the regression line indicates the effective radiative forcing, while the equilibrium climate sensitivity and feedback parameter are given by the x-axis intercept and slope of the regression line. The regression method is useful to diagnose and compare forcing, feedback and equilibrium climate sensitivity among AOGCMs, without running them over centuries. In frameworks of the Coupled Model Intercomparison Project (CMIP) and Cloud Feedback Model intercomparison Project (CFMIP), inter-model spread and robustness of estimated radiative forcing with the above two methods were examined quantitatively (GW08; Andrews et al. 2012a; WLG12). Andrews et al. (2012a) pointed out some differences in effective radiative forcings estimated using the fixed-SST and regression methods, probably owing to a non-linear response of TOA radiative balance in the AOGCM integrations (Gregory et al. 2004).
However, the nature of tropospheric cloud adjustment has not been clearly stated in previous studies. The effective radiative forcing, particularly shortwave (SW) component of the cloud radiative effect (CRE; defined by the difference between all-sky and clear-sky fluxes; Cess et al. 1990), associated with cloud adjustment is estimated by GCMs but has large uncertainty (GW08, Andrews et al. 2012a; WLG12). Several processes potentially important for changes in cloud adjustment and SW CRE (hereafter SWcld) have been suggested. Dong et al. (2009) showed a decreasing total cloud fraction and positive SWcld associated with tropospheric adjustment by using the atmosphere component of HadSM3. Colman and McAvaney (2011) revealed that the essential part of positive perturbation in SW cloud radiation in tropospheric adjustment is not associated with cloud optical properties, but cloud fraction with a version of the Australian Bureau of Meteorology Research Centre (BMRC) climate model. They also pointed out similarities in patterns of tropospheric warming, drying, and instantaneous radiative heating in zonal-mean, height-latitude sections. Watanabe et al. (2011) and Wyant et al. (2012, hereafter W12) reported a positive SWcld and shoaling of the marine boundary layer in the subtropics as factors contributing to tropospheric adjustment in the Model for Interdisciplinary Research on Climate version 5 (MIROC5) and SP-CAM, respectively. It is needed to clarify changes in cloud, temperature, humidity and boundary layer as well as their effects on changes of CRE in the tropospheric adjustment.
One possible way to explore the mechanisms of tropospheric cloud adjustment is to examine transient processes evolving on different timescales following an abrupt CO2 increase. Because tropospheric adjustment is considered as rapid as stratospheric adjustment, it is difficult to detect transient evolution and processes of the tropospheric adjustment using annual or monthly-mean data. After CO2 forcing is imposed, the model atmosphere warms fastest in the first year. This means that the number of samples is very limited, which results in considerable uncertainty in estimation of the adjustment because they contain interannual variability. In addition, a single sensitivity test also contains seasonality in the response to CO2 increase. Dong et al. (2009) examined fast responses to CO2 doubling from six-member ensemble experiments with December and June initial conditions. They presented timescales in the development of land surface warming and changes in tropospheric thermodynamic structure. Wu et al. (2012) conducted a 10-member ensemble of CO2 doubling experiments, with initial conditions in different years. They concluded that the troposphere warms in 1 month, but atmospheric circulation adjusts on a longer timescale. However, the above methods have some limitation for investigating transient evolutions on sub-monthly timescales, because the samples do not cover interannual and seasonal variations.
There is a possibility that some aspects of tropospheric adjustment are driven by rapid warming of the land surface. When atmospheric CO2 is increased in the model, land–sea thermal contrast evolves rapidly because the land surface warms up much faster than the ocean. This land–sea warming contrast in response to CO2 increase (Manabe et al. 1991; Sutton et al. 2007; Dommenget 2009; Boer 2011) slows down atmospheric circulation and the hydrological cycle (Andrews et al. 2009; Dong et al. 2009; Bala et al. 2010; Fasullo 2010; Andrews et al. 2011), and also modifies cloud amount and the CRE (Lambert et al. 2011, hereafter LWJ11; W12). In response to CO2 increases, reduction in stomatal conductance acts as a “CO2 physiological forcing” (Sellers et al. 1996; Dong et al. 2009; Doutriaux-Boucher et al. 2009; Boucher et al. 2009; Andrews et al. 2011, 2012b) that reinforces the land–sea thermal contrast and associated changes in atmospheric circulation and hydrological cycle. The change in dynamical motion affects the cloud profile and total fractions as well as associated CRE in tropospheric adjustment (GW08; LWJ11; W12). However, in the previous works, it was not clarified whether the land–sea warming contrast is essential for tropospheric adjustment. Some kinds of idealized model experiments such as aqua-planet experiments may clarify the role of land and associated temperature contrast between land and ocean for the tropospheric adjustment to increasing CO2.
In this study, we aim to clarify the physical mechanisms of tropospheric cloud adjustment focusing on two aspects: its timescale and land–sea thermal contrast. We conducted abrupt CO2 quadrupling (4×CO2) experiments in a single model, but with various model and experimental configurations (AOGCM, AGCM, aqua-planet model, weather forecast approaches; Sect. 2). Transient adjustment processes on fast timescales are first investigated by a large-member ensemble experiment with increased signal-to-noise ratio, and then idealized model experiments based on an aqua planet are performed to examine the role of the land surface warming on the tropospheric adjustment. Section 2 describes the model used and experimental settings, i.e., standard CMIP5 (Taylor et al. 2012)/CFMIP2 (Bony et al. 2011) experiments and idealized runs. Section 3 explores immediate, transient, and slower processes in tropospheric adjustment, with particular attention to cloud and associated radiative changes over the ocean. Section 4 focuses on the land effect on tropospheric adjustment by investigating tropospheric responses through a series of idealized models with different continent sizes. Section 5 presents concluding discussions, highlighting study results and their significance.
2 Model and experiments
The GCM used is MIROC5 (Watanabe et al. 2010) developed jointly at the Atmosphere and Ocean Research Institute (AORI), University of Tokyo, National Institute for Environmental Studies (NIES), and Japan Agency for Marine-Earth Science and Technology (JAMSTEC). This is one of the models contributing to CMIP5 (Taylor et al. 2012) and the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5). The atmospheric component of MIROC5 has resolution T85 in the horizontal, with vertical 40 Eta (η) levels. The ocean component model has approximately 1° horizontal resolution, and 49 vertical levels with an additional bottom boundary layer. MIROC5 reproduces cloud and water vapor generally well but underestimates high cloud relative to the other CMIP5 models (Jiang et al. 2012). The equilibrium climate sensitivity to doubling CO2 in MIROC5 is 2.6 K, which is 1 K lower than in the previous model version, MIROC3.2 (Hasumi and Emori 2004) because of a difference in SWcld feedback (Watanabe et al. 2010; Shiogama et al. 2012). Dependencies of SWcld feedback and climate sensitivity on the structure of MIROC models are detailed in Watanabe et al. (2012).
2.2 4×CO2 experiments
Summary of experimental configurations using MIROC5 analyzed in this study
Type of model
Length for calculating climatology
Transpose-AMIP II (TAMIP)
Aqua-planet experiment (APE)
5 years after 1 year spin-up
L60, L120, L180, L240, L300
5 years after 1 year spin-up
2.3 Transpose-AMIP II
In recent efforts toward better understanding of systematic errors in climate models, the weather forecast approach was proposed (Phillips et al. 2004; Williams and Brooks 2008). In this approach, climate models are run in “weather forecast mode”, with initial data from operational numerical weather prediction data or reanalyses. Development of forecast errors from all initialized states, averaged over cases, can provide insight into bias processes in the climatologies of long-term simulations (Rodwell and Palmer 2007). The new framework of the international model intercomparison project is referred to as the Transpose AMIP II (TAMIP hereafter; Xie et al. 2012; Williams et al. 2012). This approach can help discern what happens in climate models on fast timescales (e.g., clouds), and why responses differ at longer timescales. More details on TAMIP are available in Williams et al. (2012).
TAMIP data consist of a series of 10-day hindcasts initialized by European Center for Medium-Range Weather Forecasts (ECMWF) analysis for the year of tropical convection (YOTC; Waliser et al. 2012) period (May 2008 to April 2010). For this period, we conducted 4 sets (October 2008, January 2009, April 2009 and July 2009) of 16 hindcasts with the start times at 30 h intervals. In addition to the standard TAMIP experiments, we also did sensitivity tests with identical settings to the hindcasts but imposing quadrupled CO2 in the atmosphere. Composite differences between 64 hindcasts and 4×CO2 runs are calculated at every 3 h up to day 10. This ensures sampling throughout annual and diurnal cycles for a given lead time, which is expected to show how the transient adjustment to the abrupt 4×CO2 evolves in time (S. Bony, personal communication).
2.4 Idealized experiments
To evaluate the role of land–sea thermal contrast in the tropospheric adjustment processes, we conducted idealized experiments with the APE as their basis. We set up five additional configurations: aqua-planet with different continent sizes (60°, 120°, 180°, 240°, and 300° longitudinal widths, hereafter L60, L120, L180, L240, L300, respectively) in the tropics (30°S–30°N). The experiments with different sizes of continent could clarify whether effects of land-sea temperature contrast depend on continent size or not. Land elevation was uniformly set to 10 m, without mountains. Vegetation types were specified with zonally-uniform distribution, derived from the most prominent types at individual latitudes used in MIROC5. Similar to the APE, 6-year integrations were performed for control and sensitivity (4×CO2) runs, with the latter 5 years analyzed in the individual configurations. Results are compared with those of the APE and AMIP in Sect. 4.
3 Tropospheric adjustment in cloud and hydrological cycle
3.1 Timescales of adjustment processes
3.2 Cloud adjustment
3.3 Dynamic and thermodynamic changes and mechanisms for marine PBL shoaling
4 Role of land–sea contrast in tropospheric adjustment
With the existence of a continent at low latitude, increasing SAT over land (Fig. 13b) generates anomalous upward motion in the troposphere (figures not shown); this increases deep convective cloud over land (Fig. 14b). The change in high cloud is greater than for mid-low clouds (Fig. 14f), leading to an increase in total cloud cover and associated negative SWcld response over the continent (Fig. 13f). These changes are consistent with those reported in the previous studies (LWJ11, W12) and are roughly opposite to the cloud adjustment over the ocean (decreasing high cloud and a positive SWcld response; Figs. 13b, d, 14b, d). It indicates that the ocean and land have positive and negative contributions to the global mean cloud adjustment, respectively. Patterns of changes in high cloud and SWcld in the AMIP experiment (Figs. 13g, 14c) do show some similarities to the idealized experiments, and are perhaps generated partly by a circulation change associated with equatorial waves (Matsuno 1966; Gill 1980). The land surface warming at low latitudes enhances convective heating, which forces a dynamical Matsuno-Gill response. Decreasing high cloud over the western coast of the continent (Fig. 14b, d) is collocated with a pair of cyclonic circulation changes straddling the equator, which resembles the Matsuno-Gill pattern (figure not shown). The above feature is also found in the AMIP experiment with some modification, because of meridional asymmetry in the land–sea contrast (Figs. 13g, 14c).
5 Concluding discussions
The analyses based on different timescales indicate that the conventional concept of effective radiative forcing can be largely decomposed into three parts: (1) time-invariant forcing (instantaneous radiative forcing, CO2 physiological forcing, and cloud masking effect); (2) adjustment on a daily timescale (rapid responses to the instantaneous radiative forcing, i.e., stratospheric adjustment, surface warming and anomalous upward motion over land, warming, drying, and cloud decrease in lower troposphere and associated CRE response, strengthened vertical stability in the lower troposphere, suppressed surface heat fluxes and hydrological cycle, and PBL shoaling); and (3) slower adjustment (additional changes in tropospheric temperature, surface heat fluxes, PBL depth, and strength of large-scale atmospheric circulation). The responses of tropospheric temperature, surface heat fluxes, and PBL depth are initially rapid and continue on slower timescales.
The land–sea warming contrast modifies the large-scale atmospheric circulation and cloud amount, which affect the tropospheric adjustment. Anomalous upward motion induced by land surface warming results in increasing high cloud and an associated negative change in SWcld over land, which partially compensates for the positive change in SWcld over the ocean. The effect of the land surface warming depends on continental size, indicating that land size is a regulating factor for tropospheric adjustment in this model. The land warming negatively contributes to the global mean tropospheric cloud adjustment, but the land effect cannot change the sign of the total adjustment. It is suggested that the land–sea warming contrast is just a secondary factor for the tropospheric adjustment.
This is the first application of the TAMIP 4×CO2 experiments to detect the fast response to instantaneous radiative forcing. Most parts of the effective radiative forcing and tropospheric adjustments detected by the TAMIP ensemble are consistent with equilibrium response in the AMIP (fixed SST) 4×CO2 experiment, but they have some discrepancies relative to those estimated by the regression method in the AOGCM 4×CO2 experiment. GW08 compared geographic distributions of tropospheric adjustment estimated by the two methods and revealed some differences, particularly in the tropics. Andrews et al. (2012a) applied the regression and fixed-SST methods to the CMIP5 multi-models, revealing differences mainly due to non-linear responses of TOA radiative budget to global-mean SAT increase over long-term integrations (150 years). They showed that, in some models, effective radiative forcing estimated by the fixed-SST method and change of TOA net radiation in the first year tend to fall above the regression line. They also stated that the largest contributor to the non-linear response is SWcld over the ocean. The non-linear response may exist on decadal and longer timescales, which may be related to delayed sea surface warming, stratocumulus response, and state of the deep ocean (Andrews et al. 2012a). The TAMIP experiment would aid quantitative evaluation of estimated effective radiative forcings among different methods and timescales.
The changes in cloud and stratification associated with tropospheric adjustment shown here are generally consistent with previous studies on the tropospheric adjustment (e.g. Dong et al. 2009; Colman and McAvaney 2011). WLG12 reported that the majority of CMIP3/CFMIP1 models show positive (negative) changes in SWcld (LWcld) and strengthening of vertical tropospheric stability with tropospheric adjustment. In contrast, W12 revealed slight increases in tropical- and global-mean total cloud fractions together with the shoaling of the marine PBL. Different responses of high-, mid- and low-level cloud, and associated CRE from those in other models would be related to differences in the model physical schemes (cloud, shallow cumulus, turbulence, and radiation) and configurations (e.g. GCMs, cloud resolving GCMs). The MIROC5 model used in this study reproduces cloud and water vapor generally well but underestimates high cloud relative to the other CMIP5 models (Jiang et al. 2012). The reproducibility of cloud in the control simulation might be a factor for the tropospheric cloud adjustment to increasing CO2. The finding in this study is based only on a particular CMIP5 model, which should therefore be validated using multi-models under the CFMIP2/CMIP5 umbrella. In particular, changes in atmospheric thermodynamic structure (temperature and humidity), surface heat fluxes, PBL depth, large-scale atmospheric circulations and cloud amounts could be key ingredients for the inter-model spread of ΔSWcld. For detection of evolution processes in response to external forcings, other applications of the TAMIP ensemble (e.g., solar constant, aerosols, patterned-SST anomaly) may also be worth developing. Such approaches may facilitate other groups to interpret evolution processes and possible mechanisms for inter-model spread within forcing and feedback studies.
We would like to acknowledge Hideo Shiogama, Tomoo Ogura, Tokuta Yokohata and Seita Emori at the National Institute for Environmental Studies (NIES), Masakazu Yoshimori and Rei Nobui at the Atmosphere and Ocean Research Institute (AORI), University of Tokyo, and Manabu Abe at the National Institute of Polar Research (NIPR) for providing helpful comments and suggestions. The authors are grateful to two anonymous reviewers for their constructive comments. This work was supported by the Program for Risk Information on Climate Change (PRICC) and Grants-in-Aid 23310014 and 23340137 from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.