evaluation of el niño–southern oscillation in the access ... · evaluation of el...

Australian Meteorological and Oceanographic Journal 63 (2013) 161–180

161

Evaluation of El Niño–Southern Oscillation in the ACCESS coupled model simulations

for CMIP5Harun A. Rashid1, Arnold Sullivan1, Anthony C. Hirst1, Daohua Bi1,

Xiaobing Zhou2, and Simon J. Marsland1

1Centre for Australian Weather and Climate Research (CAWCR), a partnership between CSIRO and the Bureau of Meteorology,

CSIRO Marine and Atmospheric Research, Aspendale, Australia.2CAWCR/Bureau of Meteorology, Melbourne, Australia

(Manuscript received July 2012; revised February 2013)

One of the key performance measures for Coupled General Circulation Mod-els (CGCMs) is their ability to realistically simulate the prominent modes of cli-mate variability. Here, we investigate the realism of El Niño–Southern Oscilla-tion (ENSO), the dominant mode of observed interannual climate variability, as simulated by the Australian Community Climate and Earth System Simulator (AC-CESS). We examine the key ENSO properties calculated from observations and from the pre-industrial control and historical simulations performed by versions 1.0 and 1.3 of the ACCESS coupled model (ACCESS-CM), as part of the Coupled Model Intercomparison Project phase 5 (CMIP5). The ENSO properties examined are the spatial structure and timescales, phase locking to the annual cycle, growth and decay of Niño3 SST anomalies, and the coupled ENSO evolution, as well as the ENSO induced teleconnection patterns (as diagnosed by linear regressions of the three-dimensional atmospheric circulation field, rainfall, sea-level pres-sure, and the upper-tropospheric geopotential heights onto the Niño3 index). The coupled evolutions of SST, zonal wind stress, and the thermocline depth during a typical ENSO event are used to illustrate the essential features of ENSO dynamics. Our results show that both versions of the ACCESS-CM perform reasonably well in simulating the ENSO properties and teleconnections. In particular, the spatial structure, timescales, phase locking, and the growth and decay rates of ENSO are well simulated. The evolution of SST, zonal wind stress, and the thermocline depth during ENSO events is also reasonably well simulated, although the SST (zonal wind stress) response is found to be stronger (weaker) than observed. The simulated ENSO teleconnection patterns are also mostly realistic, despite some adverse impact of the equatorial Pacific cold SST bias on the local precipitation response. The present study shows that the ACCESS-CMs simulate the key ENSO properties (e.g. amplitude and spatial structure, timescales, and atmospheric tele-connections) better than most of the previous generations of CGCMs.

Introduction

Over the last several decades, CGCMs have been used to a great effect for climate change projections on centennial timescales and for understanding the physical processes underlying these changes. More recently, this effort has

been coordinated through various phases of the Coupled Model Intercomparison Project (CMIP) that, among other things, underpin the Inter-governmental Panel on Climate Change (IPCC) assessment reports (Taylor et al. 2012). These models are also increasingly being used for understanding and prediction of weather and climate variability on timescales ranging from days to decades. For example, many modelling centres now routinely use CGCMs for dynamical seasonal and decadal predictions. The success of a CGCM

Corresponding author address: Harun Rashid, email: harun.rashid@ csiro.au

162 Australian Meteorological and Oceanographic Journal 63:1 March 2013

coupled model (ACCESS-CM) is a state-of-the-art GCM with sophisticated parameterisations for unresolved physical processes involving radiation, clouds, aerosols, greenhouse gases, convection, gravity-wave drags, and the atmospheric boundary layer (Martin et al. 2011). The ocean and sea-ice components also incorporate the latest advancements in ocean−sea ice modelling (Marsland et al. 2013, Uotila et al. 2013). In addition, the atmosphere model includes sophisticated treatments of the external radiative forcings arising from both natural variability (due to variations in solar output and volcanic eruption) and anthropogenic sources, e.g. changes in greenhouse gas and aerosol concentrations (Dix et al. 2013). This model is, therefore, expected to perform relatively well in its simulation of mean climate and climate variability. Indeed, there have been considerable improvements in the tropical Pacific mean-state bias and other aspects of mean climate (Bi et al. 2013a, Rashid et al. 2013). For example, the equatorial Pacific cold bias is only around 1 °C in the ACCESS models, which is considerably lower than the cold bias found in most CMIP3 models (e.g. Lin 2007, Reichler and Kim 2008).

In this paper, we investigate the performance of the ACCESS-CM in simulating ENSO and its associated teleconnection patterns using the pre-industrial control and historical simulations contributed to the CMIP5 archive. These multi-century simulations are well-suited for examining the possible improvements in the ENSO mode in the ACCESS-CM. We use observations and simulations from versions 1.0 and 1.3 of the ACCESS-CM to examine the realism of fundamental ENSO properties: amplitude and spatial structure, timescales, seasonality, growth and decay rates, and the evolution of ENSO. We also investigate the ENSO-induced teleconnection patterns in the ACCESS-CM simulations and compare these patterns with those derived from a reanalysis dataset and an AMIP simulation. The rest of the paper is structured as follows. First, we briefly describe the model formulation and experiment designs, followed by some background results concerning the tropical mean climate and its annual cycle. Next, the main results involving ENSO properties and teleconnection are presented. Finally, the conclusions of this work are given.

Models, experiments, and validation datasets

Here, we provide a brief description of the ACCESS-CM; a more detailed description of this coupled model may be found in Bi et al. (2013a). The CSIRO-BOM ACCESS-CM is an atmosphere–ocean sea-ice model, based on the atmospheric component of the U.K. Met Office Unified Model (UM, version 7.3) and the ACCESS Ocean sea-ice Model (ACCESS-OM; Bi et al. 2013b) coupled through the OASIS version 3.2.5 (Valcke 2006) software. The atmosphere model has N96 horizontal resolution, equivalent to a 1.25° by 1.875° latitude–longitude grid. There are 38 levels in the vertical, with the top level placed at a height of about 39 km. The

in performing these tasks relies on, to a large extent, its ability to realistically simulate the dominant modes of climate variability (e.g. Kirtman and Schopf 1998). Of central importance to both climate change projection and the prediction of climate variability is the realistic simulation of El Niño–Southern Oscillation (ENSO), the associated feedback processes and the atmospheric teleconnection patterns in the CGCM (e.g. Turner et al. 2005, Shukla et al. 2006, DelSole and Shukla 2010). The realistic ENSO simulation, however, has remained a challenge for state-of-the-art climate models (e.g. Joseph and Nigam 2006, AchutaRao and Sperber 2006). A main source of difficulty is the CGCMs inability to correctly simulate the time-mean (or basic) state in the tropical Pacific (e.g. Turner et al. 2005, Magnusson et al. 2012), which in turn is due to the inadequacy of the parameterised physical processes and coarse spatial resolutions used in the models (e.g. Guilyardi et al. 2004). Therefore, the most fundamental ways of improving the performance of a CGCM are through better and more complete parameterisations of physical processes of the component models (atmosphere, ocean, and sea-ice) and through increasing the model’s spatial resolutions (Guilyardi et al. 2009a, Roberts et al. 2009).

ENSO is the dominant mode of interannual climate variability, which affects the global weather patterns through atmospheric teleconnections. It can also modulate the tropical Pacific sea surface temperature (SST) response to the greenhouse gas (GHG) increases through air–sea exchange of heat (e.g. DiNezio et al. 2009). The ENSO mode owes its existence to the unstable ocean–atmosphere interactions in the tropical Pacific and/or through the excitation due to higher-frequency noise of an otherwise stable coupled mode (Suarez and Schopf 1988, Battisiti and Hirst 1989, Penland and Sardeshmukh 1995, Jin 1997). The ENSO-related ocean−atmosphere interactions are characterised by a number of feedback processes, including the SST-zonal wind stress or Bjerknes feedback, the SST-thermocline feedback, and the SST-heatflux or thermodynamic feedback. The first two processes constitute a positive feedback, whereas the last process provides a negative feedback to ENSO. Historical observations show that the ENSO phenomenon recurs every 3–7 years, with a significant anomalous warming (locally up to 5 °C) of the tropical central and eastern Pacific and a concomitant (smaller) cooling of the far western Pacific (e.g. Rasmusson and Carpentar 1982, Philander 1990). This anomalous warming temporarily shifts the atmospheric deep convection region from the western Pacific to the central Pacific. Such a shift, in turn, displaces the upper-tropospheric Rossby wave source, leading to the ENSO-induced teleconnection pattern and the associated disruptions of world weather (e.g. Trenberth et al. 1998).

The Australian Community Climate and Earth System Simulator (ACCESS) is a state-of-the-art global coupled model, which incorporates comprehensive parameterisations of physical processes of the climate sub-systems as well as a reasonably high spatial resolution (Bi et al. 2013a). The atmospheric component of the ACCESS

Rashid etal.: Evaluation of El Niño-Southern Oscillation in the ACCESS coupled model simulations for CMIP5 163

performing the core CMIP5 experiments (Taylor et al. 2012, Dix et al. 2013), including the 500-year pre-industrial control (piControl) and the historical (hist) simulations. We use model outputs from the two experiments to assess the realism of simulated ENSO properties. Sea surface temperatures corresponding to years 1870–2005 from each experiment are used to examine the fundamental ENSO properties. The simulated atmospheric data used for the teleconnection study correspond to the 1979–2005 period, to facilitate comparison with the reanalysis and AMIP data.

Several observational and reanalysis datasets have been used for validation of the model results. For sea surface temperatures, we use the HadISST dataset from the Met Office Hadley Centre (Rayner et al. 2003). The NOAA extended reconstructed sea surface temperature (ERSST, V3b) dataset of Smith and Reynolds (2003) is also used for a comparison. The simulated wind stress was validated against a recent air-sea flux product (TropFlux; Praveen Kumar et al. 2012) and version 2 of the Coordinated Ocean-ice Reference Experiment (CORE) flux product (CORE.v2; Large and Yeager 2009). Precipitation response was evaluated using the GPCP data set (Adler et al. 2003, Huffman et al. 2009). The ERA-Interim reanalysis dataset (Dee et al. 2011) was used to assess the ENSO teleconnection in the pressure and geopotential height fields. The thermocline depth is estimated by the depth of 20 °C isotherm, which was computed from two ocean reanalyses, Simple Ocean Data Assimilation (SODA; version 2.2.4) reanalysis (Carton et al. 2000) and the Australian Bureau of Meteorology’s newly developed POAMA Ensemble Ocean Data Assimilation System (PEODAS; Yin et al. 2011).

Mean climate and annual cycle in the tropics

Mean climateProperties of the simulated ENSO are closely related to the representation of the tropical Pacific time-mean state and annual cycle in the coupled model (e.g. Battisti and Hirst 1989, Gu and Philander 1995, Fedorov and Philander 2001). We therefore briefly describe the means and annual cycles of the variables (SST, zonal wind stress and the thermocline depth) that are directly related to ENSO properties; more information about the simulated time-mean state in ACCESS-CM may be found elsewhere (Bi et al. 2013a, Rashid et al. 2013, Marsland et al. 2013).

Figure 1 shows the time-mean SSTs over the tropics from two observational datasets (top row) and the CMIP5 historical (middle row) and pre-industrial control (bottom row) simulations. The first panel shows the time-mean SSTs from the HadSST dataset; all other panels show the mean values (contours) and the difference (shades) of the mean values from those in the first panel (Fig. 1(a)). The values plotted in the top-right panel (Fig. 1(b)) give us an indication of observational uncertainties. As we can see, the biases (with respect to HadSST) in the simulated mean SST are larger than the estimated observational errors. An

atmospheric component has a non-hydrostatic dynamical core, and includes a comprehensive set of parameterisation schemes for physical processes. Among the parameterised physical processes are radiation, interactive aerosols, land surface processes and hydrology, boundary layer turbulent mixing, convection, cloud and precipitation microphysics, and gravity wave drag. A more complete description of the UM is given in Collins et al. (2008), Martin et al. (2011), Hewitt et al. (2011) and Walters et al. (2011); the particular versions of the UM used in the ACCESS-CM are described in Bi et al. (2013a). ACCESS-OM is based on the U.S. NOAA/GFDL Modular Ocean Model (MOM4p1) coupled to the U.S. Los Alamos National Laboratory sea-ice model (CICE4.1). ACCESS-OM has 50 vertical levels, with finer vertical resolution in the upper ocean layers (the resolution increasing to 10 m throughout the top 200 m). The nominal horizontal resolution is 1° in the zonal direction, apart from a tripolar grid in the Arctic (Murray 1996). The meridional resolution is 1/3° in a band extending 10° either side of the equator, and nominally 1° elsewhere (apart from the tripolar grid region in the Arctic, and a cosine varying Mercator region in the southern hemisphere). The CICE model has the same horizontal resolution as the ocean model and has five sea-ice categories. More detailed information about the ACCESS-OM may be found in Bi et al. (2013b), Marsland et al. (2013) and Uotila et al. (2013).

Two versions of the ACCESS-CM have been used in this study, with differing atmosphere model configurations. In the first version (hereafter, ACCESS1.0) the atmospheric model has the HadGEM2 configuration (Collins et al. 2008, Martin et al. 2011), whereas in the second version (ACCESS1.3) the atmospheric configuration is closely related to the Global Atmosphere version 1.0 (GA1.0) configuration (Hewitt et al. 2011, Walters et al. 2011). Thus, ACCESS1.3 uses a prognostic cloud scheme (Wilson et al. 2008) and ACCESS1.0 uses the diagnostic cloud scheme of Smith (1990). The other major difference is that ACCESS1.3 uses the CABLE land surface scheme (Kowalczyk et al. 2006, 2013), whereas ACCESS1.0 uses the MOSES 2 land surface scheme (Essery et al. 2003). In addition, the ACCESS1.3 version includes some further changes in the cloud, radiation, and boundary layer schemes, as detailed in Bi et al. (2013a).

The ocean models used in ACCESS1.0 and ACCESS1.3 are identical, except for one difference: the two models have different values for the critical Richardson number, used for determining the occurrence of enhanced mixing. In ACCESS1.3, this value is 0.3; however, a smaller value (0.15) is used in ACCESS1.0, which reduces the mixed-layer depth (Large et al. 1994, their Table 2) and enhances stratification in the upper equatorial thermocline. The effect on ENSO was to moderately increase its strength. This is demonstrated by a seven per cent increase in the standard deviation of the Niño3 index (from 0.71 to 0.76) with respect to that in a parallel simulation of ACCESS1.0 with the critical Richardson number set to 0.30 (same as that used in ACCESS1.3).

The ACCESS-CM versions have been used for


TropFlux and the differences of the COREv2 and simulated τx are shown in Fig. 2. The time-mean τx is easterly almost everywhere in the tropics, except in a few regions including the equatorial and northern Indian Ocean, where the wind stress is associated with the southwest monsoon flows (Fig. 2(a)). The wind stresses are part of the trade wind system and have a maximum magnitude on both sides of the equator. The uncertainties associated with the two observational estimates are displayed in Fig. 2(b). In general, the τx from COREv2 is stronger than that from the TropFlux dataset. Also, both versions of the ACCESS-CM shows stronger τx

almost everywhere in the tropics, with the largest easterly biases in the north equatorial Pacific. The τx bias is also easterly along the equator in the Pacific and Indian Ocean sectors, implying stronger than observed easterly winds there. The occurrence of stronger than observed easterly wind stresses is a common problem in CGCMs (e.g. Lin 2007, Guilyardi et al. 2009a). While the equatorial τx bias is weaker than the off-equatorial bias in ACCESS-CMs, this

overall cold bias is evident in both simulations, and this cold bias is stronger and more wide-spread in ACCESS1.0 than in ACCESS1.3. However, the spatial pattern of the full SST values (contours) in ACCESS1.0 has a slightly better correspondence with the observation. In particular, the east Pacific cold tongue in ACCESS1.3 appears to extend too far into the West Pacific; also, the South Pacific Convergence Zone looks a bit worse. The regions with large warm bias are confined to the eastern parts of the ocean basins in both models. While substantial cold and warm biases still exist in the ACCESS models, the equatorial Pacific cold biases are only 1 °C in ACCESS1.0 and 1.3 °C in ACCESS1.3, which are lower than the multi-model mean cold bias of around 2 °C in CMIP3 models (e.g. Lin 2007, Reichler and Kim 2008). A reduced cold bias implies a mean state more representative of the nature, which is conducive to a realistic simulation of ENSO (Magnusson et al. 2012).

The zonal wind stress (τx) also plays a central role in determining ENSO characteristics. The time-mean τx from

Fig. 1. The climatological mean SST (°C) distributions (contours) in observations and coupled model simulations for 1870−2005 over the tropics, and the departures (shades) of the simulated SSTs from the observed SST (HadSST) distribution. Also shown are the departures of a second observed SST dataset from the first observed SST dataset as an estimate of the observational uncertainty. (a) HadSST, (b) NOAA extended reconstructed SST V3b (ER-SST), (c) ACCESS1.0 historical simu-lation, (d) ACCESS1.3 historical simulation, (e) ACCESS1.0 pre-industrial control simulation, and (f) ACCESS1.3 pre-industri-al control simulation. The shadings in plots (b−f) represent departures from HadSST shown in (a). The spatial correlation coefficients between the HadSST climatology and the other observed and simulated climatologies are shown on top of the respective plots.


downward) Ekman pumping velocity on the poleward sides of the maximum easterly τx biases in both hemispheres. The Z20 biases in both ACCESS models are small throughout the equatorial Pacific. Indeed, the spatial patterns of the bias are quite similar in all the simulations, with some minor differences only.

The vertical temperature profile in the equatorial Pacific is also of interest. Among other things, this reveals the sharpness or otherwise of the equatorial thermocline that affects ENSO dynamics. Figure 4 presents the depth profiles of equatorial Pacific potential temperatures (averaged over 5°S–5°N latitudes) as a function of longitudes. The temperatures from the coupled simulations (Fig. 4(c–f)) are compared with those from SODA (Fig. 4(a)) and PEODAS (Fig. 4(b)). The 20 °C isotherm, used as a proxy for the thermocline depth, is shown by the thick contour. The temperature biases in both models (shades) show a consistent pattern, with cold biases in the upper ocean and large warm subsurface biases in the eastern Pacific. The latter cause the thermocline to be deeper and more diffused than observed in the eastern Pacific (cf. Fig. 3), a common bias in some CGCMs (Yeh et al. 2010). Presumably, these biases result from a weaker upwelling velocity in the far eastern equatorial Pacific (Yeh et al. 2010)

can still significantly distort the simulated ENSO properties by contributing to the equatorial SST cold bias through enhanced upwelling and zonal advection. The SST cold bias, in turn, contributes to the double-ITCZ problem commonly seen in CGCMs (Lin 2007).

The thermocline depth is another key variable that participates in the dynamics of ENSO through interactions with SST and τx. Figure 3 shows the time-mean thermocline depth (contours), approximated by the depth of 20 °C isotherm (Z20), from the SODA and PEODAS ocean reanalyses (Fig. 3(a, b)) and the coupled simulations (Fig. 3(c–f)). The shades in Fig. 3(b–f) show deviations of the time-mean Z20 fields from those in SODA. In general, Z20 in PEODAS tends to be shallower than in SODA. The simulated Z20 is deeper than the observed Z20 over most of the equatorial zone (Fig. 3(c–f)); however, these model biases are comparable to the reanalysis uncertainty (Fig. 3(b)). The regions with large Z20 biases are located away from the equator in both hemispheres, as for τx. Indeed, the spatial pattern of Z20 biases corresponds very well with that of τx. The deeper-than-observed (i.e. positive) Z20 biases tend to occur on the poleward side of maximum easterly τx biases (cf. Fig. 2). This is consistent with the occurrence of negative (i.e.

Fig. 2. As in Fig. 1, but for the zonal wind stress (N m–2). Two recent observation-based air–sea flux products are used as the ob-servational estimates: (a) the TropFlux and (b) the CORE version 2. The full time-mean values are shown in contours, and the departures from observed values in shades, in (b-f).


annual cycle in COREv2 (Fig. 5(b)) has smaller amplitude than that in TropFlux (Fig. 5(a)) in the eastern Pacific. The broadscale pattern of the simulated SST annual cycle is consistent with the observed annual cycle. For example, the largest annual variations of SST occur in the eastern Pacific, with the minimum and maximum SSTs occurring during the (southern) spring and autumn seasons, respectively. The annual range (i.e. maximum–minimum) of simulated SSTs is somewhat smaller than in observations, and is more so for ACCESS1.3. The simulated SSTs are too cold during early autumn in the Niño4 region in both models. In particular, the east Pacific SST maximum is too weak in ACCESS1.3, which may affect the correct simulation of the observed ENSO phase locking to the annual cycle. Ham et al. (2012) suggest that a weak east Pacific SST maximum can affect the ENSO phase locking by enhancing the equatorial SST gradient, and shoaling the thermocline and reducing the thermal damping in the east Pacific.

As for SSTs, large annual variations for τx (contours) also occur in the eastern Pacific. Indeed, the two annual cycles closely correspond to each other, with the strongest easterly (westerly) τx located to the west of the SST minimum (maximum). Also, the strongest easterly τx precedes the

and are associated with the warm SST bias there (Fig. 1). The cold temperature bias in the upper ocean is associated with stronger equatorial upwelling (caused by the stronger wind stress) that contributes to the cold tongue bias at the surface (e.g. Guilyardi et al. 2009a).

Annual cycleIn addition to the mean state, the annual cycle in the tropical Pacific also has a close connection to the characteristics of ENSO. Figure 5 shows the climatological annual cycles of SST (shading) and τx (contours) in the equatorial Indian and Pacific Oceans. The variables were averaged over the 5°S–5°N latitude band and the respective annual means were removed. As before, the top two panels show the annual cycles from two observational SST datasets and the middle and bottom panels show the annual cycles, respectively, from the historical and pre-industrial control simulations. Note that all the panels now display the full values, except the annual means were removed. Also, unlike the previous figures, the observed annual cycle was not subtracted from the simulated annual cycles shown in Fig. 5(c–f). The SST annual cycles in the two observational datasets are similar, but with some differences in detail. For example, the τx

Fig. 3. As in Fig. 1, but for the depth of the 20 °C isotherm (m), used as a proxy for the thermocline depth. The ocean temperatures from the (a) Simple Ocean Data Assimilation (SODA) and (b) POAMA Ensemble Ocean Data Assimilation System (PEODAS) were used for calculating the thermocline depth, as observational estimates.


existence of significant biases in the modelled mean-state and annual cycle (e.g. Guilyardi et al. 2009a). These biases have been reduced considerably in the ACCESS-CM (see the previous section and Bi et al. 2013a) and, as a result, the ENSO characteristics in this model are expected to improve. Figure 6 shows the spatial structures of ENSO in observations and coupled simulations, as represented by regression patterns. The patterns were obtained by regressing the monthly SST anomalies upon the standardised Niño3 index, so the values represent the SST anomalies associated with a typical El Niño event. The Niño3 index was calculated by area-averaging the monthly SST anomalies in the Niño3 region (90°W−150°W and 5°N−5°S). In general, both the coupled models show realistic spatial patterns (Fig. 6(c–e)), with the highest SST anomalies confined in the eastern equatorial

occurrence of the SST minimum, as in observations. There are, however, some discrepancies with observations as well. For example, the modelled peak easterly τx tends to be somewhat stronger than the observational estimates. Moreover, in ACCESS1.3, the peak westerly τx during the austral autumn does not strengthen enough, which is likely to be associated with the weaker than observed SST maximum at that time.

El Niño–Southern Oscillation

The properties of simulated ENSOThe spatial structure and timescales of variability are among the most important features of ENSO. Historically, CGCMs had difficulties in simulating these features, partly due to the

Fig. 4. As in Fig. 3, but for the vertical profiles of mean potential temperatures (°C) of the equatorial Pacific Ocean. The tempera-tures were averaged over the 5°S–5°N latitude band. The thick contour shows the mean thermocline depth, as represented by the 20 °C isotherm.


Niño3 region in both models. Also, the meridional extent of the primary SST anomalies is still somewhat too narrow, a common bias exhibited by most CGCMs.

A realistic simulation of ENSO strength and its timescales of 3–7 years has also been a challenge for CGCMs: about 25 per cent of the CMIP3 models failed to show a spectral peak in this time range for their simulated ENSO (AchutaRao and Sperber 2006). Figure 7 presents power spectra of the Niño3 indices derived from observations and the ACCESS-CM simulations. The maximum power is in the 3–7 year range, closely matching the observations (Fig. 7 (a, b)). However, the ENSO variance in this range is somewhat weaker (stronger) than observed in ACCESS1.0 (ACCESS1.3). Nevertheless, for both models the observed spectra fall within the 95 per cent

Pacific. While the magnitudes tend to be somewhat larger than observed in the eastern part of the Niño3 region, the bias involving an excessive westward extension of the primary SST anomalies into the western equatorial Pacific has been reduced, especially in ACCESS1.0. The primary SST anomalies in ACCESS1.0 appear restricted slightly more than observed to the eastern equatorial Pacific, in contrast to the excessive westward extension of these anomalies into the western equatorial Pacific seen in many CGCMs (e.g. van Oldenborgh et al. 2005), and to some extent in ACCESS1.3. Another improvement is the better simulation of SST anomalies along the South American coast seen in both the models. Despite these improvements, the ENSO-induced SST variability is still too weak to the north of the

Fig. 5. The climatological annual cycles of equatorial SSTs (shades) and zonal wind stress (contours) in observations and coupled model simulations. The variables were averaged over the 5°S–5°N latitude band. Also, the respective annual means were removed from the computed annual cycles. (a) HadSST and TropFlux, (b) ER-SST and CORE.v2, (c) ACCESS1.0 historical simulation, (d) ACCESS1.3 historical simulation, (e) ACCESS1.0 pre-industrial control simulation, and (f) ACCESS1.3 pre-industrial control simulation. Units are °C for SST and N m–2 for wind stress. The vertical lines in each plot show the Niño3 and Niño4 regions.


century-long simulations can still be useful, provided caution is exercised as to their possible variations.

The temporal evolution of spectral power may be studied using the wavelet spectrum analysis (e.g. Torrence and Compo 1998). The wavelet spectra of the observed and simulated Niño3 indices are shown in Fig. 8. The spectra were adjusted following the suggestion of Liu et al. (2007) to facilitate comparison of spectral peaks at different frequencies. Statistically significant spectral peaks occur with realistic magnitudes. The statistical significance was assessed with respect to the background red-noise process (or the first-order autoregressive process). Also, the familiar decadal modulations of ENSO activity are evident throughout the simulation period. Even some hints of centennial modulations can be seen in the longer pre-industrial simulations (Fig. 8 (e, f)). Given such natural variations of ENSO power over decades and centuries, it is hard to detect the changes in ENSO power, if any, due to the increasing anthropogenic forcings over the historical period. An examination of ENSO wavelet spectra in simulations with stronger anthropogenic

confidence intervals for the simulated spectra, indicating consistency between the observed and simulated spectra.

Wittenberg (2009) demonstrates that the strength and periodicity of ENSO simulated in the GFDL-CM2.1 model varies significantly on centennial timescales. This may call into question the robustness of the detail of such results based on a single century of simulated ENSO. To test for such variability for ACCESS-CM, power spectra of Niño3 indices from the preindustrial control simulation are calculated for the full 500 years and separately for the successive 100-year segments. The results for ACCESS1.0 and ACCESS1.3 are shown in Fig. 7(c, d), respectively. As in Wittenberg’s (2009) analysis, the power spectra for individual 100-year segments show considerable variations (consistent with random sampling fluctuations); for example the lowest and highest spectral peaks among the five segments vary by a factor of two. However, most of the individual spectra approximate well the spectrum computed from the 500-year simulation; also, they are within the range of uncertainty shown in Fig. 7(a, b). This implies that power spectra estimated from

Fig. 6. Spatial structures of the observed and simulated ENSOs, as represented by the regression of the tropical Pacific SST anomalies (°C) upon the standardised Niño3 index. Monthly anomalies for all months are used in the calculation. The rect-angular box in the eastern Pacific shows the Niño3 region. (a) HadSST, (b) ER-SST, (c) ACCESS1.0 historical simulation, (d) ACCESS1.3 historical simulation, (e) ACCESS1.0 pre-industrial control simulation, and (f) ACCESS1.3 pre-industrial control simulation.


being weaker than observed and the minimum activity occurring in July. This deficiency in the seasonal phase locking of ENSO in ACCESS1.3 will be addressed during future model development.

We next show the growth and decay of typical ENSO events in observations and the coupled simulations (Fig. 10). These are illustrated by regressing the Niño3 SST time series for the preceding and following months upon the standardised Niño3 index for December only. Therefore, the numerical values have a unit of °C and express the magnitudes of Niño3 SST anomaly during the growth and decay phases of a typical ENSO event. The growth-decay curves for the two ACCESS1.0 simulations (blue and purple) have a similar shape as those for the observations (solid and dashed black curves). However, the ENSO growth and decay is slightly delayed compared to the observations. The deviations from observations are larger for ACCESS1.3 (red

forcings may be needed to determine the possible effect of anthropogenic forcings on future ENSO behaviour.

An outstanding characteristic of ENSO is that the SST anomaly in the central equatorial Pacific typically peaks around December (e.g. Rasmusson and Carpenter 1982). Many CMIP3 models did not represent this seasonal phase-locking of ENSO realistically (Joseph and Nigam 2006). The seasonal phase locking in ACCESS simulations is depicted by the annual variations of the monthly standard deviation of the Niño3 indices (Fig. 9). In general, the annual variation of ENSO activity is simulated in both versions of the ACCESS-CM with a varying degree of accuracy. As in observations, the ENSO activity in ACCESS1.0 peaks in December, with a comparable variation over the calendar year. However, the minimum activity occurs two months later than in observations. On the other hand, ACCESS1.3 shows the maximum ENSO activity in March, with the annual variation

Fig. 7. Power spectra of the observed and simulated Niño3 indices. (a) ACCESS1.0 historical simulation, (b) ACCESS1.3 historical simulation, (c) ACCESS1.0 pre-industrial control simulation, and (d) ACCESS1.3 pre-industrial control simulation. The spec-tra were obtained from the periodograms by applying a 19-point boxcar smoother. The two black curves in (a) and (b) rep-resent the observed spectra computed from HadSST and ER-SST, and the data for 1870–2005 were used for both the ob-served and simulated spectra. The two blue dashed curves in each of these plots show the 95 per cent lower and upper confidence limits for the simulated spectra, estimated from the chi-squared distribution. The black solid curve in (c) and (d) represents the power spectrum from each 500-yr long pre-industrial control simulation. The coloured dashed curves are the power spectra computed from the five successive 100-year segments of the pre-industrial control simulation. Unit is °C2/(cycles/mon).


Fig. 8. Wavelet spectra of the observed and simulated Niño3 indices. (a) HadSST, (b) ER-SST, (c) ACCESS1.0 historical simulation, (d) ACCESS1.3 historical simulation, (e) ACCESS1.0 pre-industrial control simulation, and (f) ACCESS1.3 pre-industrial con-trol simulation. The Morlet wavelet was used as the basis function. The thick contour encloses the wavelet power significant at the 95 per cent level (with respect to the background red-noise process), and the hatched regions at both ends show the 'cone of influence'. The wavelet spectra are scaled by timescales to facilitate comparison of spectral peaks across timescales.

Fig. 9. Annual variation of the monthly standard deviations (°C) of the observed and simulated Niño3 indices.

Fig. 10. Growth and decay of the observed and simulated Niño3 SST anomalies (°C) before and after the ENSO peaks (here assumed to occur in December).


observationally derived regression patterns for SST and τx

(Fig. 11(a, b); see also Guilyardi et al. 2009b) show that the SST anomalies (shading) develop contemporaneously along all longitudes, with the strongest anomalies in the eastern Pacific. There is no clear sign of a zonal propagation of the SST anomalies. The SST anomalies in ACCESS1.0 develop in the same fashion as observed, while those in ACCESS1.3 show westward propagations during the ENSO decay phase. In both models, however, strong SST anomalies are induced over a wider region than in observations. This is mainly due to a higher correlation between the Niño3 index and the SST anomalies in and surrounding the Niño3 region in the models than in the observations.

The regressed τx field (contours) from observations shows the familiar pattern of westerly wind stress appearing to the west of the Niño3 region (Fig. 11(a, b)). In contrast to the

and green curves), especially during the growth phase and the early part of the decay phase. These deviations may be attributed to the error in ENSO phase locking in this model, for which the highest SST anomaly occurs in March instead of December.

According to simple theoretical models of ENSO (Suarez and Schopf 1988, Battisiti and Hirst 1989, Jin 1997), the zonal wind stress (τx) and the thermocline depth (Z20) in the tropical Pacific play a key role in determining the ENSO-related SST evolution. It is of interest to see how these coupled variables evolve through the ENSO developing and decay phases in the model simulations. Figures 11 and 12 show the lag-regressions of latitudinally-averaged (5°S—5°N) SST, zonal wind stress and thermocline depth anomalies at each longitude and the standardised Niño3 index. Unlike in Fig. 10, all months were included in the calculations here. The

Fig. 11. Lag-regressions of the equatorial SST (shades) and zonal wind stress (contours) anomalies on the standardised Niño3 in-dices. (a) HadSST and TropFlux, (b) ER-SST and CORE.v2, (c) ACCESS1.0 historical simulation, (d) ACCESS1.3 historical simulation, (e) ACCESS1.0 pre-industrial control simulation, and (f) ACCESS1.3 pre-industrial control simulation. The verti-cal lines in each plot show the Niño3 and Niño4 regions. Units are °C for SST and N m–2 for wind stress.


(shading). The Z20 regressions in observations (Fig. 12(a, b)) are characterised by a predominantly dipole pattern, with positive Z20 anomalies (i.e. a deeper thermocline) collocated with the strongest positive SST anomalies in the eastern Pacific. Negative Z20 anomalies coincide with weak negative SST anomalies in the western Pacific. While the SST and Z20 anomalies evolve contemporaneously in the eastern Pacific, the negative Z20 anomalies in the western Pacific lag the Niño3 SST anomalies by up to four months. Consistent with this is the occurrence of the weak negative SST anomalies there; as the mean thermocline is deeper in the western Pacific, the thermocline anomaly only weakly affects the SSTs there. The Z20 regression pattern in the coupled simulations is broadly similar to that found in observations, although there are differences in details. For example, the positive anomaly pole in the eastern Pacific is of smaller size than in the observations and the negative anomaly pole is somewhat shifted east into the Niño3 region. Also, the weak negative SST anomalies in the western Pacific are not as well defined

SST anomalies, the τx anomalies show zonal propagations: the anomalies first appear in the western-central Pacific and then propagate eastward to the central-eastern Pacific as the SST anomalies strengthen. This eastward propagation is also seen in ACCESS1.0 (Fig. 11(c, e)); however, the regressed τx anomalies in ACCESS1.3 show a westward propagation, as for the SST anomalies in this model. In addition, the regressed τx anomalies in both the models are much weaker than in observations. The occurrence of such a weak τx

response to ENSO (or dynamic feedback) in the models is surprising, given that the simulated ENSO amplitudes are comparable to the observed amplitude. One possibility is that the ENSO thermodynamic feedback, associated with the surface heatflux response to ENSO, is also weak in the models, thereby compensating for the weaker than observed dynamic feedback.

Figure 12 shows the similar regression patterns as above, but for the Z20 field (contours). For ease of comparison, the SST regressions from Fig. 11 are reproduced in Fig. 12

Fig. 12. As in Fig. 11, but for SST (°C; shades) and the thermocline depth (m; contours). The thermocline depth data used in (a) and (b) are from SODA and PEODAS, respectively.


presents the atmospheric circulation response to ENSO, in the form of regression patterns of the surface horizontal (vectors) and zonal (contours) winds and the pressure velocity at the 500 hPa level (shading) regressed upon the standardised monthly Niño3 index. As evident from the observations (Fig. 13(a)), the deep convection region of the western Pacific is shifted towards the central Pacific during the ENSO warm phase, creating anomalous ascending motions (and surface wind convergence) around the dateline and anomalous descending motions (and surface wind divergence) in the far western Pacific. This east–west dipole pattern of vertical motion is accompanied by anomalous westerlies in the western Pacific, to the west of ascending motions, and anomalous easterlies in the Indian Ocean, to the west of descending motions. The convective region extends to the eastern Pacific along the north of the equator and along the South Pacific Convergence Zone (SPCZ). Elsewhere in the tropics, there are weak descending motions over the southwest and northwest Pacific and over the tropical Atlantic Ocean.

The ENSO-induced atmospheric circulation anomalies in the ACCESS1.0 AMIP simulation are presented in Fig. 13(b).

in the model simulations as in the observations, which could result from a combination of model errors in the thermocline depth, wind stress, and the heatflux in that region.

In summary, we have compared with observations the salient properties of the simulated ENSO, including its spatial structure, amplitude, timescales of variability, seasonal phase locking, growth and decay, and the evolution of the coupled variables during ENSO. Overall, these prominent properties of ENSO seem to be reasonably well simulated in the two versions of ACCESS-CM, although version 1.0 tends to perform better than version 1.3 on the measures examined here. However, there is still much scope for improving the simulated ENSO, and the remaining challenges include a better simulation of the wind stress and thermocline depth and their interactions with tropical Pacific SST anomalies.

ENSO teleconnectionsIt is well known that ENSO has a significant climatic impact on remote regions through atmospheric teleconnections (e.g. Trenberth et al. 1998). Here, we look at the ENSO impact on some of the important climate variables in model simulations and compare that with observations. Figure 13

Fig. 13. The monthly anomalies of 'observed' and simulated surface horizontal winds (vectors), zonal winds (contours), and the pressure vertical velocity (shades) regressed upon the standardized Niño3 indices. Only the tropical domain is shown. (a) ERA-Interim, (b) ACCESS1.0 AMIP simulation, (c) ACCESS1.0 historical simulation, (d) ACCESS1.3 historical simulation, (e) ACCESS1.0 pre-industrial control simulation, and (f) ACCESS1.3 pre-industrial control simulation. In (a–d), data for 1979–2005 are used; for (e, f), data for an equivalent period are used. The vector scale is shown at the top-right corner of the plots. Units are m s–1 for horizontal winds and Pa s–1 for pressure vertical velocity.


and is seen in all the coupled simulations. Another significant difference is the area of dry rainfall anomaly in the western Pacific, which is a prominent feature of the observed rainfall anomalies, but is weak in the coupled simulations, especially in ACCESS1.0. To some extent, this deficiency is also seen in the AMIP simulation, indicating that the bias may have originated from deficiencies in the atmospheric model physics. Finally, the positive rainfall anomalies in the central equatorial region in the coupled simulations are significantly weaker than observed, a consequence of the cold bias of simulated SSTs in that region.

The analysis presented above emphasises the tropical region. However, there is also significant ENSO teleconnections in the extratropics, which may be suitably illustrated by the sea-level pressure field (Fig. 15). As before, regression results from an AMIP simulation (Fig. 15(b)) and four coupled simulations (Fig. 15(c–f)) are compared with the result from reanalysis. In the tropics, the warm phase of ENSO is associated with higher (lower) than normal pressure in the western Pacific and the Indian Ocean (central and eastern Pacific). In the extratropics, the notable features include a strong anomalous low pressure system in the north Pacific and a pronounced anomalous high pressure system in the far southeastern Pacific (Fig. 15(a)). These anomalous features are well captured by the AMIP simulation (Fig. 15(b)), which is forced by the observed SST and sea-ice

The large-scale patterns of vertical velocity and surface winds are similar to those for observations. However, the response for the vertical velocity (zonal wind) is somewhat stronger (weaker) than observed. In the coupled models (Fig. 13(c–f)) the large-scale patterns are also reasonably well simulated, although the magnitude of response in this case is weaker than observed for both the vertical velocity and horizontal winds. Also note that the weak zonal wind response in the Indian Ocean is seen in both the AMIP and coupled simulations, indicating that the atmospheric model could be the source of this bias.

The ENSO impact on rainfalls is important, as this is related to the occurrence of droughts and floods in many regions, including Australia. Figure 14 shows the El Niño-related rainfall anomalies as the regression coefficients of rainfall on the standardised Niño3 index for the observation (Fig. 14(a)), an AMIP simulation (Fig. 14(b)), and the coupled simulations (Fig. 14(c–f)). The rainfall anomaly pattern in the AMIP simulation is similar to that observed, with several differences, the most striking of which is an east–west oriented dry anomaly region to the north of the equator. For the coupled simulations, the large-scale rainfall anomaly pattern agrees reasonably with the observed pattern. However, the magnitude of rainfall response is weaker overall for all the coupled model simulations. Also, the dry anomaly region to the north of the equator is now more pronounced

Fig. 14. As in Fig. 13, but for rainfall (mm day–1). The rectangular box in the Pacific shows the Niño3 region.


westward shift of the negative height anomalies in the north Pacific (Fig. 16(c–f)).

ENSO teleconnection with Australian rainfallENSO is a major driver of interannual rainfall variability in Australia (e.g. McBride and Nicholls 1983, Wang and Hendon 2007, Risbey et al. 2009). It is therefore of interest to examine the representation of this important aspect of Australian climate in ACCESS models. In Fig. 17, we present the correlation between the Niño3.4 index and rainfall anomalies in the Australian region. Note that we use the Niño3.4 index for Australian teleconnections, as the Niño3.4 index tends to show a higher correlation with Australian climate variability than for the Niño3 index (e.g. Wang and Hendon 2007, Risbey et al. 2009). The analysis was performed separately for four standard seasons (December–January–February, March–April–May, June–July–August, and September–October–November (SON)) and the annual means. Here, we show only the results for the spring season (SON), as the ENSO impact on Australian climate is largest during this season. The rainfall teleconnection in the ACCESS1.0 and ACCESS1.3 historical experiments (Fig. 17(b, c)) is compared with that in the Bureau of Meteorology rainfall dataset (Fig. 17(a)); the rainfall teleconnection in the AMIP simulation with an earlier version of the ACCESS atmosphere model was investigated by Risbey et al. (2011).

distributions. The coupled models also do a reasonable job in simulating the ENSO-induced pressure patterns, with some exceptions. The latter include a somewhat stronger extratropical response in the northern hemisphere (NH), with a large high-pressure response over the Arctic in the ACCESS1.3 simulation (Fig. 15(f)). The large Arctic response in ACCESS1.3 appears to be part of a generally stronger extratropical response to ENSO shown by this model.

Figure 16 shows the ENSO teleconnection in the upper troposphere using the regressed 200 hPa height field. In the tropics, the height anomalies take the form of a pair of anticyclones (cyclones) over the central-eastern Pacific (the western Pacific and eastern Indian Ocean), on both sides of the equator. These upper tropospheric anomalies are broadly of opposite signs to the surface anomalies, consistent with a baroclinic response to the ENSO-induced heat sources in the tropics (Gill 1980). In the extratropics, the upper tropospheric height anomalies have the same sign as for the surface anomalies, consistent with an equivalent barotropic response there. For the observation (Fig. 16(a)), the familiar Pacific-North American (PNA) pattern and the Pacific-South American (PSA) pattern are apparent. These prominent ENSO teleconnection patterns are also captured well by the models. In the AMIP simulation (Fig. 16(b)), the tropical (extratropical) response is somewhat weaker (stronger) than observed. A notable difference for the coupled models is a

Fig. 15. Same as Fig. 13, but for sea level pressure (hPa) on the global domain.


Fig. 16. Same as Fig. 15, but for the 200 hPa geopotential height (m) field.

Fig. 17. Correlation coefficients between the Niño3.4 indices and the rainfall anomalies in the Australian region for the spring (SON) season. The seasonal mean observed and simulated data for 1906−2005 are used in the analysis. (a) Bureau of Meteorology rainfall data, (b) ACCESS1.0 historical simulation, and (c) ACCESS1.3 historical simulation.


simulated ENSO timescale is also realistic in both models, with the maximum ENSO power being in the 3–7 year band as observed. The ENSO phase locking to the annual cycle is a robust observed feature of ENSO; this is simulated very well by one of the ACCESS models (ACCESS1.0). The ENSO growth and decay tend to be somewhat delayed in the ACCESS models; however, overall these features are also well simulated. The success of simulating the time-lagged evolutions of SST and zonal wind stress anomalies during ENSO is mixed: the spatial patterns are realistic, but the response of SST (zonal wind stress) to ENSO is too strong (weak) in the simulations. Moreover, the SST anomalies in ACCESS1.3 show a tendency of westward propagations. While no such propagation is seen in the regression pattern derived from observations for 1870–2005, westward (eastward) propagating ENSO signals were observed before (after) the 1977 climate shift (Guan and Nigam 2008). The simulated response of the thermocline depth is somewhat more realistic in ACCESS1.0 than in ACCESS1.3. The simulated ENSO teleconnection patterns show a close correspondence with observed patterns, especially in the extratropics. In the tropics, while the teleconnections are realistic, the impact of the equatorial cold bias is seen in precipitation and vertical velocity responses, contributing to the weaker than observed responses there.

Despite these deficiencies, the ACCESS models perform better in simulating key ENSO properties and teleconnections than most CMIP3 models (cf. AchutaRao and Sperber 2006, Joseph and Nigam 2006). For example, Joseph and Nigam (2006) examined the ENSO evolution and teleconnections in six CGCM simulations from the CMIP3 archive: GFDL CM2.1, GISS-EH, CCSM3, PCM, HadCM3, and MIROC3.2 (hires). Most of these models showed strong biases in some of the ENSO properties examined above; e.g. only two of these six CMIP3 models simulated a realistic spatial structure and ENSO phase locking (HadCM3 and CCSM3), although the ENSO temporal evolution in CCSM3 was too quick and periodic. Likewise, the simulated teleconnections are also better in the ACCESS models than in many CMIP3 models.

While these improvements in ENSO simulations indicate a clear progress from the previous generation models, there are still significant deficiencies in the ACCESS-CM that need to be addressed as part of the overall model development effort. One obvious deficiency that has to be understood is the weaker than observed zonal wind stress response to ENSO in both versions of ACCESS-CM. Also, imperfections in the seasonal ENSO phase locking, especially in ACCESS1.3, need to be investigated. We plan further data analysis and atmosphere and coupled model experiments to address these issues in future.

Acknowledgments

We thank two anonymous reviewers for providing insightful comments on this paper. This work has been undertaken as part of the Australian Climate Change Science Program,

All results shown below are for a recent one hundred year period, 1906–2005.

The observed ENSO teleconnection with spring rainfall shows the familiar pattern, with the largest dry anomalies located in the northern and eastern parts of Australia (Fig. 17(a)). There are small dry anomalies over the central part of Western Australia and larger anomalies elsewhere, including the southwestern tip. The rainfall teleconnection simulated in the historical experiments shows many of the prominent teleconnection features in the observed dataset, including the larger dry anomalies in the northern, eastern and southeastern Australia. However, there are also significant deviations from the observed pattern. For ACCESS1.0, these include the overall weakness of the teleconnection strength, and the positive correlations over much of Western Australia and some coastal regions of northern Australia (Fig. 17(b)). The main bias in ACCESS1.3 appears to be the weakness of the teleconnection signal over eastern Australia (Fig. 17(c)).

The weaker than observed teleconnection between ENSO and Australian rainfall in ACCESS1.0 is perhaps expected, given the presence of the equatorial Pacific cold bias and also that the simulated SSTs in this model tend to be colder (by 0.12 °C in the Niño3.4 region) than those in ACCESS1.3. While the spurious positive correlations in Western Australia are not straightforward to explain, it appears that the continent-wide teleconnection pattern is somewhat biased towards positive correlations in ACCESS1.0. The coupled simulations represent only a single ensemble member from each model; inclusion of additional ensemble members (which are not yet available) in future analyses is expected to improve the statistical significance of the overall rainfall teleconnection pattern over Australia. Given that a realistic simulation of regional rainfall variability has remained among the most significant challenges for the state-of-the-art CGCMs (Randall et al. 2007), the ACCESS model performance in capturing the ENSO-Australian rainfall teleconnection may be judged as reasonably good.

Conclusions

In this work, we have investigated the realism of simulated ENSO in ACCESS simulations performed for participating in the CMIP5. Using the pre-industrial control and historical simulations of ACCESS-CM versions 1.0 and 1.3, we compared with observations the fundamental ENSO properties, such as the spatial structure and timescales, seasonal phase locking, growth and decay, and coupled ENSO evolutions, as well as the ENSO induced teleconnection patterns. Our analysis shows that both versions of the ACCESS-CM perform reasonably well in simulating the ENSO properties and teleconnections when compared with observations.

The simulated ENSO spatial pattern is found to be realistic in both ACCESS1.0 and ACCESS1.3. In particular, a commonly seen bias of high ENSO related SST variance in the western Pacific is absent in ACCESS1.0 simulations. The


Fedorov, A.V., and Philander, S.G. 2001. A stability analysis of tropical ocean–atmosphere interactions: Bridging measurements and theory for El Niño. J. Clim., 14, 3086–101.

Gill, A.E. 1980. Some simple solutions for heat-induced tropical circula-tion. Q. J. R. Meteorol. Soc., 106, 447–62.

Gu, D. and Philander, S.G.H. 1995. Secular changes of annual and inter-annual variability in the Tropics during the past century. J. Clim., 8, 864–76.

Guan, B. and Nigam, S. 2008. Pacific Sea Surface Temperatures in the Twen-tieth Century: An Evolution-Centric Analysis of Variability and Trend. J. Clim., 21, 2790–809. doi: http://dx.doi.org/10.1175/2007JCLI2076.1

Guilyardi, E., S. Gualdi, J. Slingo, A. Navarra, P. Delecluse, J. Cole, G. Madec, M. Roberts, M. Latif, and L. Terray. 2004. Representing El Niño in Coupled Ocean–Atmosphere GCMs: The Dominant Role of the At-mospheric Component. J. Clim., 17, 4623–29.

Guilyardi, E., A. Wittenberg, A. Fedorov, M. Collins, C. Wang, A. Capo-tondi, G. J. van Oldenborgh, and T. Stockdale. 2009a. Understanding El Niño in Ocean–Atmosphere General Circulation Models: Progress and Challenges. Bull. Am. Meteorol. Soc., 90, 325–40. doi: http://dx.doi.org/10.1175/2008BAMS2387.1

Guilyardi, E, Braconnot, P., Jin, F.-F., Kim, S.T., Kolasinski, M., Li, T. and Musat, I. 2009b. Atmosphere Feedbacks during ENSO in a Coupled GCM with a Modified Atmospheric Convection Scheme. J. Clim., 22, 5698–718. doi: http://dx.doi.org/10.1175/2009JCLI2815.1

Ham, Y.G., Kug, J.S., Kim, D., Kim, Y.H., and Kim, D.H. 2012. What con-trols phase-locking of ENSO to boreal winter in coupled GCMs?. Clim. Dyn., 1–18. DOI: 10.1007/s00382-012-1420-2.

Hewitt, H.T., D. Copsey, I.D. Culverwell, C.M. Harris, R.S.R. Hill, A.B. Keen, A.J. McLaren, and E.C. Hunke. 2011. Design and implementa-tion of the infrastructure of HadGEM3: the next-generation Met Office climate modelling system. Geosci. Model Dev., 4, 223–53, doi:10.5194/gmd-4-223-2011.

Huffman, G.J., Adler, R.F., Bolvin, D.T., and Gu, G. 2009. Improving the global precipitation record: GPCP Version 2.1. Geophys. Res. Lett., 36, L17808, doi:10.1029/2009GL040000.

Jin, F.-F. 1997. An Equatorial Ocean Recharge Paradigm for ENSO. Part I: Conceptual Model. J. Atmos. Sci., 54, 811–29. doi: http://dx.doi.org/10.1175/1520-0469(1997)054<0811:AEORPF>2.0.CO;2

Joseph, R., and Nigam, S. 2006. ENSO evolution and teleconnections in IPCC’s twentieth-century climate simulations: Realistic representa-tion? J. Clim., 19, 4360–77.

Kirtman, B.P. and Schopf, P.S. 1998. Decadal Variability in ENSO Predict-ability and Prediction. J. Clim., 11, 2804–22.

Kowalczyk, E.A., Wang, Y.P., Law, R.M., Davies, H.L., McGregor, J.L. and Abramowitz, G. 2006. The CSIRO Atmosphere Biosphere Land Ex-change (CABLE) model for use in climate models and as an offline model. CSIRO Marine and Atmospheric Research paper 013, www.cmar.csiro.au/e-print/open/kowalczykea_2006a.pdf

Kowalczyk, E.A., L. Stevens, R.M. Law, M. Dix, Y.P. Wang, I.N. Harman, K. Hayens, J. Srbinovsky, B. Pak, and T. Zhien. 2013. The land surface model component of ACCESS: description and impact on the simu-lated surface climatology. Aust. Met. Oceanogr. J., 63, 65–82.

Large, W.G., McWilliams, J.C., and Doney, S.C. 1994. Oceanic vertical mixing: A review and a model with a nonlocal boundary layer param-eterization. Reviews of Geophysics, 32, 363–403.

Large, W. and Yeager, S. 2009. The global climatology of an interannu-ally varying air-sea flux data set. Clim. Dyn., 33, 341–64, doi:10.1007/s00382-008-0441-3.

Lin, J.-L. 2007, The Double-ITCZ Problem in IPCC AR4 Coupled GCMs: Ocean–Atmosphere Feedback Analysis. J. Clim., 20, 4497–525. doi: http://dx.doi.org/10.1175/JCLI4272.1

Liu, Y., Liang, X.S., and Weisberg, R.H. 2007. Rectification of the Bias in the Wavelet Power Spectrum. J. Atmos. Oceanic Technol., 24, 2093–102. doi: http://dx.doi.org/10.1175/2007JTECHO511.1

Magnusson, L., Alonso-Balmaseda, M., and Molteni, F. 2012. On the dependence of ENSO simulation on the coupled model mean state. Clim. Dyn., DOI: 10.1007/s00382-012-1574-y.

Marsland, S.J., D. Bi, P. Uotila, R. Fiedler, S. M. Griffies, K. Lorbacher, S. O’Farrell, A. Sullivan, P. Uhe, X. Zhou. 2013. Evaluation of the ocean performance of the ACCESS Climate Model in CMIP5. Aust. Met. Oceanogr. J., 63, 101–19.

funded jointly by the Department of Climate Change and Energy Efficiency, the Bureau of Meteorology and CSIRO. The simulations were performed at the NCI National Facility at the Australian National University. We thank Dr. Harry Hendon for his comments on an early draft of this paper, and we thank the members of the coupled modelling team at CAWCR for their roles leading to the ACCESS-CM CMIP5 contributions. The authors thank Dr. Lawrence Rikus and Mr. Ben Hu for downloading and maintaining an archive of reanalysis data, some of which are used in this work.

ReferencesAchutaRao, K.M. and Sperber, K.R. 2006. ENSO Simulation in Coupled

Ocean-Atmosphere Models: Are the Current Models Better? Clim.Dyn., Vol. 27, 1–15, DOI: 10.1007/s00382-006-0119-7

Adler, R.F., Huffman, G.J., Chang, A., Ferraro, R., Xie, P.-P., Janowiak, J., Rudolf, B., Schneider, U., Curtis, S., Bolvin, D., Gruber, A., Susskind, J., Arkin, P., and Nelkin, E. 2003. The Version-2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis (1979–Present). J. Hydrometeor., 4, 1147–67.

Battisti, D.S., and Hirst, A.C. 1989. Interannual variability in the tropical atmosphere-ocean system: Influence of the basic state and ocean ge-ometry. J. Atmos. Sci. 46, 1687–712.

Bi, D., Dix, M., Marsland, S.J., O’Farrell, S., Rashid, H.A., Uotila, P., Hirst,A.C. , Kowalczyk, E., Golebiewski, M., Sullivan, A., Yan, H., Hanna, N., Franklin, C., Sun, Z., Vohralik, P., Watterson, I., Zhou, X., Fiedler, R., Collier, M., Ma, Y,. Noonan, J., Stevens, L., Uhe, P., Zhu, H., Hill, R., Harris, C., Griffies, S., and Puri, K. 2013a. The ACCESS Coupled Model: Description, Control Climate and Preliminary Valida-tion. Aust. Met. Oceanog. J., 63, 41–64.

Bi, D., S.J. Marsland, P. Uotila, S. O’Farrell, R. Fielder, A. Sullivan, S.M. Griffies, X. Zhou, and A.C. Hirst. 2013b. ACCESS-OM, the Ocean-Sea Ice Core of the ACCESS Coupled Model. Aust. Met. Oceanog. J., 63, 213–32.

Carton, J. A., G. Chepurin, and X. Cao. 2000. A Simple Ocean Data Assim-ilation Analysis of the Global Upper Ocean 1950–95. Part II: Results. J. Phys. Oceanogr., 30, 311–26. doi: http://dx.doi.org/10.1175/1520-0485(2000)030<0311:ASODAA>2.0.CO;2

Collins, W.J., N. Bellouin, M. Doutriaux-Boucher, N. Gedney, T. Hinton, C. D. Jones, S. Liddicoat, G. Martin, F. O’Connor, J. Rae, C. Senior, I. Tot-terdell, S. Woodward, T. Reichler, and J. Kim. 2008. Evaluation of the HadGEM2 model. Met Office Hadley Centre Technical Note no. HCTN 74, available from Met Office, FitzRoy Road, Exeter EX1 3PB http://www.metoffice.gov.uk/publications/HCTN/index.html

Dee, D.P., S.M. Uppala, A.J. Simmons, P. Berrisford, P. Poli, S. Kobayashi, U. Andrae, M.A. Balmaseda, G. Balsamo, P. Bauer, P. Bechtold, A.C.M. Beljaars, L. van de Berg, J. Bidlot, N. Bormann, C. Delsol, R. Dragani, M. Fuentes, A.J. Geer, L. Haimberger, S.B. Healy, H. Hersbach, E.V. Hólm, L. Isaksen, P. Kållberg, M. Köhler, M. Matricardi, A.P. McNally, B.M. Monge-Sanz, J.-J. Morcrette, B.-K. Park, C. Peubey, P. de Rosnay, C. Tavolato, J.-N. Thépaut, and F. Vitart. 2011. The ERA-Interim reanal-ysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc., 137, 553–97.

DelSole, T. and Shukla, J. 2010. Model fidelity versus skill in seasonal fore-casting. J. Clim., 23, 4794–806.

DiNezio, P.N., Clement, A.C., Vecchi, G.A., Soden, B.J., Kirtman, B.P., Lee, S.-K. 2009. Climate Response of the Equatorial Pacific to Global Warming. J. Clim., 22, 4873–92.

Dix, M., P. Vohralik, D. Bi, H. A. Rashid, S.J. Marsland, S. O’Farrell, P. Uotila, A.C. Hirst, E. Kowalczyk, A. Sullivan, H. Yan, C. Franklin, Z. Sun, I. Watterson, M. Collier, J. Noonan, L. Rotstayn, L. Stevens, P. Uhe and K. Puri. 2013. The ACCESS Coupled Model: Documentation of core CMIP5 simulations and initial results. Aust. Met. Oceanogr. J., 63, 83–99.

Essery, R.L.H., Best, M.J., Betts, R.A., Cox, P.M., and Taylor, C.M. 2003. Ex-plicit representation of sub-grid heterogeneity in a GCM land-surface scheme. J. Hydrometeor., 4, 530–43.


Trenberth, K.E., Branstator, G.W., Karoly, D., Kumar, A., Lau, N.-C. and Ropelewski, C. 1998. Progress during TOGA in understanding and modeling global teleconnections associated with tropical sea surface temperatures. J. Geophys. Res., 103, 14291–324.

Turner A.G, Inness, P.M., and Slingo, J.M. 2005. The role of the basic state in the ENSO-monsoon relationship and implications for predictability. Q. J. R. Meteorol. Soc., 131, 781–804

Uotila, P., O’Farrell, S. Marsland, S.J. and Bi, D. 2013. The sea-ice perfor-mance of the Australian climate models participating in the CMIP5. Aust. Met. Oceanogr. J., 63, 121–43.

Valcke, S. 2006. OASIS3 User Guide (prism_2-5). CERFACS Technical Re-port TR/CMGC/06/73, PRISM Report No 3, Toulouse, France. 60 pp.

van Oldenborgh, G.J., Philip, S. and Collins, M. 2005. El Niño in a chang-ing climate: A multi-model study. Ocean Sci., 1, 81–95.

Walters, D.N., M.J. Best, A.C. Bushell, D. Copsey, J.M. Edwards, P.D. Falloon, C.M. Harris, A.P. Lock, J.C. Manners, C.J. Morcrette, M.J. Roberts, R.A. Stratton, S. Webster, J.M. Wilkinson, M.R. Willett, I.A. Boutle, P.D. Earnshaw, P.G. Hill, C. MacLachlan, G.M. Martin, W. Moufouma-Okia, M.D. Palmer, J.C. Petch, G.G. Rooney, A.A. Scaife, and K.D. Williams. 2011. The Met Office Unified Model Global Atmo-sphere 3.0/3.1 and JULES Global Land 3.0/3.1 configurations. Geosci. Model Dev., 4, 919–41, doi:10.5194/gmd-4-919-2011.

Wang, G., and Hendon, H.H. 2007. Sensitivity of Australian Rainfall to Inter–El Niño Variations. J. Clim., 20, 4211–26. doi: http://dx.doi.org/10.1175/JCLI4228.1

Wilson, D.R., Bushell, A.C., Kerr-Munslow, A.M., Price, J.D. and Mor-crette, C.J. 2008. PC2: A prognostic cloud fraction and condensation scheme. I: Scheme description. Q. J. R. Meteorol. Soc., 134, 2093–107.

Wittenberg, A.T. 2009. Are historical records sufficient to constrain ENSO simulations? Geophys. Res. Lett., 36, L12702, doi:10.1029/2009GL038710.

Yeh, S.-W., Dewitte, B., Yim, B.Y. and Noh, Y. 2010. Role of the upper ocean structure in the response of ENSO-like SST variability to global warm-ing. Clim. Dyn., 35, 355–69, doi: 10.1007/s00382-010-0849-4.

Yin, Y., Alves, O. and Oke, P. 2011. An ensemble ocean data assimilation system for seasonal prediction. Mon. Weather Rev., 139, 786–808.

HadGEM2 Development Team: Martin, G.M., N. Bellouin, W.J. Collins, I.D. Culverwell, P.R. Halloran, S.C. Hardiman, T.J. Hinton, C.D. Jones, R.E. McDonald, A.J. McLaren, F.M. O’Connor, M.J. Roberts, J.M. Rodri-guez, S. Woodward, M.J. Best, M.E. Brooks, A.R. Brown, N. Butchart, C. Dearden, S.H. Derbyshire, I. Dharssi, M. Doutriaux-Boucher, J.M. Edwards, P.D. Falloon, N. Gedney, L.J. Gray, H.T. Hewitt, M. Hobson, M.R. Huddleston, J. Hughes, S. Ineson, W.J. Ingram, P.M. James, T.C. Johns, C.E. Johnson, A. Jones, C.P. Jones, M.M. Joshi, A.B. Keen, S. Liddicoat, A.P. Lock, A.V. Maidens, J.C. Manners, S.F. Milton, J.G.L. Rae, J.K. Ridley, A. Sellar, C.A. Senior, I.J. Totterdell, A. Verhoef, P.L. Vidale, and A. Wiltshire. 2011. The HadGEM2 family of Met Office Unified Model climate configurations. Geosci. Model Dev., 4, 723–57, doi:10.5194/gmd-4-723-2011.

McBride, J. and Nicholls, N. 1983. Seasonal relationships between Aus-tralian rainfall and the Southern Oscillation. Mon. Weather Rev., 111, 1998–2004.

Murray, R.J. 1996. Explicit generation of orthogonal grids for ocean mod-els. J. Comput. Phys., 126, 251–73.

Penland, C. and Sardeshmuhk, P. 1995. The optimal growth of the tropical sea surface temperature anomalies. J. Clim., 8, 1999–2024.

Philander, S.G. 1990. El Niño, La Niña, and the Southern Oscillation. Aca-demic Press, International Geophysics Series Vol. 46, 293 pp.

Praveen Kumar, B., Vialard, J., Lengaigne, M., Murty, V.S.N., McPhaden, M.J. 2012. TropFlux: air-sea fluxes for the global tropical oceans—de-scription and evaluation. Clim. Dyn., 38, 1521–43. DOI 10.1007/s00382-011-1115-0.

Randall, D.A., R.A. Wood, S. Bony, R. Colman, T. Fichefet, J. Fyfe, V. Katt-sov, A. Pitman, J. Shukla, J. Srinivasan, R.J. Stouffer, A. Sumi, and K.E. Taylor. 2007. Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 589–662.

Rashid, H.A., Hirst, A.C. and Dix, M. 2013. Atmospheric circulation fea-tures in the ACCESS model simulations for CMIP5: Historical simula-tion and future projections. Aust. Met. Oceanogr. J., 63, 145–60.

Rasmussen, E.M., and Carpenter, T.H. 1982. Variations in tropical sea sur-face temperature and surface wind fields associated with the South-ern Oscillation/El Niño. Mon. Weather Rev., 110, 354–84.

Rayner N.A., D.E. Parker, E.B. Horton, C.K. Folland, L.V. Alexander, D.P. Rowell, E.C. Kent, A. Kaplan. 2003. Global analyses of sea surface tem-perature, sea ice, and night marine air temperature since the late nine-teenth century. J Geophys. Res, 108, doi:10.1029/2002JD002670

Reichler, T., and Kim, J. 2008. How Well Do Coupled Models Simulate To-day’s Climate? Bull. Am. Meteorol. Soc., 89, 303–11. doi: http://dx.doi.org/10.1175/BAMS-89-3-303

Risbey, J.S., Pook, M.J., McIntosh, P.C., Wheeler, M.C., Hendon, H.H. 2009. On the Remote Drivers of Rainfall Variability in Australia. Mon. Weath-er Rev., 137, 3233–53. doi: http://dx.doi.org/10.1175/2009MWR2861.1

Risbey, J.S., McIntosh, P.C., Pook, M.J., Rashid, H.A., Hirst, A.C. 2011. Evaluation of rainfall drivers and teleconnections in an ACCESS AMIP run. Aust. Met. Oceanogr. J., 61, 91–5.

Roberts, M.J., A. Clayton, M.-E. Demory, J. Donners, P.L. Vidale, W. Nor-ton, L. Shaffrey, D.P. Stevens, I. Stevens, R.A. Wood, and J. Slingo. 2009. Impact of Resolution on the Tropical Pacific Circulation in a Matrix of Coupled Models. J. Clim., 22, 2541–56. doi: http://dx.doi.org/10.1175/2008JCLI2537.1

Shukla, J., DelSole, T., Fennessy, M., Kinter, J. and Paolino, D. 2006. Cli-mate model fidelity and projections of climate change. Geophys. Res. Lett., 33, L07702, doi:10.1029/2005GL025579.

Smith, R.N.B. 1990. A scheme for predicting layer clouds and their wa-ter content in a general circulation model. Q. J. R. Meterol. Soc., 116, 435–60.

Smith, T.M., and Reynolds, R.W. 2003. Extended Reconstruction of Glob-al Sea Surface Temperatures Based on COADS Data (1854–1997). J. Clim., 16, 1495–510.

Suarez, M.J., and Schopf, P.S. 1988. A Delayed Action Oscillator for ENSO. J. Atmos. Sci., 45, 3283–7. doi: http://dx.doi.org/10.1175/1520-0469(1988)045<3283:ADAOFE>2.0.CO;2

Taylor, K.E., Stouffer, R.J. and Meehl G.A. 2012. An overview of CMIP5 and the experiment design. Bull. Am. Meteorol. Soc., 93, 485–98, doi:10.1175/BAMS-D-11-00094.1

Torrence, C., and Compo, G.P. 1998. A Practical Guide to Wavelet Analy-sis. Bull. Am. Meteor. Soc., 79, 61–78.

evaluation of el niño–southern oscillation in the access ... · evaluation of el...

Documents