Climate simulations and projections with a super-parameterized climate model

Environmental Modelling & Software 60 (2014) 134e152

Contents lists available at ScienceDirect

Environmental Modelling & Software journal homepage: www.elsevier.com/locate/envsoft

Climate simulations and projections with a super-parameterized climate model Cristiana Stan a, b, *, Li Xu b a b

Department of Atmospheric, Oceanic and Earth Sciences, George Mason University, 4400 University Drive, Fairfax, VA 22030, USA Center for Ocean-Land-Atmosphere Studies, George Mason University, Fairfax, VA, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 15 November 2013 Received in revised form 5 June 2014 Accepted 11 June 2014 Available online

The mean climate and its variability are analyzed in a suite of numerical experiments with a fully coupled general circulation model in which subgrid-scale moist convection is explicitly represented through embedded 2D cloud-system resolving models. Control simulations forced by the present day, fixed atmospheric carbon dioxide concentration are conducted using two horizontal resolutions and validated against observations and reanalyses. The mean state simulated by the higher resolution configuration has smaller biases. Climate variability also shows some sensitivity to resolution but not as uniform as in the case of mean state. The interannual and seasonal variability are better represented in the simulation at lower resolution whereas the subseasonal variability is more accurate in the higher resolution simulation. The equilibrium climate sensitivity of the model is estimated from a simulation forced by an abrupt quadrupling of the atmospheric carbon dioxide concentration. The equilibrium climate sensitivity temperature of the model is 2.77
C, and this value is slightly smaller than the mean value (3.37
C) of contemporary models using conventional representation of cloud processes. The climate change simulation forced by the representative concentration pathway 8.5 scenario projects an increase in the frequency of severe droughts over most of the North America. © 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

Keywords: Super-parameterization Global modeling Climate sensitivity Climate change

1. Introduction Variability in the climate system spans a broad range of spatial and temporal scales and results from the interaction between fluctuations due to internal dynamics and external forcing. The processes involved can be thought of as having both deterministic and stochastic components (Palmer, 1999). Cloud processes represent a key component contributing to the variability due to natural internal processes. Therefore, the numerical representation of convective clouds in climate models is expected to have a significant impact on the simulations of all time scales beyond that of individual weather events. In particular, the influence of cloud modeling on climate projections is also an outstanding problem, and cloud feedback is still considered the largest source of uncertainty (Soden and Held, 2006; Bony et al., 2006; Solomon et al., 2007; Webb and Lock, 2012; Brient and Bony, 2012) in the estimation of the climate

* Corresponding author. Department of Atmospheric, Oceanic and Earth Sciences, George Mason University, 4400 University Drive, Fairfax, VA 22030, USA. E-mail address: [email protected] (C. Stan).

sensitivity. Shiogama et al. (2012) showed that a relatively large number of parameters associated with cloud micro- and macrophysics can influence the climate sensitivity of a model. Currently, four major approaches are used in climate models to represent the main effects clouds have on the atmosphere: (i) the conventional parameterizations in which the statistical effects of clouds are described in terms of the properties of the resolved scales (Arakawa, 2004, provides a review on the cumulus parameterization), (ii) the stochastic parameterizations, which are modified versions of the conventional parameterizations to include a stochastic forcing (see Neelin et al., 2008, for a review on stochastic parameterization), (iii) cloud-resolving models in which cumulus convection is explicitly resolved (Miura et al., 2007; Satoh et al., 2008), and (iv) the so-called “super-parameterization” (Grabowski, 2001; Khairoutdinov and Randall, 2001), which bridges the gap between the parameterized and resolved convection (Randall et al., 2003). In the last approach, a large number of individual cloud-resolving models are embedded in the host model, which resolves the large-scale dynamics. Previous studies provide strong evidence that climate simulations are sensitive to the representation of cloud processes. Sensitivity of the simulated seasonal and intraseasonal variability of the

http://dx.doi.org/10.1016/j.envsoft.2014.06.013 1364-8152/© 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

tropical atmosphere to cloud parameterization has been reported in various climate models (Maloney and Hartmann, 2001; Zhu and Hendon, 2010; Zhou et al., 2011; Kim et al., 2012). On longer time scales, Neale et al. (2008) found that changes in the parameterization of deep convection of the Community Climate Model, version 3 (CCSM3), led to an improved simulation of the interan~ o Southern Oscillation nual variability associated with El Nin (ENSO). Stan et al. (2010) showed that by replacing the conventional representation of shallow and deep convective processes by the super-parameterization the ENSO variability simulated by the super-parameterized CCSM3, referred to as SP-CCSM3, became more realistic. The observed seasonal and intraseasonal variability of the tropical atmosphere, poorly represented in CCSM3, was also well captured in the climate simulation produced by SP-CCSM3 (DeMott et al., 2011, 2013; Krishnamurthy et al., 2014). The coupled climate simulations using the superparameterization have also a number of limitations, some of which are related to the approach itself and some are inherent problems of the host model. For example, the eastern subtropical oceans in the SP-CCSM3 climate simulation are affected by warm SST biases, which are related to the lack of stratocumulus clouds in these regions. Dirmeyer et al. (2012) showed that in SP-CCSM3 the 12Z maximum precipitation over the Great Plains is displaced to the northeast of its observed location. The displacement may be caused by the 2D geometry and/or by the east-west orientation of the cloud model. The period of the ENSO variability in SP-CCSM3 is shorter than in observations, and this is a common problem in models with a low resolution in the ocean grid. We conducted new simulations in which the same superparameterization used in Stan et al. (2010) was implemented in an improved version of the host model. In the new simulations, the host model uses a different dynamical core and the horizontal resolution is increased in both the atmosphere and the ocean. In addition, the new simulations are extended for a much longer period of time and some experiments follow the guidelines of the Coupled Model Intercomparison Project Phase 5 (CMIP5, Taylor et al., 2011). The effects of horizontal resolution on the climate simulations produced by global models using explicit representation of convection are less explored. In a case study of an Madden-Julian Oscillation (MJO) event Liu et al. (2009) found that the global cloud-system-resolving Nonhydrostatic Icosahedral Atmospheric Model (NICAM) showed a slight improvement in the high resolution simulation. The purpose of this study is to report on climate simulations and projections with CCSM, version 4 (CCSM4, Gent et al., 2011) in which the parameterization of convective processes is replaced by quasi-independent cloud-scale models. We document the strengths and biases of the super-parameterization approach when applied to climate simulation and projections, provide an interpretation - to first order - of the source of errors, and evaluate the surface warming and changes in hydrological cycle projected by the super-parameterization approach. It is not the intent of this paper to provide a comprehensive comparison of SP-CCSM4 with the conventionally parameterized version of the model. Such a comparison was provided by Stan et al. (2010), Stan (2012), Krishnamurthy et al. (2014) for CCSM3, except for the climate change experiments that were not conducted with SP-CCSM3. However, the super-parameterization has been used in idealized and very short climate change experiments (Blossey et al., 2009; Wyant et al., 2009, 2012), which allowed only the evaluation of the fast cloud response. As pointed out by Wyant et al. (2012) a realistic study of the cloud response to climate change requires longer simulations with coupled models. Section 2 provides a description of the model configuration and numerical experiments conducted with the model. In

135

Sections 4 and 5 we validate the simulated mean climate and its variability against observations and reanalyses. The climate sensitivity of the model is evaluated in Section 6 and the projection of North America drought under a climate change scenario is explored in Section 7. Section 8 summarizes the main findings of the study. 2. Model and experiments description The super-parameterized version of CCSM4 (Gent et al., 2011) was used to produce the simulations presented in this paper. The only difference between the CCSM4 and SP-CCSM4 is the representation of cloud-scale processes. CCSM4 uses the conventional approach in the form of shallow and deep convection parameterization described by Neale et al. (2013), whereas SP-CCSM4 uses a 2D cloud-system resolving model (Grabowski, 2001; Khairoutdinov and Randall, 2001) embedded in each grid column of the atmospheric model as described by Stan et al. (2010). Simulations were carried out at two horizontal resolutions for the atmospheric model: the finite volume grid (Lin, 2004) with 1.9  2.5, referred to as fv19, and 0.9  1.25, referred to as fv09. Each embedded cloud-resolving model (CRM) is two-dimensional, with the independent dimensions being longitude and vertical level. 32 EeW grid points are used for each model, with a fixed dimensional resolution of 4 km in the fv19 configuration, and 3 km in the fv09 configuration. These resolutions span the equator, but at higher latitudes neighboring CRMs overlap in longitude. However, the CRMs do not share among them information from the host model and their domain is periodic. The CRMs share the same 28-vertical levels from the surface with the dynamical core, which has 30 levels. The time step of the cloud model is not fixed and cannot exceed 20 s in fv19 and 15 s in fv09. A comprehensive description of the CRM model equations is presented in Khairoutdinov and Randall (2003). In summary, each CRM solves the anelastic equations of motion in the Cartesian coordinates (x, z) using the finite-difference representation on a fully staggered Arakawa C-grid. The model includes prognostic equations for liquid/ice moist static energy, precipitable hydrometeors, and non-precipitable water. The subgrid scale parameterization uses a first-order closure scheme and the microphysics parameterization is based on a one moment scheme. The CRMs runs continuously and exchange information with the large-scale model at the host model time step. The CRMs states are relaxed toward the large-scale forcing and the CRMs provide the host model with convective temperature and moisture tendencies. The tendencies represent the zonal averages of the CRM domain. The radiative scheme is handled by the host model and only the cloud-radiation interaction takes place at the CRM scale. The CRMs do not resolve any cloud-aerosol-precipitation interactions. The horizontal resolution of the ocean model grid is 1
, with 60 levels in the vertical, and these parameters are consistent for all simulations. The resolution of the land model matches the resolution of the atmospheric model, and the sea-ice model uses the same resolution as the ocean model. The initial conditions of all simulations were taken from the CCSM4 twentieth-century runs. The fv09 runs were initialized in January 2005 and the fv19 runs in January 2006. Two control simulations with present-day constant forcing (represented by 2000 green house gas concentrations) were run for 100 years. They are referred to as CTL-fv19 and CTL-fv09, respectively. In this paper, the last 50 years of the control runs will be analyzed. Note that the two initial states could explain to some degree the differences in the simulations, especially with the short spin-up time considered here which is not enough to allow the deep ocean to adjust to the change in forcings. Aerosol

136

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

Table 1 SP-CCSM4 computational cost. The number of simulated years is denoted by Myears. The spatial resolution of the CRM is 4 km in the fv19 configuration and 3 km in the fv09. fv19/4 km

fv09/3 km

Number of cores

Myears/wall clock day

Number of cores

Myears/ wall-clock day

384 1152 2112 4224 8192 16,512

0.59 2.18 3.52 5.65 8.41 10.54

384 1152 2112 4224 8192 16512

0.16 0.59 1.10 1.86 2.57 2.60

concentrations were prescribed as in Gent et al. (2011). The land model includes an active carbon-nitrogen cycle. A climate sensitivity experiment was integrated with the fv19 configuration and a climate change projection was conducted with the fv09 configuration. In the climate sensitivity simulation, referred to as 4CO2-fv19, the present day carbon dioxide (CO2) concentration (368.9ppmv) was abruptly quadrupled at the beginning of the simulation. In the climate change projection run, referred to as RCP85-fv09, the external forcing followed the mitigation scenario RCP8.5 (Moss et al., 2010; Taylor et al., 2011). The climate sensitivity experiment was integrated for 150 years. The external forcing starts one year after the initialization, i.e., in January 2006, and the projection run ends at the end of year 2100. Table 1 shows the computational cost of the model expressed as model years simulated per wall-clock day. The cost estimation is relevant to the Cray peta-flop systems such as XT5 or XE6. 3. Data and model evaluation methods Model simulations are compared with reanalyses and satellitebased and in-situ observations, which are considered the “truth” and referred to as observations. The datasets are routinely used in studies focused on the evaluation of climate models. Here, in the direct comparisons the model grid is interpolated to the observation grid. The model evaluation methods consist of general metrics (e.g., Jakeman et al., 2006; Bennett et al., 2013) and specific diagnostics designed for climate simulations. The general metrics include quantitative measures such as the mean, bias, and root mean square error (RMSE). The specific diagnostics are designed to extract the model ability of simulating mechanisms responsible for the observed phenomena (i.e., ENSO, seasonal and intraseasonal variability of the tropical atmosphere). Visual inspection is then

Table 2 Global mean values of energy balance, hydrology and non-radiative surface fluxes. Variable Annual mean budget (W/m2) TOA Surface

CTL-fv19 0.99 2.30

CTL-fv09

Observationsa

1.35 2.50

0.57 1.20

TOA longwave radiation (W/m2) All sky 231.1 Clear sky 263.1

232.0 264.9

239.6 265.9

TOA absorbed solar radiation (W/m2) All sky 232.1 Clear sky 287.4

233.2 287.8

240.2 287.4

Cloud radiative forcing (W/m2) Longwave 32.0 Shortwave 55.3

32.9 54.8

26.3 47.2

Cloud amount (%) Total High-level Middle-level Low-level Cloud water path (mm) Precipitable water (mm) Precipitation (mm/day)

53.8 26.3 19.0 35.3 0.112 23.8 2.82

Surface longwave radiation (W/m2) All sky 53.2 Clear sky 85.2 Net surface shortwave radiation (W/m2) All sky 158.2 Clear sky 218.6 Latent heat flux (W/m2) Sensible heat flux (W/m2) a

81.7 21.0

54.2 27.9 17.4 34.5 0.108 24.5 2.89

68.0 30.7 28.4 50.3 0.112 24.6 2.67

52.4 84.6

54.5 83.6

159.1 218.2

163.0 192.1

84.0 20.5

87.9 19.4

Xu and Cheng (2013).

applied to compare the simulated variability with the observed counterpart. 4. Control simulation climatology 4.1. Top of the atmosphere and surface radiation budgets Fig. 1a compares the annual cycle of the global mean residual energy (the imbalance between the absorbed solar radiation, ASR, and outgoing longwave radiation, OLR) at the top-of-atmosphere (TOA) in the control simulations to the satellite-based data products, CERES (2000e2012) and ERBE (1985e1988). Both model configurations show reasonable agreement with observations for both amplitude and phase of the seasonal variation. It is interesting to note the better agreement between the models and observations

Fig. 1. The annual cycle of the global mean (a) net radiation at the top-of-atmosphere (TOA) and (b) net short and longwave radiation at the surface. Numbers in parentheses denote the annual mean, which represents the radiative imbalance.

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

during January through June than in the second half of the year, a feature that is also present when the observational datasets are compared against each other. The difference between the observations comes from the uncertainties in the calibration of the retrieval algorithms (Loeb et al., 2009). An inspection of Table 2, which provides a partitioning of the global and annual mean budget into its terms, indicates that in the model both the all-sky OLR and ASR at the TOA are underestimated. The source of errors can be either the model's radiation scheme, the simulation of clouds, or the interaction between the two. The clearsky radiative fluxes at the TOA and the atmospheric precipitable water are within the measurement uncertainty of the observations (4.2 W/m2, Loeb et al., 2009; 1.9 mm, Randel et al., 1996) suggesting

137

that the OLR and ASR errors are mainly due to deficiencies in the simulation of cloud properties. The total cloud amount in the simulations is underestimated by 15%. In the model, the warming due to long wave cloud radiative forcing is stronger than observed, and this is consistent with the smaller values of the all-sky OLR. In the shortwave range, the simulated cooling produced by cloud radiative forcing is stronger than observed. The increase in the shortwave component is consistent with the decrease in the cloud amount, a behavior noticed in the observations (Cess et al., 1992). The behavior of the longwave component simulated by the model is not consistent with observations where an increase in the cloud amount is accompanied by an increase in the longwave component (Cess

Fig. 2. Seasonal mean precipitation and root mean square error (RMSE) between 60
S and 60
N. (a) Observed DecembereJanuaryeFebruary (DJF) 1979e2010 from GPCP. (b) As in (a) but for JuneeJulyeAugust. (c) DJF mean from CTL-fv19. (d) As in (c) but for JJA. (e) DJF mean from CTL-fv09. (g) As in (e) but for JJA. Units are in mm/day.

138

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

et al., 1992). These suggest that other factors such as the cloud vertical structure is deficiently represented in the model; the underestimated cloud amount for all cloud types (Table 2) supports this argument. The annual cycle of the radiative terms of the surface budget are shown in Fig. 1b for the model and CERES (2000e2012). Although the model reproduces the pattern of the surface radiation reasonably well, there are noticeable differences. The amplitude of the simulated downward radiation is slightly underestimated in both configurations of the model. Table 2 also provides information for the radiation budget at the surface. The net surface longwave radiation for both all- and clear-sky is within the measurement uncertainty of the observations. Similar to the TOA, at the surface, biases in the shortwave portion of the spectrum are larger than in the longwave band. The clear-sky net surface shortwave radiation is underestimated by more than 10 W/m2, suggesting a poor representation of surface albedo. 4.2. Surface seasonal climatology Fig. 2 compares the seasonal mean of the simulated precipitation against the observed climatology of GPCP (Huffman et al., 1997) for boreal winter and summer simulations. Both configurations of the model simulate the large-scale features of the observed distribution, with global errors less than 2 mm/day. In the simulation with higher resolution, CTL-fv09, the global RMSE is smaller than in the CTL-fv19 configuration for both seasons. At regional scale, the biases also tend to be smaller in the CTL-fv09. Over land, this result is expected and is consistent with previous studies using atmospheric GCMs and coupled GCMs (Wehner et al., 2010; Gent et al., 2010). Over the ITCZ region, we note that biases in the CTL-fv19 configuration are larger than in SP-CCSM3 (Fig. 3c and d in Stan et al., 2010). This suggests that one of the possible source of errors for the spurious ITCZ south of the equator in DJF may be the numerical representation of dry dynamics, becuase the nominal horizontal resolution in the atmosphere model of the CTL-fv19 configuration is equivalent to the nominal horizontal resolution (T42) of the host atmosphere GCM in SP-CCSM3 (Williamson, 2008). Wei et al. (2011) also noticed a degradation in the simulation of intraseasonal variability of the East Asian summer monsoon in the finite volume dynamical core compared to the semi-Lagrangian representation of the dry dynamics, in the atmosphere model of CCSM3. The comparison of the double-ITCZ bias in the low- and highresolution simulations shows a smaller error in the high-

resolution simulation (CTL-fv09), and this suggests that resolution may also mitigate the double-ITCZ problem. This result differs from that of Gent et al. (2010), who found that increasing the resolution alone did not improve “at all” the double-ITCZ bias of CCSM4. The impact we see in the super-parameterization approach could be explained by the smoother representation of scale interactions e in the sense that the gap between the scales resolved in the model is smaller in the high resolution simulation, and the improved boundary conditions for the CRMs, which do not include varying boundary conditions. Fig. 3 shows the annual biases (defined as the model deviation from the observation) in the two control simulations CTL-fv19 and CTL-fv09, respectively. In both configurations there is an excess of precipitation over most of the tropical oceans, except over the eastern Indian Ocean. Over the land, the biases are relatively small except over the South American continent; they tend to be larger in the fv19 configuration. The format of Fig. 3 follows Fig. 5 of Gent et al. (2011), which provides the biases in CCSM4. Although the overall pattern of the biases in SP-CCSM4 and CCSM4 are similar, there are also noticeable differences. The double ITCZ problem in the super-parameterized model is not as strong as in the conventional model. In the western Indian Ocean the errors in the SPCCSM4 model are smaller than in CCSM4. The region with largest biases in SP-CCSM4, South America, is the only place where the biases in CCSM4 are smaller than in SP-CCSM4. The annual mean simulated sea-surface temperature fields (SSTs) are compared to the Hadley Center SST version 2 (HadSST2) data set (Rayner et al., 2006) in Fig. 4. The simulated SSTs tend to be colder than observations in the Tropics and warmer almost everywhere else. The warm biases off the western coasts suggest a deficit of low-level cloud over these regions. Whereas in the Tropics, the higher resolution configuration simulates warmer SSTs than the lower resolution, the SSTs in the eastern subtropical oceans are colder in the CTL-fv09 configuration than in the CTLfv19. However, the differences between the two configurations are small and their locations indicate they could be attributed to the wind-driven circulation and its effect on parameterized stratocumulus clouds along the western coasts of the Americas and Africa. Panels (d) and (e) in Fig. 4 are designed to be directly comparable to the SST biases of CCSM4, which are shown in Fig. 1 of Gent et al. (2011). The cold bias in the Labrador basin, very large in CCSM4, is reduced in the super-parameterized version of the model. In SP-CCSM4, the cold bias in the western Indian Ocean and eastern tropical Pacific improved when going from the fv19 to the

Fig. 3. Annual mean precipitation bias of (a) CTL-fv19 and (b) CTL-fv09 relative to observed 1979e2010 from GPCP. Units are in mm/day.

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

139

Fig. 4. Annual mean sea surface temperature for the (a) observed 1959e2002 Hadley SST, (b) CTL-fv19, and (c) CTL-fv19. Mean differences from the Hadlley SST for (d) CTL-fv19 and (e) CTL-fv09. (f) Annual mean sea surface temperature difference between CTL-fv19 and CTL-fv09. Stippling denotes 95% significance level. Units are in degree Celsius.

fv09; in CCSM4 this bias did not change with the increase in resolution. SST biases have a significant impact on the simulation of seasonal variability associated with the monsoon circulations (DeMott et al., 2013; Prodhomme et al., 2014) and in Fig. 5 we show the SST biases simulated during DJF and JJA. In the fv19 there is a cold SST bias during the boreal summer over the Indian Ocean, whereas in the fv09 there is a warm bias near the equator. Another region with a change in the bias sign is the Pacific Ocean. The western equatorial Pacific is dominated by cold biases throughout the year whereas in the eastern Pacific the positive bias during DJF becomes negative during JJA, in both configurations of the model. This result implies a misrepresentation of the seasonal cycle of the eastern Pacific cold tongue, which is controlled by the meridional component of the surface wind (Fu and Wang, 2001).

4.3. Atlantic meridional overturning circulation The representation of the structure of the Atlantic meridional overturning circulation (AMOC) has been the focus of recent numerical simulations (Griffies et al., 2011; Huang et al., 2012; Danabasoglu et al., 2012; Xu et al., 2012). The AMOC plays a significant role in the climate system, as it transports heat poleward. A measure of the AMOC is the meridional transport massstreamfunction, which at any given latitude quantifies the vertically integrated amount of water moving meridionally across that latitude and is expressed in units of Sverdrups (Sv ¼ 106 m3/s). The strength of the AMOC refers to the maximum value of the streamfunction. In observations, the AMOC strength can vary between 15 Sv (Ganachaud and Wunsch, 2000) and 18 Sv (Talley et al., 2003). Here, we compare the modeled mean meridional

140

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

Fig. 5. Seasonal mean sea surface temperature bias and root mean square error (RMSE). (a) DecembereJanuaryeFebruary (DJF) from CTL-fv19. (b) As in (a) but for JuneeJulyeAugust. (c) DJF mean from CTL-fv09. (d) As in (c) but for JJA. Units are in degree Celsius.

Fig. 6. Annual mean Atlantic meridional overturning stream function. (a) Reanalysis 1979e2004 from NCEP GODAS. (b) CTL-fv19. (c) CTL-fv09. The contour interval is 2 Sv and zero contours are omitted. Positive values denote clockwise circulation. (d) CTL-fv19 minus CTL-fv09. The contour interval is 0.5 Sv and zero contours are omitted. Boxes in (b) and (c) show the region used in the MOC index calculation.

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

141

13

MOC Index, Sv

12

11

10

CTL−fv09

9

CTL−fv09 smooth CTLfv−19 CTL−fv19 smooth

8

0

10

20

30

40

50

60

70

80

90

100

year Fig. 7. MOC index over the area marked by a box in Fig. 6. The dark color lines depict 5-year running average value of the index.

streamfunction against an estimation (Huang et al., 2012) based on the ocean reanalysis GODAS (Behringer and Xue, 2004) (Fig. 6). The simulated North Atlantic Deep Water (NADW) cell in the upper 3 km is similar to the reanalysis with the maximum near 35
N and the sinking branch near 60
N. However, the transport is stronger than observed (~20 Sv) in both configurations but weaker than in CCSM4 (~24 Sv, Gent et al., 2011). The Antarctic Bottom Water (ABW) cell in both simulations is much weaker than in GODAS and the maximum transport blue is displaced in the Southern Hemisphere. The AMOC simulation in the CTL-fv09 is closer to GODAS than the simulation with a lower resolution (fv19) in the atmosphere model, which is weaker consistent with the warmer SSTs in the Atlantic basin. The shallow overturning subtropical circulations in the few hundreds meters below the surface are consistent with GODAS, and are similar between the two configurations of the model. Small biases in the estimation of the AMOC strength have a large impact on the oceanic heat transport due to a large sensitivity of the heat transport to changes in the AMOC, estimated to be ~0.06 PW/Sv (Srokosz et al., 2012). The difference between the model configurations emphasizes the strong influence of the atmospheric forcing in driving the AMOC transport (see also Klinger et al., 2003). Fig. 7 shows the time series of MOC index, defined as the annual mean AMOC maximum transport between 20
Ne50
N and 0.5e2 km (Danabasoglu et al., 2012), for the two model configurations. Both time series suggest a slight increase of the AMOC maximum transport over the first hundred years. This drift may be just a residual of the “coupling shock” (Gupta et al., 2012) and a much longer simulation is required to establish whether the MOC in the model has a real trend. 5. Model simulated variability 5.1. ENSO simulation The interannual variability of the tropical Pacific Ocean is ~ o3.4 index. Fig. 8 shows the time series of analyzed using the Nin the index from the CTL-fv19 and CTL-fv09 simulations along with their power spectra. The amplitude of the SST anomalies simulated by both configurations are weaker than the anomalies derived from

a 50-year period of Hadley SST. The CTL-fv19 spectrum has a period of 48 months whereas the main spectral peak of the CTL-fv09 simulation is located at 41 months; however all model and observed spectra have very broad peaks. Compared to CCSM4 (Fig.8 in Gent et al. (2011) or Fig. 4 in Deser et al. (2011)), the period simulated by the CTL-fv09 configuration is shorter and the variability is much smaller. In fact, while the variability in the SPCCSM4 is half of the natural variability, that in CCSM4 is three times larger than in the observations. The spatial structure of the simulated SST anomaly associated with the ENSO variability is compared in Fig. 9 against the observations. The regression pattern of the SST anomalies with the ~ o3.4 index in the model simulations agrees with the observaNin tions. Apparently the slight differences between the two models are large enough to alter the period. The zonal extent of the cold tongue in the western Pacific in the CTL-fv19 simulation is in good agreement with the observations. In the CTL-fv09 simulation the cold tongue extends farther west but not as far as in the CCSM4 (Fig. 9b in Gent et al. (2011)). This error alone cannot explain the shorter period in the CTL-fv09 configuration. Another key component controlling the ENSO period is the pattern of the wind-stress (Kirtman and Schneider, 1996; Kirtman, 1997; An and Kang, 2001; Capotondi et al., 2006; Neale et al., 2008), and we will explore some features of the simulated wind stress in connection with the ENSO variability. Fig. 10 shows the regression pattern of the zonal component of the wind stress anomaly against ~ o3.4 index, and its zonal average in the Pacific basin. The the Nin observational data set used for the zonal wind stress is the 1950e1995 SODA analysis (Carton et al., 2000a,b). In the Northern Hemisphere, the zonal wind stress anomaly simulated by the CTLfv09 has a merdional extent comparable to the observations whereas south of the equator does not go beyond 5
S. In the Northern Hemisphere, the zonal wind stress anomaly simulated by the CTL-fv09 has a meridional extent comparable to the observations whereas the simulated pattern is damped south of 5
S. In CTLfv19 the meridional extent of the zonal wind stress anomaly pattern is in better agreement with observations. The amplitude of the regression coefficients is larger in both model simulations than in observations. A broader eastward extend of the positive correlation is also noticed in the model simulations.

142

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

Anomaly (oC)

Nino3.4: HadSST 4

a

2 0 −2 1960

1965

1970

1975

1980

1985

1990

1995

2000

Anomaly (oC)

Nino3.4: CTL−fv19 4

b

2 0 −2 2050

2055

2060

2065

2070

2075

2080

2085

2090

Anomaly (oC)

Nino3.4: CTL−fv09 4

c

2 0 −2 2050

2055

2060

2065

2070

2075

2080

2085

2090

25

d

20

HadSST (48)

Variance (C2)

CTL−fv19 (48) CTL−fv09 (41) 15

Red Noise

10

5

0

48

24

12

6

Period (months) ~ o 3.4 index and the associated power spectrum. (a) Observations 1959e2002 from Hadley SST. (b) CTL-fv19, and (c) CTL-fv09. (d) Power spectra Fig. 8. Time series of the Nin computed for the time series in (aec). The blue dashed line corresponds to red noise power spectrum. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

5.2. Tropical seasonal and intraseasonal variability The mean seasonal cycle of precipitation for the boreal summer monsoon systems is depicted by latitude-time sections of pentad means in Fig. 11. In the south Asian monsoon region (70
Ee95
E) models capture the observed “onset” at the beginning of the boreal summer (around June 1st) and the location of maximum precipitation between 20
N and 30
N. The secondary maximum located at 5
S and associated with the equatorial convection is not well

simulated by either model configuration. In both configurations the model produces a much stronger than observed rain band north of the equator (Fig. 10d and g). The seasonal march of the East Asian-Western North Pacific monsoon system (115
Ee140
E) shows a tropical rain band located north of the equator from June through November and a subtropical rain band that occurs from mid-May to early July migrating northward from 20
N to 30
N (Fig. 10b). These features are captured by both simulations (Fig. 11e and h) although the

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

143

~ o3.4 index. (a) 1959e2002 Hadley SST. (b) CTL-fv19, and (c) CTL-fv09. Fig. 9. Regression map of the sea surface temperature monthly anomaly with the Nin

maximum intensity over the Western North Pacific (East Asian) sector is stronger (weaker) than observed. In the CTL-fv09 configuration the dry zone between the subtropical and tropical rain is not as distinct as in observations (Fig. 11h). The seasonal cycle of the West African monsoon (WAM; 11
We10
E) system (Fig. 11c) consists of a northward propagating rain band during May-June followed by a more rapid southward migration from mid-August through the end of September (Thorncroft et al., 2011). The southerly migration is accompanied by a “jump”in the location of maximum precipitation from the

Guinean coast to the Sahel at the end of June (Sultan and Janicot,  et al., 2002; Hagos and Cook, 2000; Sultan et al., 2003; Le Barbe 2007). Model simulations have a seasonal cycle with much stronger amplitude and a symmetric meridional migration (Fig. 11f). In the CTL-fv19 simulation there is a break period at 5
N but there is no jump in the location of maximum precipitation. The higher resolution simulation misses the break phase and also the jump (Fig. 11i). The errors in the simulation of the evolution of precipitation over West Africa are also present in the parameterized version of the model, CCSM4. Cook et al. (2012) found that in the

144

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

~ o3.4 index and zonal average (120
Ee100
W) of the regression pattern. (a) Observations from 1950 to Fig. 10. Regression map of zonal wind stress monthly anomaly with the Nin 1995 SODA analysis (Carton et al., 2000a,b) and Hadley SST. (b) CTL-fv19. (c) CTL-fv09. Units are in dyn/cm2 K and the contour interval is 0.03.

CCSM4 model (their Fig. 1) the location of maximum precipitation in January is displaced south of the equator compared to the observed position. In the SP-CCSM4 model the maximum precipitation over the coast of Guinea persists from March through June consistent with observations, whereas in CCSM4 the rainfall band is located south of the equator during this period. The intraseasonal variability of the tropical atmosphere is evaluated using the WheelereKiladis diagrams (Wheeler and Kiladis, 1999) based on daily OLR averaged over the 15
S-15
N latitudinal belt. Fig. 12 shows the wavenumber-frequency spectra of the symmetric and antisymmetric components of the simulated OLR and estimates of the NOAA OLR (Liebmann and Smith,

1996). Each power spectrum is divided by the corresponding background spectrum, computed using the same smoothing algorithm. In the symmetric component, both simulations show a robust MJO signal, the maximum power in CTL-fv09 being slightly higher and closer to the observed value. In the simulations, the Kelvin waves signal is weaker than in the observations, especially at higher wavenumbers. The observed clear separation of the MJO from the Kelvin wave signal is not simulated. Westward propagating equatorial Rossby waves and synoptic-scale disturbances tend to have a stronger signal in the simulations than in observations.

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

145

Fig. 11. Mean annual cycle of precipitation vs latitude for (left) the Indian Ocean (70
Ee95
E), (middle) East Asia and western Pacific Ocean (115
Ee140
E), and (right) west Africa (10
We10
E). (a)e(c) Observed climatology 1979e2010 from GPCP. (d)e(f) Climatology from CTL-fv19. (g)e(i) Climatology from CTL-fv09. The vertical line in each panel denotes onset of the observed monsoon rainfall. The horizontal line in each of the middle column panels draws the delimitation between the East Asian and Western Pacific monsoonal regions.

In the antisymmetric component again the simulations capture the observed features of the convectively coupled mixed-Rossby gravity (MRG) waves. However, the signal is weaker and in the CTL-fv19 configuration is confined mostly to the westward propagating wavenumbers. The intraseasonal variability simulated by the SP-CCSM4 is consistent with the accuracy documented in the previous version of the model, SP-CCSM3 (e.g., Stan et al., 2010; DeMott et al.,

2011, 2013). The union between the MJO and Kelvin wave signal was not present in the SP-CCSM3 version and again we speculate that, at least for the CTL-fv19 configuration, the dynamical core could be responsible for this spurious problem. In the CCSM4 the Kelvin wave activity is stronger than in observations and the MJO associated variability is spread over a broader frequency domain than observed (Subramanian et al., 2011).

Fig. 12. WheelereKiladis zonal wavenumber-frequency power spectrum for (left) equatorially symmetric and (right) equatorially antisymmetric OLR daily anomalies. (a)e(b) Observations 1983e2008 from NOAA. (c)e(d) CTL-fv19. (e)e(f) CTL-fv09.

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

147

Fig. 13. (a) Time series of the change in the annual mean, global mean surface temperature (red curve) and the change in the annual mean, global mean net radiation at the TOA (blue curve). (b) Scatter plot of the change in the annual mean, global mean net radiation at the top of the atmosphere vs. the change in the annual, global mean surface temperature. The black curve denotes the least-square regression line with the slope of 1.3 W/m2 K and intercept 7.2 W/m2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

6. Climate sensitivity The response of the model simulated climate to a given forcing is investigated in a simulation in which the atmospheric CO2 concentration used in the CTL-fv19 run is quadrupled. Fig. 13a shows the time series of the change in the global and annual mean surface temperature and the net radiation at the TOA in response to the increased atmospheric CO2 concentration. The global mean surface temperature increases by one degree in the first year and stabilizes after about 40 years at a 4 degree increase, with small fluctuations about this value. The change in the net radiative flux at TOA is about 6.5 W/m2 after the first year and like the surface temperature decreases exponentially to about 2 W/ m 2. The climate sensitivity of the model is estimated following the linear regression method of Gregory et al. (2004). In this method, the net radiative change (N) due to a climate forcing (F) is proportional to the global average, annual mean surface temperature change (DT), i.e. N ¼ F  aDT, where a is the climate response or feedback parameter. Fig. 13b shows the variation of the global average net radiative flux at the TOA with the global average, annual mean surface temperature and their linear fit. The regression slope gives the climate response or the feedback parameter of

the climate system simulated by the model. The equilibrium temperature sensitivity or climate sensitivity (ECS), defined as the equilibrium DT in response to an abrupt doubling of the atmospheric CO2 concentration, is estimated as in Andrews et al. (2012), ECS ¼ DTeqm/2 ¼ F/2a. For the SP-CCSM4 model with the fv19 configuration, the climate feedback parameter a ¼ 1.3 W/m2 K and the equilibrium temperature sensitivity ECS ¼ 2.77
C. This value is slightly smaller than the mean value of the generation of GCMs in CMIP3 (3.2
C, Randall et al., 2007) and CMIP5 (3.37
C, Andrews et al., 2012). The ECS of the SP-CCSM4 model is closer to the mean value of 2.3
C estimated from temperature reconstructions by Schmittner et al. (2011). The ECS of the atmospheric component of CCSM4 coupled to a slab ocean model is 3.2
C (Gettelman et al., 2012). According to Sherwood et al. (2014), SP-CCSM4 falls in the category of models of low-sensitivity (ECS <3.0). Models with ECS >3.5 are identified as models of highsensitivity. Climate sensitivity of the global surface temperature and the climate feedback parameter depend on several feedbacks such as water vapor, cloud and radiation, surface and cloud albedo, and their interactions. It is beyond the scope of this paper to investigate each of the feedbacks. Their contributions will be reported in subsequent papers.

Fig. 14. (a) Surface temperature change at the end of 21st century, calculated as mean of (2091e2100) minus mean of (2005e2014). (b) Same as in Fig. 13a but for precipitation (%).

148

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

7. Future projections under the RCP8.5 mitigation scenario The greenhouse gas emissions and concentrations in the RCP8.5 scenario increase significantly with time, leading to a radiative forcing of 8.5 W/m2 in 2100. In this type of simulation the climate change is transitory and the final state is not an equilibrium state, as it was in the climate sensitivity experiment. Fig. 13a shows the surface temperature difference between the present day conditions (2005e2014) and the end of 21st century (2091e2100). Consistent with CCSM4 (Meehl et al., 2012), SPCCSM4 shows a significant increase of surface temperature at the end of the century. There is only an exception: in a small region in the North Atlantic Ocean, around the southern tip of Greenland, the temperature difference is negative (about 2 degrees). This feature

is not present in the CCSM4 simulation. The model with conventional representation of convection projects a small increase (~1 degree) in the surface temperature of the Labrador Sea at the end of 21st century. The amplitude of temperature changes over the land is larger than over the ocean consistent with CCSM4. The global average warming for the last 10 years of the 21st century relative to the period 2005e2014 period (in which the CO2 concentration is assumed to be constant) projected by SP-CCSM4 is 3.22 degrees, which is very close to the CCSM4 projection of 3.53 degrees (Meehl et al., 2012). The distribution of precipitation change (Fig. 14b) shows much more regional variability than the temperature field. Over land, the projections show reduced precipitations in the tropical regions of the American continents and in the subtropics over other regions.

Fig. 15. Frequency of severe 12-months (SPI12) and 24-month (SPI24) drought during present day (a)e(b) and projections under the climate change mitigation scenario (c)e(d).

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

The Atlantic Ocean shows a symmetric pattern with wet regions in the high and middle latitudes separated by a dry zone in the subtropics. The distribution of wet and dry zone in the Pacific and Indian Oceans are not symmetric about the equator. The Indian Ocean is dominated by an increase in precipitation north of the equator and reduced precipitation south of 15
S. The equatorial Pacific is the region showing the largest change in precipitation, one that is positive. The large-scale pattern of precipitation changes projected by the SP-CCSM4 is consistent with the CCSM4 simulation (Fig. 12f in Meehl et al., 2012) with some regional differences such as over the central America, Indian Peninsula, and middle East. Extreme events such as heat waves or droughts, which may or may not be related to climate change, are important climate variability events with significant socio-economic impact. Fig. 15 shows the frequency of severe droughts over North America based on the standardized precipitation index (SPI, McKee et al. (1995)). The 12month and 24-month SPI is computed for the CTL-fv09 and RCPfv09 simulations, and the percentage of days spent in severe drought conditions (SPI <1.3) are plotted. The spatial pattern of drought conditions over continental Unites States and Mexico is consistent across the two time scales and this is the case for both simulations. Areas with increased drought conditions in the CTLfv09 simulation are consistent with regions observed to have experienced moderate and extreme droughts during 1950e1999 period, as analyzed by Wehner et al. (2010). In the projections, the frequency of severe droughts is comparable to the present day conditions over most regions. Fig. 16 shows the difference between the frequency of severe droughts between the climate change scenario and the present day simulation for 12-month and 24month time scales. In the warmer mean climate most regions experience an increase in the frequency of severe droughts in the range of 2e4%. On the 12-month time scale the northwest coast of United States and Mexico show the largest increase in the frequency. On the 24-month time scale, the area along the northwestern coast of U.S. with increase frequency shrinks, whereas in Mexico the area with more frequent droughts expands. The frequency is reduced on the 24-month time scale along the east coast,

149

the Central US and the interior Plains of Canada. In observations, precipitation changes over this region show inter- and multidecadal variations driven by the SST patterns over the ocean. In this analysis these variations are not separated from the signal. 8. Summary and conclusions This paper describes the super-parameterized version of the CCSM4 and presents results from present-day control simulations at two horizontal resolutions, a CO2 sensitivity simulation in which the concentration (368.9 ppmv) is abruptly quadrupled and a climate projection for the RCP8.5 mitigation scenario. The simulated mean climate is in agreement with the observations and shows some dependence on the horizontal resolution. The mean state simulated by the higher resolution configuration has smaller biases. Additional experiments are necessary to understand the relative importance of horizontal resolution of the large-scale model and CRM, as they both vary in the current experiment. The interannual variability of the Pacific SST associated with ENSO is better simulated by the model with coarser horizontal resolution. However, both model configurations produce SST anomalies weaker than observed. The annual cycle of precipitation associated with the summer global monsoon circulations is well captured by the model, and the errors in the CTL-fv19 (lower resolution) configuration are smaller than in the CTL-fv09 configuration. The intraseasonal variability associated with the MJO and equatorial waves are realistically simulated by the model with a slightly better representation in the CTL-fv09 (higher resolution) configuration. The equilibrium climate sensitivity of the model is in the middle of the range simulated by other models and close to the mean value estimated from temperature reconstructions. The projection of temperature and precipitation under the RCP8.5 mitigation scenario shows a global increase of surface temperature, except in a small region around the southern tip of

Fig. 16. Change in the frequency of (a) 12-month severe droughts and (b) 24-month severe droughts. The change is defined as the difference between conditions at the end of the 21st century and present day.

150

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

Greenland, and increased regional variability in the rainfall distribution. Progress in understanding climate variability and change depends on the accuracy of complex climate models, which are rich in physical processes occurring on scales that are small in comparison with that of the model and are represented through parameterizations. An obvious effect of the statistical nature of parameterizations is the uncertainty of the answer we get when using models for simulating the climate. In particular, cloud processes represent a large category of missing processes that require parameterizations to be used for their representation. In conventional models, subgrid-scale moist convection parameterization has four components: cloud microphysics, macrophysics or cloud dynamics, radiation and turbulence. In the super-parameterization approach, the macrophysics is explicitly resolved, therefore the number of sources of uncertainties is reduced. This is the case of an improvement of a process already represented in the model and therefore the number of degrees of freedom in the model does not change. Validation of model simulations against observations shows that SP-CCSM4 still has significant biases, which may yet be attributed to the representation of cloud processes. The other parameterizations required for the complete description of cloud processes can also introduce errors and Xu and Cheng (2013) showed that a more complex scheme for the representation of boundary layer turbulence can improve the representation of low clouds. The results reported here are intended to stimulate the interest for other studies which arguably could benefit from the realistic simulation of the observed features of the climate system and a slightly reduced uncertainty associated with the representation of cloud processes offered by the SP-CCSM4 model. Acknowledgments We thank all the scientists and software engineers who contributed to the development of the CCSM4 and the superparameterization. This work has been supported by the Regional and Global Climate Modeling Program funded by U.S. Department of Energy, Office of Science under award number SC0006722 and by the National Science Foundation and Technology Center for Multiscale Modeling of Atmospheric Processes managed by Colorado State University under cooperative agreement No. ATM-0425247. This research used resources of the National Energy Research Scientific Computing Center, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC0205CH11231 and the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation grant number OCI-1053575. The authors are grateful to David Straus for his suggested improvements of the paper and thank the three anonymous reviewers for their critical and thoughtful comments. References An, S.I., Kang, I.S., 2001. Sensitivity of the equatorial airesea coupled system to the zonal phase difference between SST and wind stress. Adv. Atmos. Sci. 18, 155e165. http://dx.doi.org/10.1007/s00376-001-0010-8. Andrews, T., Gregory, J.M., Webb, M.J., Taylor, K.E., 2012. Forcing, feedbacks and climate sensitivity in CMIP5 coupled atmosphere-ocean climate models. Geophys. Res. Lett. 39, L09712. http://dx.doi.org/10.1029/2012GL051607. Arakawa, A., 2004. The cumulus parameterization problem: past, present, and future. J. Clim. 17, 2493e2525. http://dx.doi.org/10.1175/1520-0442(2004)017< 2493:RATCPP> 2.0.CO;2. Behringer, D.W., Xue, Y., 2004. Evaluation of the global ocean data assimilation system at NCEP: the Pacific Ocean. In: Eighth Symposium on Integrated Observing and Assimilation Systems for Atmosphere, Oceans, and Land Surface,

AMS 84th Annual Meeting, Washington State Convention and Trade Center, Seattle, Washington. Bennett, N.D., Croke, B.F.W., Guariso, G., Guillaume, J.H.A., Hamilton, S.H., Jakeman, A.J., Marsili-Libelli, S., Newham, L.T.H., Norton, J.P., Perrin, C., Pierce, S.A., Robson, B., Seppelt, R., Voinov, A.A., Fath, B.D., Andreassian, V., 2013. Characterising performance of environmental models. Environ. Model. Softw. 40, 1e20. http://dx.doi.org/10.1016/j.envsoft.2012.09.011. Blossey, P.N., Bretherton, C.S., Wyant, M.C., 2009. Subtropical low cloud response to a warmer climate in a superparameterized climate model. Part II: column modeling with a cloud resolving model. J. Adv. Model. Earth Syst. 1, 8. http:// dx.doi.org/10.3894/JAMES.2009.1.8. Bony, S., Colman, R., Kattsov, V.M., Allan, R.P., Bretherton, C.S., Dufresne, J.L., Hall, A., Hallegatte, S., Holland, M.M., Ingram, W., Randall, D.A., Soden, B.J., Tselioudis, G., Webb, M.J., 2006. How well do we understand and evaluate climate change feedback processes? J. Clim. 19, 3445e3482. http://dx.doi.org/10.1175/ JCLI3819.1. Brient, F., Bony, S., 2012. How may low-cloud radiative properties simulated in the current climate influence low-cloud feedbacks under global warming? Geophys. Res. Lett. 39, L20807. http://dx.doi.org/10.1029/2012GL053265. Capotondi, A., Wittenberg, A., Masina, S., 2006. Spatial and temporal structure of tropical pacific interannual variability in 20th century coupled simulations. Ocean Model. 15, 274e298. http://dx.doi.org/10.1016/j.ocemod.2006.02.004. Carton, J.A., Chepurin, G., Cao, X., 2000a. A simple ocean data assimilation analysis of the global upper ocean 1950e95. Part II: results. J. Phys. Oceanogr. 30, 311e326. http://dx.doi.org/10.1175/1520-0485(2000)030< 0311:ASODAA> 2.0.CO;2. Carton, J.A., Chepurin, G., Cao, X., Giese, B., 2000b. A simple ocean data assimilation analysis of the global upper ocean 1950e95. Part I: methodology. J. Phys. Oceanogr. 30, 294e309. http://dx.doi.org/10.1175/1520-0485(2000)030< 0294: ASODAA> 2.0.CO;2. Cess, R.D., Harrison, E.F., Minnis, P., Barkstrom, B.R., Ramanathan, V., Kwon, T.Y., 1992. Interpretation of seasonal cloud-climate interactions using earth radiation budget experiment data. J. Geophys. Res. Atmos. 97, 7613e7617. http:// dx.doi.org/10.1029/92JD00547. Cook, K.H., Meehl, G.A., Arblaster, J.M., 2012. Monsoon regimes and processes in CCSM4. Part II: African and American monsoon systems. J. Clim. 25, 2609e2621. http://dx.doi.org/10.1175/JCLI-D-11-00185.1. Danabasoglu, G., Yeager, S.G., Kwon, Y.O., Tribbia, J.J., Phillips, A.S., Hurrell, J.W., 2012. Variability of the Atlantic meridional overturning circulation in CCSM4. J. Clim. 25, 5153e5172. http://dx.doi.org/10.1175/JCLI-D-11-00463.1. DeMott, C.A., Stan, C., Randall, D.A., 2013. Northward propagation mechanisms of the boreal summer intraseasonal oscillation in the ERA-interim and SP-CCSM. J. Clim. 26, 1973e1992. http://dx.doi.org/10.1175/JCLI-D-12-00191.1. DeMott, C.A., Stan, C., Randall, D.A., Kinter, J.L., Khairoutdinov, M., 2011. The asian monsoon in the superparameterized CCSM and its relationship to tropical wave activity. J. Clim. 24, 5134e5156. http://dx.doi.org/10.1175/2011JCLI4202.1. Deser, C., Phillips, A.S., Tomas, R.A., Okumura, Y.M., Alexander, M.A., Capotondi, A., Scott, J.D., Kwon, Y.O., Ohba, M., 2011. ENSO and Pacific decadal variability in the Community Climate System Model version 4. J. Clim. 25, 2622e2651. http:// dx.doi.org/10.1175/JCLI-D-11-00301.1. Dirmeyer, P., Cash, B., Kinter III, James L., Jung, T., Marx, L., Satoh, M., Stan, C., Tomita, H., Towers, P., Wedi, N., Achuthavarier, D., Adams, J., Altshuler, E., Huang, B., Jin, E., Manganello, J., 2012. Simulating the diurnal cycle of rainfall in global climate models: resolution versus parameterization. Clim. Dyn. 39, 399e418. http://dx.doi.org/10.1007/s00382-011-1127-9. Fu, X., Wang, B., 2001. A coupled modeling study of the seasonal cycle of the Pacific cold tongue. Part I: simulation and sensitivity experiments. J. Clim. 14, 765e779. http://dx.doi.org/10.1175/1520-0442(2001)014< 0765:ACMSOT> 2.0.CO;2. Ganachaud, A., Wunsch, C., 2000. Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature 408, 453e457. http://dx.doi.org/10.1038/35044048. Gent, P., Yeager, S., Neale, R., Levis, S., Bailey, D., 2010. Improvements in a half degree atmosphere/land version of the CCSM. Clim. Dyn. 34, 819e833. http:// dx.doi.org/10.1007/s00382-009-0614-8. Gent, P.R., Danabasoglu, G., Donner, L.J., Holland, M.M., Hunke, E.C., Jayne, S.R., Lawrence, D.M., Neale, R.B., Rasch, P.J., Vertenstein, M., Worley, P.H., Yang, Z.L., Zhang, M., 2011. The community climate system model version 4. J. Clim. 24, 4973e4991. http://dx.doi.org/10.1175/2011JCLI4083.1. Gettelman, A., Kay, J.E., Shell, K.M., 2012. The evolution of climate sensitivity and climate feedbacks in the community atmosphere model. J. Clim. 25, 1453e1469. http://dx.doi.org/10.1175/JCLI-D-11-00197.1. Grabowski, W.W., 2001. Coupling cloud processes with the large-scale dynamics using the cloud-resolving convection parameterization (CRCP). J. Atmos. Sci. 58, 978e997. http://dx.doi.org/10.1175/1520-0469(2001)058< 0978:CCPWTL> 2.0.CO;2. Gregory, J.M., Ingram, W.J., Palmer, M.A., Jones, G.S., Stott, P.A., Thorpe, R.B., Lowe, J.A., Johns, T.C., Williams, K.D., 2004. A new method for diagnosing radiative forcing and climate sensitivity. Geophys. Res. Lett. 31 http:// dx.doi.org/10.1029/2003GL018747. Griffies, S.M., Winton, M., Donner, L.J., Horowitz, L.W., Downes, S.M., Farneti, R., Gnanadesikan, A., Hurlin, W.J., Lee, H.C., Liang, Z., Palter, J.B., Samuels, B.L., Wittenberg, A.T., Wyman, B.L., Yin, J., Zadeh, N., 2011. The GFDL CM3 coupled climate model: characteristics of the ocean and sea ice simulations. J. Clim. 24, 3520e3544. http://dx.doi.org/10.1175/2011JCLI3964.1. Gupta, A.S., Muir, L.C., Brown, J.N., Phipps, S.J., Durack, P.J., Monselesan, D., Wijffels, S.E., 2012. Climate drift in the CMIP3 models. J. Clim. 25, 4621e4640. http://dx.doi.org/10.1175/JCLI-D-11-00312.1.

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152 Hagos, S.M., Cook, K.H., 2007. Dynamics of the West African monsoon jump. J. Clim. 20, 5264e5284. http://dx.doi.org/10.1175/2007JCLI1533.1. Huang, B., Hu, Z.Z., Schneider, E., Wu, Z., Xue, Y., Klinger, B., 2012. Influences of tropicaleextratropical interaction on the multidecadal AMOC variability in the NCEP climate forecast system. Clim. Dyn. 39, 531e555. http://dx.doi.org/ 10.1007/s00382-011-1258-z. Huffman, G.J., Adler, R.F., Arkin, P., Chang, A., Ferraro, R., Gruber, A., Janowiak, J., McNab, A., Rudolf, B., Schneider, U., 1997. The global precipitation climatology project (GPCP) combined precipitation dataset. Bull. Am. Meteorol. Soc. 78, 5e20. http://dx.doi.org/10.1175/1520-0477(1997)078< 0005:TGPCPG> 2.0.CO;2. Jakeman, A.J., Letcher, R.A., Norton, J.P., 2006. Ten iterative steps in development and evaluation of environmental models. Environ. Model. Softw. 21, 602e614. http://dx.doi.org/10.1016/j.envsoft.2006.01.004. Khairoutdinov, M.F., Randall, D.A., 2001. A cloud resolving model as a cloud parameterization in the NCAR community climate system model: preliminary results. Geophys. Res. Lett. 28, 3617e3620. http://dx.doi.org/10.1029/ 2001GL013552. Khairoutdinov, M.F., Randall, D.A., 2003. Cloud resolving modeling of the arm summer 1997 IOP: model formulation, results, uncertainties, and sensitivities. J. Atmos. Sci. 60, 607e625. http://dx.doi.org/10.1175/1520-0469(2003)060< 0607:CRMOTA> 2.0.CO;2. Kim, D., Sobel, A.H., Del Genio, A.D., Chen, Y., Camargo, S.J., Yao, M.S., Kelley, M., Nazarenko, L., 2012. The tropical subseasonal variability simulated in the NASA GISS general circulation model. J. Clim. 25, 4641e4659. http://dx.doi.org/ 10.1175/JCLI-D-11-00447.1. Kirtman, B.P., 1997. Oceanic Rossby wave dynamics and the ENSO period in a coupled model. J. Clim. 10, 1690e1704. http://dx.doi.org/10.1175/15200442(1997)010< 1690:ORWDAT> 2.0.CO;2. Kirtman, B.P., Schneider, E.K., 1996. Model-based estimates of equatorial Pacific wind stress. J. Clim. 9, 1077e1091. http://dx.doi.org/10.1175/1520-0442(1996) 009< 1077:MBEOEP> 2.0.CO;2. Klinger, B.A., Drijfhout, S., Marotzke, J., Scott, J.R., 2003. Sensitivity of basinwide meridional overturning to diapycnal diffusion and remote wind forcing in an idealized Atlanticesouthern ocean geometry. J. Phys. Oceanogr. 33, 249e266. http://dx.doi.org/10.1175/1520-0485(2003)033< 0249:SOBMOT> 2.0.CO;2. Krishnamurthy, V., Stan, C., Randall, D.A., Shukla, R.P., Kinter, J.L., 2014. Simulation of the south Asian monsoon in a coupled model with an embedded cloudresolving model. J. Clim. 27, 1121e1142. http://dx.doi.org/10.1175/JCLI-D-1300257.1. , L., Lebel, T., Tapsoba, D., 2002. Rainfall variability in West Africa during the Le Barbe years 1950e90. J. Clim. 15, 187e202. http://dx.doi.org/10.1175/1520-0442(2002) 015< 0187:RVIWAD> 2.0.CO;2. Liebmann, B., Smith, C.A., 1996. Describtion of a complete (interpolated) outgoing longwave radiation dataset. Bull. Am. Meteorol. Soc. 77, 1275e1277. Lin, S.J., 2004. A “vertically lagrangian”finite-volume dynamical core for global models. Monthly Weather Rev. 132, 2293e2307. http://dx.doi.org/10.1175/15200493(2004)132< 2293:AVLFDC> 2.0.CO;2. Liu, P., Satoh, M., Wang, B., Fudeyasu, H., Nasuno, T., Li, T., Miura, H., Taniguchi, H., Masunaga, H., Fu, X., Annamalai, H., 2009. An MJO simulated by the NICAM at 14- and 7-km resolutions. Monthly Weather Rev. 137, 3254e3268. http:// dx.doi.org/10.1175/2009MWR2965.1. Loeb, N.G., Wielicki, B.A., Doelling, D.R., Smith, G.L., Keyes, D.F., Kato, S., ManaloSmith, N., Wong, T., 2009. Toward optimal closure of the earth's top-ofatmosphere radiation budget. J. Clim. 22, 748e766. http://dx.doi.org/10.1175/ 2008JCLI2637.1. Maloney, E.D., Hartmann, D.L., 2001. The sensitivity of intraseasonal variability in the NCAR CCSM3 to changes in convective parameterization. J. Clim. 14, 2015e2034. http://dx.doi.org/10.1175/1520-0442(2001)014< 2015:TSOIVI> 2.0.CO;2. McKee, T., Doesken, N.J., Kleist, J., 1995. Drought Monitoring with Multiple Time Scales, pp. 233e236. Meehl, G.A., Washington, W.M., Arblaster, J.M., Hu, A., Teng, H., Tebaldi, C., Sanderson, B.N., Lamarque, J.F., Conley, A., Strand, W.G., White, J.B., 2012. Climate system response to external forcings and climate change projections in CCSM4. J. Clim. 25, 3661e3683. http://dx.doi.org/10.1175/JCLI-D-11-00240.1. Miura, H., Satoh, M., Nasuno, T., Noda, A.T., Oouchi, K., 2007. A Madden-Julian oscillation event realistically simulated by a global cloud-resolving model. Science 318, 1763e1765. Moss, R.H., Edmonds, J.A., Hibbard, K.A., Manning, M.R., Rose, S.K., van Vuuren, D.P., Carter, T.R., Emori, S., Kainuma, M., Kram, T., Meehl, G.A., Mitchell, J.F.B., Nakicenovic, N., Riahi, K., Smith, S.J., Stouffer, R.J., Thomson, A.M., Weyant, J.P., Wilbanks, T.J., 2010. The next generation of scenarios for climate change research and assessment. Nature 463, 747e756. http://dx.doi.org/10.1038/ nature08823. Neale, R.B., Richter, J.H., Jochum, M., 2008. The impact of convection on ENSO: from a delayed oscillator to a series of events. J. Clim. 21, 5904e5924. http:// dx.doi.org/10.1175/2008JCLI2244.1. Neale, R.B., Richter, J., Park, S., Lauritzen, P.H., Vavrus, S.J., Rasch, P.J., Zhang, M., 2013. The mean climate of the community atmosphere model (CAM4) in forced SST and fully coupled experiments. J. Clim. 26, 5150e5168. http://dx.doi.org/ 10.1175/JCLI-D-12-00236.1. Neelin, J.D., Peters, O., Lin, J.W.B., Hales, K., Holloway, C.E., 2008. Rethinking convective quasi-equilibrium: observational constraints for stochastic convective schemes in climate models. Phil. Trans. Math. Phys. Eng. Sci. 366, 2581e2604. http://dx.doi.org/10.2307/25190856.

151

Palmer, T.N., 1999. A nonlinear dynamical perspective on climate prediction. J. Clim. 12, 575e591. http://dx.doi.org/10.1175/1520-0442(1999)012< 0575:ANDPOC> 2.0.CO;2. Prodhomme, C., Terray, P., Masson, S., Izumo, T., Tozuka, T., Yamagata, T., 2014. Impacts of Indian Ocean SST biases on the Indian monsoon: as simulated in a global coupled model. Clim. Dyn. 42, 271e290. http://dx.doi.org/10.1007/ s00382-013-1671-6. Randall, D., Khairoutdinov, M., Arakawa, A., Grabowski, W., 2003. Breaking the cloud parameterization deadlock. Bull. Am. Meteorol. Soc. 84, 1547e1564. http://dx.doi.org/10.1175/BAMS-84-11-1547. Randall, D.A., et al., 2007. Climate Change 2007: The Physical Science Basis. Cambridge University Press, pp. 591e662. Contribution of working group 1 to the fourth assessment report to the Intergovernamental Panel on Climate Change. Randel, D.L., Greenwald, T.J., Vonder Haar, T.H., Stephens, G.L., Ringerud, M.A., Combs, C.L., 1996. A new global water vapor dataset. Bull. Am. Meteorol. Soc. 77, 1233e1246. http://dx.doi.org/10.1175/1520-0477(1996)077< 1233:ANGWVD> 2.0.CO;2. Rayner, N.A., Brohan, P., Parker, D.E., Folland, C.K., Kennedy, J.J., Vanicek, M., Ansell, T.J., Tett, S.F.B., 2006. Improved analyses of changes and uncertainties in sea surface temperature measured in situ since the mid-nineteenth century: the HadSST2 dataset. J. Clim. 19, 446e469. http://dx.doi.org/10.1175/JCLI3637.1. Satoh, M., Matsuno, T., Tomita, H., Miura, H., Nasuno, T., Iga, S., 2008. Nonhydrostatic icosahedral atmospheric model (NICAM) for global cloud resolving simulations. J. Comput. Phys. 227, 3486e3514. http://dx.doi.org/10.1016/j.jcp.2007.02.006. Schmittner, A., Urban, N.M., Shakun, J.D., Mahowald, N.M., Clark, P.U., Bartlein, P.J., , A., 2011. Climate sensitivity estimated from temperature Mix, A.C., Rosell-Mele reconstructions of the last glacial maximum. Science 334, 1385e1388. Sherwood, S.C., Bony, S., Dufresne, J.L., 2014. Spread in model climate sensitivity traced to atmospheric convective mixing. Nature 505, 37e42. http://dx.doi.org/ 10.1038/nature12829. Shiogama, H., Watanabe, M., Yoshimori, M., Yokohata, T., Ogura, T., Annan, J., Hargreaves, J., Abe, M., Kamae, Y., O'ishi, R., Nobui, R., Emori, S., Nozawa, T., AbeOuchi, A., Kimoto, M., 2012. Perturbed physics ensemble using the MIROC5 coupled atmosphereeocean GCM without flux corrections: experimental design and results. Clim. Dyn. 39, 3041e3056. http://dx.doi.org/10.1007/ s00382-012-1441-x. Soden, B.J., Held, I.M., 2006. An assessment of climate feedbacks in coupled oceaneatmosphere models. J. Clim. 19, 3354e3360. http://dx.doi.org/10.1175/ JCLI3799.1. Solomon, S., Qin, D., Manning, M., Marquis, M., Averyt, K., Tignor Jr., M.M.B., Miller, H.L., Chen, Z., 2007. Climate Change 2007: The Physical Science Basis. Cambridge University Press. Srokosz, M., Baringer, M., Bryden, H., Cunningham, S., Delworth, T., Lozier, S., Marotzke, J., Sutton, R., 2012. Past, present, and future changes in the Atlantic meridional overturning circulation. Bull. Am. Meteorol. Soc. 93, 1663e1676. http://dx.doi.org/10.1175/BAMS-D-11-00151.1. Stan, C., 2012. Is cumulus convection the concertmaster of tropical cyclone activity in the atlantic? Geophys. Res. Lett. 39, L19716. http://dx.doi.org/10.1029/ 2012GL053449. Stan, C., Khairoutdinov, M., DeMott, C.A., Krishnamurthy, V., Straus, D.M., Randall, D.A., Kinter III, James L., Shukla, J., 2010. An oceaneatmosphere climate simulation with an embedded cloud resolving model. Geophys. Res. Lett. 37 http://dx.doi.org/10.1029/2009GL040822. Subramanian, A.C., Jochum, M., Miller, A.J., Murtugudde, R., Neale, R.B., Waliser, D.E., 2011. The MaddeneJulian oscillation in ccsm4. J. Clim. 24, 6261e6282. http:// dx.doi.org/10.1175/JCLI-D-11-00031.1. Sultan, B., Janicot, S., 2000. Abrupt shift of the ITCZ over West Africa and intraseasonal variability. Geophys. Res. Lett. 27, 3353e3356. http://dx.doi.org/ 10.1029/1999GL011285. Sultan, B., Janicot, S., Diedhiou, A., 2003. The west African monsoon dynamics. Part I: documentation of intraseasonal variability. J. Clim. 16, 3389e3406. http:// dx.doi.org/10.1175/1520-0442(2003)016< 3389:TWAMDP> 2.0.CO;2. Talley, L.D., Reid, J.L., Robbins, P.E., 2003. Data-based meridional overturning streamfunctions for the global ocean. J. Clim. 16, 3213e3226. http://dx.doi.org/ 10.1175/1520-0442(2003)016< 3213:DMOSFT> 2.0.CO;2. Taylor, K.E., Stouffer, R.J., Meehl, G.A., 2011. An overview of CMIP5 and the experiment design. Bull. Am. Meteorol. Soc. 93, 485e498. http://dx.doi.org/10.1175/ BAMS-D-11-00094.1. , P., 2011. Annual cycle of the west Thorncroft, C.D., Nguyen, H., Zhang, C., Peyrille African monsoon: regional circulations and associated water vapour transport. Q. J. R. Meteorol. Soc. 137, 129e147. http://dx.doi.org/10.1002/qj.728. Webb, M.J., Lock, A., 2012. Coupling between subtropical cloud feedback and the local hydrological cycle in a climate model. Clim. Dyn., 1e17. http://dx.doi.org/ 10.1007/s00382-012-1608-5. Wehner, M., Smith, R., Bala, G., Duffy, P., 2010. The effect of horizontal resolution on simulation of very extreme US precipitation events in a global atmosphere model. Clim. Dyn. 34, 241e247. http://dx.doi.org/10.1007/s00382-009-0656-y. Wei, T., Wang, L., Dong, W., Dong, M., Zhang, J., 2011. A comparison of east Asian summer monsoon simulations from cam3.1 with three dynamic cores. Theor. Appl. Climatol. 106, 295e306. http://dx.doi.org/10.1007/s00704-011-0409-9. Wheeler, M., Kiladis, G.N.,1999. Convectively coupled equatorial waves: analysis of clouds and temperature in the wavenumberefrequency domain. J. Atmos. Sci. 56, 374e399. http://dx.doi.org/10.1175/1520-0469(1999)056< 0374:CCEWAO> 2.0.CO;2. Williamson, D.L., 2008. Convergence of aqua-planet simulations with increasing resolution in the Community Atmospheric Model, version 3. Tellus A 60, 848e862. http://dx.doi.org/10.1111/j.1600-0870.2008.00339.x.

152

C. Stan, L. Xu / Environmental Modelling & Software 60 (2014) 134e152

Wyant, M.C., Bretherton, C.S., Blossey, P.N., 2009. Subtropical low cloud response to a warmer climate in a superparameterized climate model. Part I: regime sorting and physical mechanisms. J. Adv. Model. Earth Syst. 1, 7. http://dx.doi.org/ 10.3894/JAMES.2009.1.7. Wyant, M.C., Bretherton, C.S., Blossey, P.N., Khairoutdinov, M., 2012. Fast cloud adjustment to increasing CO2 in a superparameterized climate model. J. Adv. Model. Earth Syst. 4, M05001. http://dx.doi.org/10.1029/2011MS000092. Xu, K.M., Cheng, A., 2013. Evaluating low cloud simulation from an upgraded multiscale modeling framework model. Part I: sensitivity to spatial resolution and climatology. J. Clim. http://dx.doi.org/10.1175/JCLI-D-12e00200.1.

Xu, X., Schmitz, W.J., Hurlburt, H.E., Hogan, P.J., 2012. Mean Atlantic meridional overturning circulation across 26.5
n from eddy-resolving simulations compared to observations. J. Geophys. Res. 117, C03042. http://dx.doi.org/ 10.1029/2011JC007586. Zhou, L., Neale, R.B., Jochum, M., Murtugudde, R., 2011. Improved MaddeneJulian oscillations with improved physics: the impact of modified convection parameterizations. J. Clim. 25, 1116e1136. http://dx.doi.org/10.1175/ 2011JCLI4059.1. Zhu, H., Hendon, H., 2010. Convection and MJO Performance in UM7.1. Technical Report 022. The Center for Australian Weather and Climate Research.