Assessment of forced responses of the Australian · PDF filemodelling groups have contributed to the fifth phase ... 2013. As such, ‘tier 1’ and ‘tier 2’ CMIP5 ......

Australian Meteorological and Oceanographic Journal 64 (2014) 87–101

87

Assessment of forced responses of the Australian Community Climate and Earth System Simulator (ACCESS) 1.3 in CMIP5 historical detection

and attribution experimentsSophie C. Lewis and David J. Karoly

School of Earth Sciences and ARC Centre of Excellence for Climate System Science, University of Melbourne, Melbourne, Victoria, 3010, Australia

(Manuscript received January 2014; revised May 2014)

The Australian Community Climate and Earth System Simulator (ACCESS) cou-pled climate model version 1.3 participated in phase five of the Coupled Model In-tercomparison Project (CMIP5) with an initial contribution of high priority experi-ments. Further standard experiments have since been conducted with ACCESS1.3, including an ensemble of three simulations for the historical period (1850–2005) forced with time-evolving natural and anthropogenic forcings. Additional en-sembles of simulations have been conducted for the same period with subsets of known forcings, including with natural forcings only (‘historicalNat’) and with greenhouse gas forcings only (‘historicalGHG’). In this study, we describe this ACCESS1.3 contribution to CMIP5 and assess several key aspects of ACCESS1.3 forced responses in these experiments against observations and an ensemble of participating CMIP5 models, consisting of 40 realisations from ten models. Over-all, ACCESS1.3 historical experiments demonstrate skill in simulating the global and regional metrics assessed that is comparable to the CMIP5 multi-model en-semble utilised. Global annual average temperature and precipitation trends sim-ulated with ACCESS1.3 (0.05–0.07 K/decade; −0.007 to −0.0004 mm day-1/decade) largely lie within the CMIP5 ensemble window (0.06 − 0.18 K/decade; −0.01 to 0.009 mm day-1/decade) and near those observed (0.10 K/decade; −0.0007 to −0.001 mm day-1/decade) over the 1950–2005 period. For the ACCESS1.3 historicalNat and historicalGHG experiments, simulated temperature trends are also predominately within the CMIP5 multi-model ensemble range. Similarly, ACCESS1.3 (−0.07 to −0.12 K) and the CMIP5 models (−0.03 to −0.21 K) largely capture the composited observed decrease in global temperature (−0.04 K) following three major late 20th century volcanic eruptions. However, like all global climate models, ACCESS1.3 has deficiencies that should be considered. In particular, one of most notable fea-tures of ACCESS1.3 historical simulations is the reduced warming trend over the period 1950–2005 that is evident in all ACCESS1.3 realisations at the global-scale for Australia, relative to both observations and the CMIP5 multi-model mean. This appears to be related to the overly strong response to increases in anthropogenic aerosols. Overall, these historical period experiments using ACCESS1.3 with vari-ous forcings are useful for inclusion with other CMIP5 models for studies aimed at detecting and attributing climatic changes.

Corresponding author address: Sophie C. Lewis, [email protected] Tel: +61 2 6250 0920

88 Australian Meteorological and Oceanographic Journal 64:2 June 2014

for attributing observed extreme climate events to particular causes (Lewis and Karoly, 2013). However, the utility of the CMIP5 simulations of the historical experiment suite for understanding observed climatic changes, and more broadly for understanding climate processes, is dependent on the participating models’ ability to capture observed changes for physically correct reasons. In particular, we can ascribe greater confidence in coupled model projections of future climate change under various scenarios where models are found to realistically represent relevant climate process, including responding to external forcings during the historical period. As such, assessing the ACCESS1.3 detection and attribution experiments against observations and against the ensemble of participating CMIP5 models is helpful for future studies that may incorporate output from these ACCESS1.3 simulations.

This paper extends the preliminary descriptions of ACCESS1.3 and its contribution to CMIP5 (Bi et al. 2013; Dix et al. 2013). The ACCESS Community Climate and Earth System Simulator is first described, together with the CMIP5 historical experiment suite configuration and details of forcings (solar, volcanic aerosol, anthropogenic aerosols, ozone and long-lived greenhouse gases) incorporated. We then evaluate key aspects of model performance, including modelled temperature and precipitation responses to imposed anthropogenic greenhouse gas forcings and modelled responses to volcanic stratospheric aerosols. Finally, the model skill and shortcomings, relative to both observations and a broader CMIP5 ensemble, are summarised. We do not present comparisons of ACCESS1.3 with observations and CMIP5 models as an exhaustive evaluation of model performance, but rather as an introductory assessment of model skill in several key metrics.

Model description and experimental design

Model descriptionACCESS is a coupled climate and earth system simulator developed as a joint initiative of the Australian Bureau of Meteorology (BoM) and the Commonwealth Scientific and Industrial Research Organisation (CSIRO), together with various Australian universities. The ACCESS climate model has been described extensively in a special issue of the Australian Meteorological and Oceanographic Journal (AMOJ) published in 2013, and particularly by Bi et al. (2013). Comprehensive details are provided there of physical parameterisations of the atmosphere, ocean, sea ice, land surface processes, aerosols and coupling strategy.

The ACCESS1.3 model is a coupling of the UK Met Office atmospheric Unified Model (UM 7.3) with the ACCESS-OM ocean-sea ice component. ACCESS-OM comprises an ocean (GFDL MOM4p1) (Griffies et al. 2011) and sea ice (LANL CICE 4.1) (Hunke and Lipscomb 2008) component with the OASIS 3.25 coupler (Valcke et al. 2013). The atmospheric model horizontal resolution is 1.875° longitude by 1.25° latitude (referred to as ‘N96’), with 38 vertical levels, while ACCESS-

Introduction

Coupled climate models are used for various purposes, including for providing projections of future climate change under anthropogenic greenhouse gas (GHG) warming. The utility of projections of future climate change does, however, depend on the accuracy of the coupled models in simulating forced responses to radiative perturbations, such as anthropogenic greenhouse gases. Various international modelling groups have contributed to the fifth phase of the Coupled Model Intercomparison Project (CMIP5) with standard model experiments (Taylor et al. 2012) that allow forced model responses to be systematically evaluated against each other and against observational climate records. This paper documents a contribution to CMIP5 using the Australian Community Climate and Earth System Simulator (ACCESS) coupled climate model, developed at the Centre for Australian Weather and Climate Research (CAWCR)1. An initial contribution was made to CMIP5 using two versions of the ACCESS coupled model (ACCESS-CM), namely ACCESS1.3 and ACCESS1.0, with resulting fields now available for use on the Earth System Grid. This initial participation in CMIP5 with two versions of ACCESS-CM focused on performing high priority, core experiments in support of Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) (Dix et al. 2013).

Nonetheless, CMIP5 is a long-term modelling endeavour with utility beyond having contributed to the IPCC AR5 in 2013. As such, ‘tier 1’ and ‘tier 2’ CMIP5 experiments, in addition to high priority, core experiments, are also useful for the evaluation of climate models and improving our understanding of climate processes. In particular, standard experiments of the historical period (‘historical’, 1850–2005), forced by observed time-evolving natural and anthropogenic forcings, are used for evaluating model performance for various metrics, relative to observations. Also, the CMIP5 experiment design describes additional twentieth century climate simulations that are predominantly used for detection and attribution of climatic change. In the detection and attribution experiment suite, participating models simulate the climate of the twentieth century with subsets of known forcings. Commonly, simulations are conducted for the period of 1850–2005 with natural forcings only (‘historicalNat’) and with greenhouse gas forcings only (‘historicalGHG’). These simulations allow the response of models to individual forcings to be determined. Previous studies utilising output from these various historical simulations have detected and attributed various observed changes in the climate system over this period, including, for example, observed changes in global mean temperatures (Jones et al. 2013) and northern hemisphere snow cover (Rupp et al. 2013). These experiments have also been useful

1 CAWCR is a partnership between CSIRO and the Bureau of Meteorol-ogy.

Lewis and Karoly: Assessment of forced responses of the Australian Community Climate and Earth System Simulator (ACCESS) 89

OM ocean and sea ice component share a regular horizontal grid, modified to the north of 65°N to implement a tripolar grid. This is defined by a uniform 1° longitude resolution and an equatorial meridional refinement of ⅓° for 10°S–10°N and a Mercator grid in the Southern Ocean of 1° at 30°S through to ¼° at 78°S. The ocean is 50 levels deep and covers ~6000 m, with 20 levels of 10 m thickness comprising the upper ocean and below 200 m, the vertical resolution decreases smoothly to ~333 m in the abyssal ocean. ACCESS1.3 version includes the Community Atmosphere Biosphere Land Exchange (CABLE1.0) land surface model (Kowalczyk et al. 2006). This land scheme includes six soil levels and a number of sub-models for simulating canopy processes, soil and snow, carbon pool dynamics and soil respiration. In addition, ACCESS1.3 also includes a significant upgrade in model physics compared to the ACCESS1.0 model version, including the PC2 prognostic cloud scheme for prognostic cloud fraction and condensate (Hewitt et al. 2011), where vapour and liquid water and cloud fractions become prognostics in the model.

Experimental designThe historical detection and attribution experiments described in this study follow the protocols described by Taylor et al. (2012). The CMIP5 protocols recommend that modelling groups should perform an ensemble of simulations for various experiments, rather than a single climate simulation (Taylor et al. 2012). In this case, all ensemble members are conducted with identical time-evolving forcings (Fig. 1) and experimental conditions, but differ in their initialisation. An ensemble of simulations allows the externally-forced response to be more readily identified and assessment to be made of the significance of differences between modelled and observed fields. In collaboration with CAWCR, three simulations were conducted for each of the historical, historicalGHG and historicalNat experimental designs, with each successive realisation initialised from different time points in the ACCESS1.3 pre-industrial control simulation. These are designated by CMIP5 ensemble identifiers r1i1p1, r2i1p1 and r3i1p1. The experiments referred to here, including the forcings incorporated in each detection and attribution experiment and ensemble members used are summarised in Table 1. While the historicalNat experimental design includes only natural forcings (solar and volcanic aerosols), the primary distinction between the historical and historicalGHG experiments is the inclusion of anthropogenic aerosols.

Generally, the climate forcings stipulated by CMIP5 are implemented in ACCESS1.3 in an identical manner to those document for the HadGEM2-ES model (Jones et al. 2011). For the historical experiment, standard CMIP5 prescriptions are used for atmospheric concentrations of CO2, CH4, N2O, O3 and halocarbons. In addition, time-varying total solar irradiance is applied from Lean (2009) and stratospheric volcanic aerosols are implemented from the Sato et al. (1993) monthly mean optical depth data source, extended

Fig. 1. Evolution of natural and anthropogenic changes used as forcings for the CMIP5 historical experiment suite for 1910–2005. (a) Well-mixed anthropogenic green-house gas concentrations in parts per million (CO2 equivalent ppm) (see Meinshausen et al. (2011) for further details). (b) Annual mean emissions of sul-phur dioxide. Full details of tropospheric aerosol pre-cursors and primary emissions can be found in Jones et al. (Jones et al., 2011) (c) Total solar irradiance (TSI) from Lean (2009). (d) Observed monthly mean aero-sol optical depth data from Sato et al. (1993). The largest tropical volcanic eruptions (Agung in 1963, El Chichon in 1982 and Pinatubo in 1991) since 1910 are indicated.

1920 1940 1960 1980 2000Year

0.04

0.08

0.12

0.16

0.2

AOD

at 5

50 n

m

Agung

El Chichon

Pinatubo

360

400

440

GH

G c

once

ntra

tion

(CO

2 eq

ppm

)

1365.2

1365.6

1366

1366.4

1366.8

TSI (

W/m

2 )

20

30

40

50

60

Tg[S

]/yea

r

A. GHG concentrations

B. Sulphur dioxideemissions

C. Total solar irradiance

D. Volcanic aerosoloptical depth

320


to 2000 and relaxed to background values by 2040. In the ACCESS model, following the HadGEM2-ES approach (Jones et al. 2011), stratospheric aerosols concentrations are varied monthly across four latitudinal bands. The aerosols are distributed vertically above the troposphere and do not interact with other simulated aerosols. The pre-industrial stratospheric volcanic aerosols composition is taken as zero, which is noted to result in an overall slight cold bias in the historical, relative to the control simulation (Dix et al. 2013; Gregory et al. 2013). A seasonally-varying biogenic aerosol concentration is maintained throughout the historical simulations, along with a background SO2 out-gassing flux. Land use changes have not been included.

Observational and other CMIP5 datasetsModel output from ACCESS1.3 was compared to data compiled from both gridded observational sources and from other models participating in CMIP5. First, we use the HadCRUT4 (v4.2.0.0) gridded dataset of monthly surface air temperatures (Morice et al. 2012). It should be noted that various gridded observational temperature datasets are available for the historical period, although these various observational products demonstrate only small temperature differences (Brohan et al. 2006; Xinyu et al. 2011) and hence we evaluate ACCESS1.3 against HadCRUT4 data only. Model temperature fields were masked to exclude areas of poor HadCRUT4 data coverage, using a surface data mask that included only grid boxes with at least 75 per cent temporal coverage from 1910 to 2005. As observed precipitation products do not have the same degree of data quality and coherence as temperature products, we use observational precipitation data from two independent sources, namely the CRU TS 3.1 (Harris et al. 2013) and the GPCC v6 (Schneider et al. 2013) monthly gridded land-only precipitation datasets. As data quality diminishes in the earlier parts of the observational record, we consider changes only over the period of 1910–2005, which represents the common extent of CMIP5 historical simulations and high-quality observational records in Australia (Trewin, 2012). Next, ACCESS1.3 simulated fields were compared to output from other participating CMIP5 models. Models were selected for inclusion in this comparison where mean near-surface air temperature (tas) and precipitation (pr) fields were available on the Australian node of the Earth System Grid for all historical detection and attribution experiments (historical, historicalGHG, historicalNat) assessed here, for

at least three unique realisations. A multi-model ensemble mean was determined as average conditions across all realisations, using all analysed models, and 5th and 95th percentile ranges were also calculated to indicate the range of simulated values across the realisations. Models and realisations included in this analysis are summarised in Table 2.

Modelled and observed near-surface air temperature (Tmean) and precipitation (PR) variables were processed in a common manner, with anomalies determined relative to a 1910–1940 climatology, and seasonal and annual averages computed. The temporal averages we consider here are annual (January to December), austral summer (December to February, DJF) and austral winter (June to August, JJA). In addition, we calculate global, together with Australian, land surface and ocean area average values. For each timeseries, we also calculate the linear temperature trends over the shorter period of 1950–2005, which encompasses the period of greatest observed warming. The statistical significance of trends is assessed using a t-test at the five per cent level and only statistically significant trends are reported here, unless otherwise indicated. The determination of precipitation trends is sensitive to the trend detection technique, which may produce varying results (Kumar et al. 2013), and here we estimate trends using Sen’s Kendall slope non-parametric method (Sen 1968).

We also evaluate the observed and modelled temperature

Table 1. Summary of historical, historicalGHG and historicalNat experiments and relevant imposed forcings. The greenhouse gases include CO2, N2O, CH4, CFC11, CFC12, CFC113, HCFC22, HFC125, HFC134a.

Experiment name

Well-mixed greenhouse

gases

Tropospheric and

stratospheric ozone

Anthropogenic sulfate aerosol

direct and indirect effects

Black carbon Organic Carbon

Solar irradiance

Stratospheric Volcanic aerosols

historical Varying Varying Varying Varying Varying Varying Varying

historicalGHG Varying 1850 cycled 1850 cycled 1850 cycled 1850 cycled Constant None

historicalNat Constant 1850 cycled 1850 cycled 1850 cycled 1850 cycled Varying Varying

Table 2. Summary of CMIP5 models and realisations used in this study. Further details of participat-ing models can be found with the Program for Climate Model Diagnosis and Intercomparison (PCMDI; cmip-pcmdi.llnl.gov).

Model Realisations

ACCESS1.3 r1i1p1, r2i1p1, r3i1p1

CCSM4 r1i1p1, r4i1p1, r6i1p1

CNRM-CM5 r1i1p1, r2i1p1, r4i1ip1, r8i1p1

CSIRO-Mk3-6-0 r1i1p1, r2i1p1, r3i1p1, r4i1ip1, r5i1p1

CanESM2 r1i1p1, r2i1p1, r3i1p1, r4i1ip1, r5i1p1

GFDL-CM3 r1i1p1, r3i1p1, r5i1p1

GISS-E2-H r1i1p1, r2i1p1, r3i1p1, r4i1ip1, r5i1p1

GISS-E2-R r1i1p1, r2i1p1, r3i1p1, r4i1ip1, r5i1p1

HadGEM2-ES r1i1p1, r2i1p1, r3i1p1, r4i1ip1

IPSL-CM5A-LR r1i1p1, r2i1p1, r3i1p1

MIROC-ESM r1i1p1, r2i1p1, r3i1p1


and precipitation responses to the three largest historic volcanic eruptions occurring between 1910 and 2005 (Table 3). Specifically, these eruptions are Agung, which erupted in 1963, El Chichón in 1982 and Pinatubo in 1991 (Fig. 1(d)). Using the Sato et al. (1993) monthly mean optical depth data used as a forcing for CMIP5 simulations, we determine mean temperature and precipitation perturbations in each model realisation in the years immediately following the identified eruptions, relative to a baseline climatology of six years prior to eruption. This baseline reference period is approximately the same as that utilised by Stenchikov (2006) and provides the longest period prior to each eruption in which the atmosphere can be assumed to be clear of volcanic aerosols from previous eruptions, although the calculated climatological anomalies associated with volcanic eruptions are generally insensitive to the choice of reference period. We then investigate average temperature and precipitation changes in the two years following each eruption. Finally, we calculate a composite modelled volcanic response, as a weighted mean of the three eruptions, scaled by the cumulative aerosol optical depth in the calendar year each eruption occurred. Using this composited volcanic response, we compare simulated areal average temperature and precipitation changes with those observed.

Assessment of response to anthropogenic forcings

Temperature responsesComparison of the observed and simulated global annual average temperatures over the period of analysis (1910–2005) shows that the observed warming trend of recent decades is best captured by ACCESS1.3 simulations including both anthropogenic and natural forcings (Fig. 2 after Bindoff et al. 2013). The historicalNat simulation, including only natural forcings, underestimates the observed warming trend, while simulations that exclude anthropogenic aerosols (historicalGHG) conversely demonstrate substantially greater warming over the 20th century than that observed. For the historicalGHG experiment, observed temperatures fall clearly outside the spread of CMIP5 ensemble members. Similarly, for the natural forcings only experiments, the observed global mean temperatures lie well above the modelled ensemble ranges in the years following ~1960.

Although the historical simulation best replicates the observed 20th century temperature change, these

Fig. 2. Timeseries of annual global Tmean (K) anomalies for historical (a), historicalGHG (b), historicalNat (c) ex-periments and observations. In each panel, observa-tions are shown by the dashed black line, the CMIP5 ensemble mean in grey and grey plumes indicate the CMIP5 5th and 95th percentiles. The three ACCESS1.3 realisations are indicated by blue for the historical ex-periment, red for the historicalGHG experiment and green for the historicalNat experiment. Note that each experiment is displayed on a different scale. Model and observational data are masked to exclude areas of poor observational data coverage, defined by Had-CRUT4 data coverage for grid boxes with at least 75 per cent temporal from 1910 to 2005.

-0.5

0

0.5

1

0

0.5

1

-0.5

0

0.5

1920 1940 1960 1980 2000

Year

Tmea

n an

omal

y (K

) Tm

ean

anom

aly

(K)

Tmea

n an

omal

y (K

)

A.

ObservationsCMIP5 meanACCESS1.3 historical

Global Tmean

A.


ObservationsCMIP5 meanACCESS1.3 historicalGHG

B.

C.

ObservationsCMIP5 meanACCESS1.3 historicalNat

Table 3. Summary of major tropical volcanic eruptions over the period of 1910–2005. The scaling factor represents the cumulative aerosol optical depth (at 550 nm) in the calendar year each eruption occurred and is used to determine a weighted mean composite volcanic response.

Volcanic eruption Eruption date Latitude Years analysed Reference PeriodVolcanic AOD scaling

factor at 550 nm

Pinatubo 15 June, 1991 15.13°N 1992–1993 1985–1990 1.537

El Chichón 4 April, 1982 17.36°N 1983–1984 1976–1981 0.9029

Agung 17 March, 1963 8.34°N 1964–1965 1957–1962 0.8603


ACCESS1.3 simulations including both anthropogenic and natural forcings display somewhat subdued warming since the middle of the century. Compared to the CMIP5 multi-model ensemble (defined in Table 2), the ACCESS1.3 historical period global annual temperature variations simulated using the historicalGHG and historicalNat experimental designs lie within the ensemble window and close to the multi-model ensemble mean (Figs 2(b) and 2(c)). However, the ACCESS1.3 realisations lie near the cooler edge of the multi-model envelope for the historical experiment. When we consider the linear temperature trends over the shorter period of 1950–2005, the somewhat modest warming trends for ACCESS1.3 relative to observed are evident (Fig. 3(a)). The CMIP5 multi-model trends range from 0.06−0.18 K/decade and the observed trend is 0.10 K/decade, which is notably warmer than the spread of trends simulated using ACCESS1.3 (0.05–0.07 K/decade). Although the magnitudes of the calculated temperature trends are somewhat sensitive to the time periods considered, we find that the relative performance of ACCESS1.3 is not. For example, when we

compared ACCESS1.3 global temperature trends for 1965–2005 to observations and the CMIP5 multi-model ensemble, simulated ACCESS1.3 temperature trends were again notably lower than the CMIP5 mean.

A relatively stronger cooling in the middle of the century relative to observed (from a 1961–1990 climatology) has been identified in the HadGEM2-ES model contribution to CMIP5 (Stott and Jones 2012). This was attributed to a possible combination of natural internal climatic variability and incorrect specification of model forcings. Specifically, the effect of modelled land use changes was identified as potentially contributing to significantly to global mean cooling over this period. As the HadGEM2 model family incorporates a similar atmospheric component (Martin et al. 2011) to ACCESS1.3, corresponding discrepancies in forced responses relative to observed may be related. However, ACCESS1.3 incorporates a static land use scheme. Also, given the ACCESS model performance relative to the CMIP5 ensemble for the historicalGHG and historicalNat simulations, it is possible that subdued historical temperature

Fig. 3. Annual Tmean trends (K/decade) for global (a) and Australian (b) area averages over the period of 1950–2005. Coloured crosses indicate the simulated trends in each of the three realisations using ACCESS1.3 for each experiment. Plot circles represent CMIP5 ensemble mean trends and ranges indicate the CMIP5 5th and 95th percentile trends. The observed tem-perature trend is indicated by the black square. Simulated historical experiment values are shown in blue, historicalGHG in red and historicalNat in green. Annual mean trends are also shown for (c) land only and (d) Australian area-average precipitation (PR, mm day-1/decade), as calculated by the non-parametric Sen’s slope estimator. Both the GPCC and CRU TS observational precipitation trends are indicated by black squares. The historicalNat temperature and precipitations trends are not statistically significant at the five per cent level, and Australian precipitation trends are only significant for the his-toricalGHG simulations.

-0.2

-0.1

0

0.1

0.2

0.3

0.4

Tmea

n tre

nd (K

/dec

ade)

A. Global C. Land

-0.04PR tr

end

(mm

day

-1 /d

ecad

e) D.

Australia

-0.02

0

0.04

PR tr

end

(mm

day

-1 /d

ecad

e)

Historical HistoricalGHG HistoricalNat

B. Australia

-0.2

-0.1

0

0.1

0.2

0.3

Tmea

n tre

nd (K

/dec

ade)

-0.02

-0.01

0

0.01

0.02


Historical

HistoricalGHG HistoricalNat

0.02

Observations CMIP5 mean }ACCESS1.3 Observations CMIP5 mean }ACCESS1.3

Tmean trends Precipitation trends



responses relate primarily to the responses to increasing anthropogenic aerosols, which constitute the foremost difference between the historical and historicalGHG. Previously, differences in historical model warming rates between ACCESS1.3 and another model version with a different land surface scheme (ACCESS1.0) could not be attributed to either natural variability or model sensitivity as only one realisation was available (Dix et al. 2013). It should be noted that previous studies have investigated, and discounted, potential pre-industrial control drifts as a source of bias in simulating historical temperature trends (Bi et al. 2013). In our present study, this subdued warming is ubiquitous across all historical all forcings realisations and hence may indicate that this bias results from a forced model response.

At the regional scale, we also investigate simulated historical temperatures for Australia using ACCESS1.3

(Fig. 4). Generally, the performance of ACCESS1.3 relative to observed and to the CMIP5 ensemble is similar at the global and Australia regional scale. Again, the observed Australian warming trend of recent decades is best captured by simulations including both anthropogenic and natural forcings and there is also a subdued warming in Australian late 20th century temperatures in ACCESS1.3 relative to observed and to the CMIP5 multi-model mean. The discrepancy between historical and observed temperature trends over 1950–2005 is larger for Australia than globally

Fig. 4. As for Fig. 2, but showing Australian areal mean an-nual temperature anomalies (Tmean, K).

-0.5

0

1

2

1920 1940 1960 1980 2000Year

Tmea

n an

omal

y (K

)

A.

B.

C.




Australian Tmean

1.5

0.5

-0.5

0

1

2

Tmea

n an

omal

y (K

)

1.5

0.5

-0.5

0

1

Tmea

n an

omal

y (K

) 1.5

0.5

-1

-1A.

Fig. 5. Timeseries of annual land-ocean Tmean (K) contrast anomalies for (a) historical, (b) historicalGHG, (c) historicalNat experiments and observations. In each panel, observations are shown by the dashed black line, the CMIP5 ensemble mean in grey and grey plumes indicate the CMIP5 5th and 95th percentiles. The three ACCESS1.3 realisations are indicated by blue for the historical experiment, red for the histori-calGHG experiment and green for the historicalNat experiment.

-0.5

0

0.5

1

0

0.5

-0.5

0

0.5

1920 1940 1960 1980 2000Year

Tmea

n an

omal

y (K

) Tm

ean

anom

aly

(K)

Tmea

n an

omal

y (K

)

A.

B.

C.




Land-Ocean Tmean contrast

B.

-0.5


(Fig. 3), although this is more difficult to discern in the temperature timeseries (Fig. 4) because of the larger magnitude of interannual variability occurring for Australia than globally. There is also a wider spread of forced responses in the ACCESS1.3 realisations over the smaller Australian region, than globally.

Next, we compare simulated and observed land-ocean temperature contrasts (LO) (Fig. 5). The contrast in temperatures over land and over the oceans provides a useful metric for detecting and attributing climate change, in addition to global mean surface temperatures (Drost and

Karoly 2012). Specifically, the difference in warming between the land surface and oceans increases under increasing globally uniform radiative forcing, providing a measure of the slower rate of oceanic, than terrestrial warming due to the comparatively larger heat capacity of water and increased evaporative cooling over oceans. Observed changes in land-ocean temperature contrasts indicate a decrease (relative to the 1910–1940 climatology) in the 1960s and 1970s, before a subsequent increase in the temperature contrast. This is likely due to increases in anthropogenic aerosol forcing at this time, which is not included in the historical GHG simulations. In the model experiments, an increase in LO is simulated only in the historical and historicalGHG experiments, which include anthropogenic greenhouse gases. The simulated LO increases in the historicalGHG experiments are larger than observed for both the ACCESS1.3 realisations and the CMIP5 ensemble. For the historical experiment, the ACCESS1.3 LO values are predominantly lower during recent decades than both those observed and than the CMIP5 multi-model mean, coinciding with the period of subdued global warming simulated by ACCESS1.3. The subdued global warming and low or negative LO values for ACESS1.3 historical runs suggest that the modelled response to anthropogenic aerosol forcing may be too large.

Precipitation responsesSimulated and observed precipitation over the period of analysis of 1910–2005 is temporally more variable than temperature. Previously, precipitation changes through time have been found to be more difficult to detect than forced temperature responses (Ren et al. 2013). When we consider only simulated land surface precipitation changes over this period, there is a decrease in precipitation in the historical experiment for the CMIP5 multi-model ensemble mean (Fig. 6). However, observed precipitation is generally more than for the models, with excursions occurring outside the model range. The standard deviation of observed annual precipitation (PR σ 0.05–0.06 mm day-1) is somewhat higher than simulated by the CMIP5 ensemble (PR σ 0.03 to 0.06 mm day-1, mean 0.04 mm day-1) and ACCES1.3 (PR σ 0.045 to 0.05 mm day-1). There are also large differences in simulated precipitation changes between models and a large spread of simulated trends occurring across the CMIP5 ensemble, with both decreasing and increasing late twentieth century trends simulated by different models (Kumar et al. 2013). Precipitation trends in the CMIP5 models over 1950–2005 are centered around zero, with a multi-model mean trend of −0.001 mm day-1/decade and an ensemble spread of −0.01 to 0.009 mm day-1/decade (Fig. 3(c)). In all ACCESS1.3 historical realisations, there is a decrease in global land surface precipitation in all three ensemble members (−0.007 to −0.0004 mm day-1/decade). Alternatively, for the historicalGHG experiment, there are larger increasing trends over this period for ACCESS1.3 (0.001 to 0.01 mm day-1 /decade) and for the CMIP5 ensemble (–0.002 to 0.012 mm day/decade,

Fig. 6. Timeseries of annual land-only precipitation (PR) anomalies (mm day-1) for (a) historical, (b) historical-GHG, (c) historicalNat experiments and observations. In each panel, the CMIP5 ensemble mean in grey and grey plumes indicate the CMIP5 5th and 95th per-centiles. Both the GPCC and CRU TS observational datasets are shown by dashed black lines. The three ACCESS1.3 realisations are indicated by blue for the historical experiment, red for the historicalGHG ex-periment and green for the historicalNat experiment.

0

0.1

0.2

1920 1940 1960 1980 2000Year

PR a

nom

aly

(mm

day

-1)


Land-only precipitation

-0.1

0

0.1

-0.1

0

0.1

-0.1


A.


B.

C.

PR a

nom

aly

(mm

day

-1)

PR a

nom

aly

(mm

day

-1)


mean 0.007 mm day-1 /decade). The simulated historicalNat ACCESS1.3 precipitation trends are again clustered around zero and are not statistically significant.

The historical precipitation decreases simulated with ACCESS1.3, when considered together with the historicalGHG trends, are consistent with strong aerosol forcings over the twentieth century (Dix et al. 2013) that may also be evident in the subdued global temperature changes for ACCESS1.3 relative to observed (Fig. 2). In terms of comparing simulated precipitation trends with observed trends, the observed global land surface mean annual precipitation anomalies also demonstrate decreases over the periods of both 1950–2005 (GPCC −0.0007 mm day-1 decade; CRU TS −0.001 mm day-1/decade). As such, there are small differences in precipitation between the two observational datasets. Furthermore, global precipitation changes detected in previous studies are largely dominated by changes occurring over the oceans, rather than the land surface, where more temporally variable precipitation occurs (Dix et al. 2013; Ren et al. 2013). This is exacerbated

at regional and local scales, where large uncertainties in simulated precipitation trends occur. When we consider Australian areal average precipitation, different changes in Australian average annual precipitation are simulated for the various ACCESS1.3 historical simulations. In particular, both decreasing (−0.004 mm day-1/decade) and increasing (0.05 mm day-1/decade) Australian precipitation trends are simulated for the 1950–2005 period using ACCESS1.3 (Fig. (3d)). However, these precipitation trends simulated with ACCESS1.3 are not found to be statistically significant: while the trends simulated over Australia are generally larger than for the global land surface, the intrinsic variability is substantially higher at the regional scale.

Assessment of response to volcanic aerosols

Temperature responses Next, we assess the response of ACCESS1.3 in the two years immediately following the large tropical volcanic eruptions that occurred during the twentieth century. We

Fig. 7. Global annual Tmean anomalies (K) in response to major 20th century volcanic eruptions for the (a) composite response and (b) Pinatubo, (c) El Chichón and (d) Agung. Simulated historical experiment values are shown in blue and historicalNat in green. Coloured crosses indicate the simulated anomalies in each of the three realisations using ACCESS1.3 for each experiment, although in cases where two realisations produce similar anomalies, only one cross is evident for these data. Plot circles represent CMIP5 model ensemble mean anomaly and the CMIP5 5th and 95th percentile ranges are indicated. The observed temperature anomaly is indicated by the black square. The reference years and years analysed for each vol-cano are defined in Table 3.

-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

Historical HistoricalNat

Tmea

n an

omal

y (K

))

A. Composite B. Pinatubo

Observations CMIP5 mean }ACCESS1.3


-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1


Tmea

n an

omal

y (K

)

Historical HistoricalNatC. El Chichón D. Agung



consider both temperature responses in the historical and historicalNat experiments, which both simulate volcanic eruptions through time-varying aerosol optical depth forcings. First, the observed decrease in global annual average temperatures (ΔTmean −0.04 K) in response to the composite 20th century volcanic eruption (incorporating Agung, El Chichón and Pinatubo) is captured by ACCESS1.3 in the historical experiment (ΔTmean −0.07 to −0.12 K). However, the simulated global temperature decreases are larger than observed for both ACCESS1.3 and the CMIP5 ensemble (ΔTmean −0.03 to −0.21 K, mean −0.13 K) utilised here (Fig. 7). Previous analysis of CMIP5 and CMIP3 models

have also indicated that simulated post-volcanic temperature responses are not in full agreement with observations and multi-model responses tend to be larger than in the observations (Knutson et al. 2013).

When each major volcanic eruption is considered separately, the largest overestimate of temperature anomalies simulated by ACCESS1.3, relative to that observed, occurs for Pinatubo, which is the largest eruption in terms of aerosol loading. Both ACCESS1.3 (ΔTmean −0.21 to −0.05 K) and the CMIP5 ensemble multi-model mean (ΔTmean −0.16 K) produce comparatively larger coolings following Pinatubo than the observed decrease in global annual average temperatures (ΔTmean −0.01 K). Previously, Knutson et al. (2013) identified larger than observed temperature responses to the Pinatubo eruption in CMIP5 models. Following the eruption of El Chichón, there are small observed temperature increases (ΔTmean 0.06 K), which lies within the range of values simulated by the three-member ACCESS1.3 ensemble (ΔTmean −0.15 to 0.14 K) as well as within the CMIP5 ensemble spread (ΔTmean −0.26 to 0.11 K) and near the multi-model mean (ΔTmean −0.08 K). In addition, the observed response to Agung (ΔTmean −0.19 K) is also close to that simulated by ACCESS1.3 (ΔTmean −0.18 to −0.07 K).

Similar global annual temperature anomalies of slightly larger magnitude are simulated in the historicalNat experiments, where the volcanic aerosol forcings implemented are identical to those incorporated in the historical experiments with all natural and anthropogenic forcings. The simulated temperature response to volcanic forcing in the historicalNat runs is somewhat larger than in the historical all-forcing runs, and likely partly due the greenhouse gas-forced global warming contributing a small warming over the time from the averaging period before the volcanic eruption to the two years after. Using temperature anomalies computed from the historicalGHG runs, we estimate the greenhouse gas-forced contribution to this warming to be approximately 0.03–0.07 K from the CMIP5 multi-model ensemble mean, although this does not include the effects of tropospheric aerosols. The largest greenhouse-forced warming is simulated for Pinatubo and El Chichón, compared to the earlier Agung eruption. For the composite response, the simulated post-volcanic year-1 and year-2 average responses are consistent, although the year-1 response is on average ~25 per cent larger for the CMIP5 ensemble mean.

The differences between observed and modelled global surface temperature responses to volcanic eruptions in the historical and historicalNat experiments, using ACCESS1.3 and the broader CMIP5 models, may result from model deficiencies in implementing realistic aerosol loadings. Typically, volcanic eruptions are crudely implemented in models by employing representations of aerosols changes in four latitudinal bands (Driscoll et al. 2012), which may be inadequate for capturing the complexity of observed volcanic responses (Marshall et al. 2009) The CMIP5 inter-

Fig. 8. Annual and seasonal (DJF and JJA) Tmean anoma-lies (K) in response to composite 20th century volca-nic eruptions (Pinatubo, El Chichón and Agung) for (a) global, (b) northern hemisphere and (c) southern hemisphere area averages. Simulated historical ex-periment values are shown in blue and historicalNat in green. Coloured crosses indicate the simulated anomalies in each of the three realisations using ACCESS1.3 for each experiment, although in cases where two realisations produce similar anomalies, only one cross is evident for these data. Plot circles represent CMIP5 model ensemble mean anomaly and the 5th and 95th percentile ranges are indicated. The observed average post-volcanic temperature anoma-ly is indicated by the black square.


-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

ANN DJF JJA

Tmea

n an

omal

y (K

))

-0.5

-0.4

-0.3

-0.2

-0.1

0

ANN DJF JJA

Tmea

n an

omal

y (K

))

-0.5

-0.4

-0.3

-0.2

-0.1

0

ANN DJF JJA

Tmea

n an

omal

y (K

)

A. Global

B. Northern Hemisphere

C. Southern Hemisphere


model spread may result from the model sensitivity to the aerosol forcing or differences in model skill in capturing atmospheric circulation patterns that are relevant for producing physically realistic post-volcanic climate impacts (Stenchikov et al. 2006). Furthermore, differences between modelled responses may result from differing schemes for imposing volcanic eruption that have varying degrees of sophistication. The post-volcanic surface temperature responses of ACCESS1.3 are generally in the warmer end of the range defined by the CMIP5 multi-model ensemble and as such, are closer to the observed post-volcanic temperature anomalies. There is a spread of simulated temperature responses across the ACCESS1.3 realisations that may be attributable to internal variability in large-scale circulation patterns.

The precise interaction between volcanic eruptions and El Niño episodes remains unknown (Driscoll et al. 2012). In the first instance, we did not isolate the post-volcanic temperature anomalies from the intrinsic model El Niño–Southern Oscillation (ENSO) variations that also impact global mean temperature. For example, the observed positive temperature anomaly following El Chichón likely

results from the 1982–1983 El Niño episode, and may contribute to the composite anomaly being smaller than in the observations than the models. Similarly, the 1991–1992 El Niño episode occurred after the Pinatubo eruption of 1991. We next investigate the influence of ENSO conditions on post-volcanic global mean temperature responses by considering ‘residual’ temperature anomalies. These residuals were calculated by removing the part of the temperature signal that is linearly associated with simulated NINO3.4 surface air temperature variations. When the global temperature anomaly following El Chichón was recalculated without the confounding influence of ENSO variations, a cooling response occurs in the observed record (ΔTmean −0.06 K). This residual observed anomaly is near the CMIP5 multi-model mean (ΔTmean −0.02 K), although small increases in temperature are simulated with ACCESS1.3 after ENSO influences have been removed (ΔTmean −0.01 to 0.03 K).

Next, we assess seasonal temperature responses in ACCESS1.3 following the 20th century volcanic eruptions. Driscoll et al. (2012) previously evaluated the response of CMIP5 models to volcanic eruptions, focusing on large-scale northern hemisphere impacts, where the largest observed

Fig. 9. Land-only annual precipitation anomalies (mm day-1) in response to major 20th century volcanic eruptions for the (a) com-posite response, (b) Pinatubo, (c) El Chichón and (d) Agung. Simulated historical experiment values are shown in blue and historicalNat in green. Coloured crosses indicate the simulated anomalies in each of the three realisations using ACCESS1.3 for each experiment. Plot circles represent CMIP5 model ensemble mean anomaly and the 5th and 95th percentile ranges are indicated. The observed precipitation anomalies for both GPCC and CRU TS datasests are shown by black squares.

Historical HistoricalNatA. Composite B. Pinatubo


C. El Chichón D. Agung


-0.15

-0.075

0

0.075

0.15

PR

ano

mal

y (m

m d

ay-1)


-0.15

-0.075

0

0.075

Historical HistoricalNat Historical HistoricalNat

PR

ano

mal

y (m

m d

ay-1)


anomalies occur. Here, modelled surface temperature anomalies in the northern hemisphere following major volcanic eruptions were overestimated compared to observed, with comparatively large post-volcanic coolings simulated by CMIP5 models. In particular, during the boreal winter, post-volcanic surface warming has been observed over the northern hemisphere, due to dynamical feedbacks resulting from tropospheric cooling and stratospheric tropical heating (Stenchikov et al. 2006). In our present study, we also find an overestimate of post-volcanic temperature decreases in the ACCESS1.3 historical realisations that is largest during the boreal winter (Fig. 8). Although the ACCESS1.3 composite post-volcanic temperature anomalies for the northern hemisphere lie close to the CMIP5 multi-model mean, they are of the incorrect sign to those observed. This discrepancy in temperature responses compared with observed anomalies is evident in many CMIP5 realisations sampled here, and likely reflects deficiencies in capturing post-eruption circulation changes during boreal winters that stem from the amplification of the polar vortex (Perlwitz and Graf 2001; Shindell et al. 2001). The global surface temperature changes in post-volcanic boreal winters are dominated by northern hemisphere changes, with minimal surface temperature changes observed in the southern hemisphere over December–January. The observed post-volcanic changes for the southern hemisphere lie near the warmer end of the CMIP5 range in both DJF and JJA seasons, and the ACCESS1.3 realisations also lie in the warmer end of the CMIP5 range.

Precipitation responsesWe also assess changes in global land surface precipitation

following these major volcanic eruptions (Fig. 9). There are observed decreases in precipitation following the eruptions of Pinatubo and Agung for both datasets. Conversely, a variable change in precipitation over the land surface is observed in the years following the eruption of El Chichón in the two observational precipitation datasets. The precipitation response may have also been affected by the El Niño episode observed in that period. Overall, the composited observed precipitation response to the eruptions is minimal. There are also small differences in the composited precipitation responses calculated in the two observational datasets utilised here (GPCC ΔPR −0.002 mm day-1; CRU TS ΔPR −0.008 mm day-1) and previous studies have questioned whether a statistically significant volcanic signal exists in precipitation on timescales of years following an eruption (Gu and Adler 2011; Ren et al. 2013).

In the CMIP5 models, post-volcanic drops in land-only precipitation are simulated following all three major 20th century eruptions, with a spread of responses modelled across the CMIP5 ensemble for the historical experiment. For the composited response, the CMIP5 range of simulated land surface precipitation anomalies spans from −0.07 to 0.03 mm day-1 (ΔPR mean −0.02 mm day-1). Similarly, there is a spread of composited post-volcanic precipitatio changes simulated with ACCESS1.3 (ΔPR −0.01 to −0.05 mm day-1), again indicating that forced precipitation responses are largely more variable than temperature responses. When we consider the standardised range of simulated anomalies for each variable for the CMIP5 ensemble, the precipitation anomaly range for the composite eruption is twice as large as that for temperature anomalies. Also, in contrast to the seasonally and regionally varying nature of observed post-volcanic temperature responses, the observed precipitation response to the composited 20th century volcanic eruption is similar in both the boreal winter and summer (Fig. 10).

The comparatively large precipitation decreases simulated by ACCESS1.3 and within the CMIP5 multi-model ensemble relative to observed may result, at least in part, from model volcanic aerosols remaining in the atmosphere longer than is physically realistic and hence their impact on precipitation is too persistent as a result (Ren et al. 2013). The variations in precipitation responses between the CMIP5 model may also relate to the degree of persistence of aerosols in the atmosphere, together with inter-model differences in the impact of aerosols on model cloud microphysics. While global mean temperatures respond to observed volcanic eruptions on seasonal to yearly timescales, relatively fast precipitation responses occur (Soden 2002). Observed rapid volcanic precipitation anomalies are associated with direct radiative effects from reductions in solar radiation leading to reduced evaporative rates (Gillett 2004; Iles et al. 2013) and also with indirect effects through volcanic aerosol influences on cloud microphysics (Gu and Adler 2011). We adopt a simple approach to assessing the simulated ACCESS1.3 volcanic responses by examining climatological responses in the two full years following each eruption (Table 3), and

Fig. 10. Seasonal land-only seasonal (DJF and JJA) precipita-tion anomalies (mm day-1) in response to composite 20th century volcanic eruptions (Pinatubo, El Chichón and Agung). Simulated historical experiment val-ues are shown in blue and historicalNat in green. Coloured crosses indicate the simulated anomalies in each of the three realisations using ACCESS1.3 for each experiment. Plot circles represent CMIP5 model ensemble mean anomaly and the 5th and 95th per-centile ranges are indicated. The observed precipita-tion anomalies for both GPCC and CRU TS datasests are shown by black squares.


Land-only precipitation

-0.15

-0.1

-0.05

0

0.05

0.1

0.15ANN DJF JJA

PR

ano

mal

y (m

m d

ay-1)


hence it is possible some part of the precipitation responses may be missed by this lag. Alternatively, the spread of model results demonstrated, which are generally centered on the observed anomalies, may simply reflect natural internal variability of precipitation.

Summary and conclusions

This study has described the participation in the CMIP5 detection and attribution experiments with the Australian Community Climate and Earth System Simulator version 1.3 coupled climate model, and updated the initial descriptions provided by Bi et al. (2013) and Dix et al. (2013). We have described the contribution of ACCESS1.3 to CMIP5 with the standard experiments of the historical period, forced by various atmospheric composition changes. In addition to the standard simulation of the historical period from 1850–2005 forced by all anthropogenic and natural factors, we conduct additional standard experiments with natural forcings only and with greenhouse gas forcings only, which are collectively useful for detection and attribution studies. The historicalNat experimental design includes solar and volcanic forcings, while the primary distinction between the historical and historicalGHG experiments is the inclusion of anthropogenic aerosols (Table 1). For all these experiments of the 1850–2005 period, we undertook an ensemble of simulations, comprised of three realisations initialised from different time points in the ACCESS1.3 pre-industrial control simulation. We compare temperature and precipitation responses to time varying greenhouse gas forcings and volcanic eruptions (Table 3) simulated in ACCESS1.3, with those observed and with a wider CMIP5 ensemble comprised of 40 realisations from ten participating models (Table 2).

Overall, the ACCESS1.3 model experiments incorporating both natural and anthropogenic forcings best capture the observed global annual average temperature timeseries of the period of 1910–2005. When anthropogenic forcings, such as long-lived greenhouse gases, are excluded in the experimental design, simulated global annual average temperatures are significantly lower than those observed. Similarly, when greenhouse gas forcings are included, but other natural and anthropogenic influences are excluded, the simulated warming exceeds that observed. The overestimate of the observed warming in the historicalGHG experiments can be attributed to the exclusion of anthropogenic aerosols that counteract a substantial amount of greenhouse gas warming (Jones et al. 2013). The historical all forcings simulations conducted using ACCESS1.3 display notably subdued warming in the second half of the 20th century, relative to both observed temperature changes and the CMIP5 multi-model ensemble. The range of linear temperature trends (over 1950–2005) from the CMIP5 ensemble (0.06−0.18 K/decade) is generally warmer than the equivalent spread of trends simulated by ACCESS1.3 (0.05−0.07 K/decade). This subdued warming trend is evident in all individual

ACCESS1.3 historical realisations at the global and regional scale for Australia, and may indicate too large forced responses to anthropogenic aerosols. Furthermore, when we consider changes in the land-ocean temperature contrast over this period, the observed increase in LO towards the end of the 20th century is reflected in ACCESS1.3, although the ACCESS1.3 simulated LO changes are again predominantly lower during recent decades than those observed and than the CMIP5 multi-model range. In addition, there is a large spread of historical precipitation responses to anthropogenic forcings over the period of 1910–2005 within ACCESS1.3 and also within CMIP5, although not all calculated trends were found to be statistically significant.

Next, we considered simulated precipitation and temperature responses following three major late 20th century volcanic eruptions. A composite modelled volcanic response was calculated, as a mean of the three eruptions, weighted by the cumulative aerosol optical depth in the calendar year that each major eruption occurred. The complexity of observed post-volcanic climatological changes is generally not well captured by ACCESS1.3, or more broadly by the CMIP5 models. When we compared simulated global areal average composite volcanic temperature changes with those observed, the simulated temperature changes are larger than observed (ΔTmean −0.04 K) in both ACCESS1.3 (ΔTmean −0.07 to −0.12 K) the CMIP5 ensemble (ΔTmean −0.03 to −0.21 K, mean −0.13 K). In addition, while post-volcanic surface warming has been observed over the northern hemisphere during the boreal winter (Stenchikov et al. 2006), modelled changes are consistently of the incorrect sign to those observed during this season. The ACCESS1.3 composite post-volcanic winter temperature anomalies for the northern hemisphere lie within, or close to, the range of the CMIP5 ensemble. There is also a seasonally-consistent overestimate of observed post-volcanic cooling in the southern hemisphere in ACCESS1.3 and CMIP5. These differences between the seasonally and regionally complex volcanic responses that are observed and the pervasive modelled post-volcanic temperatures decreases may relate to model deficiencies in capturing post-eruption circulation changes during boreal winters and the crude implementation of volcanic eruptions in the models using changes in stratospheric aerosol optical depths in four discrete latitudinal bands, rather than smoothly varying changes (Driscoll et al. 2012). Furthermore, substantial post-volcanic drops in precipitation are simulated following all three major 20th century eruptions, with a large spread of responses modelled across the model ensemble for the historical experiment for both ACCESS1.3 and CMIP5 resulting in part from natural variability. Previously, observed precipitation anomalies following volcanic eruptions have been linked to direct radiative effects, from reductions in solar radiation leading to reduced evaporative rates and indirect effects on cloud microphysics (Gu and Adler 2011). In particular, previous studies have noted that model volcanic aerosols remain in the atmosphere longer


than is physically realistic and hence simulated precipitation changes, such as those indicated here, are too substantial as a result (Ren et al. 2013).

This study was not intended to be an exhaustive evaluation of ACCESS1.3 against observations and against other CMIP5 models. Rather, we used a limited number of metrics to investigate forced model responses and highlight any notable model imperfections that may be relevant to users of these experiments. Overall, ACCESS1.3 demonstrates skill in simulating the global and regional metrics assessed here, relative to observations, which is comparable to the CMIP5 multi-model ensemble. As such, these experiments with various historical forcings will be useful for inclusion with other CMIP5 contributions for studies aimed at detecting and attributing climatic changes. For example, Jones et al. (2013) used a suite of CMIP5 historical, historicalGHG and historicalNat simulations to determine that anthropogenic greenhouse gases are the dominant cause of observed global warming since the mid-20th century. In this case, the inclusion of multiple realisations from numerous, distinct models allowed the uncertainty in model responses to anthropogenic aerosol forcings to be better constrained. Despite the general similarity of historical forced responses in ACCESS1.3 and CMIP5, it remains necessary for future studies to evaluate these ACCESS1.3 simulations rigorously in order to determine how well the model represents the actual climate for specific regions and climate variables and hence establish the validity of its use.

The ACCESS1.3 model, like all global climate models, has deficiencies that should be considered. This study has identified aspects of ACCESS1.3 simulated responses to forcings that warrant further investigations. In particular, one of most notable features of ACCESS1.3 historical simulations is the rather subdued warming trend that is evident in all ACCESS1.3 realisations for both global mean temperatures and land-ocean temperature contrasts, relative to observed and to the CMIP5 multi-model mean. This indicates that the response to anthropogenic aerosol forcing in ACCESS1.3 may be too large. The three ensemble members for ACCESS1.3 now available for use for each of the historical, historicalGHG and historicalNat experiments will allow any potential differences between ACCESS1.3 and both observations and the other CMIP5 models, such as these disparate warming rates, to be more readily attributed to either model sensitivity or natural variability, which was not previously possible (Dix et al. 2013). This strong forced response to anthropogenic aerosols could be addressed quantitatively using additional ‘sstClim’-style single forcing experiments to explicitly calculate the effective radiative forcing of various factors in ACCESS1.3 (e.g. Shindell et al. 2013, Rotstayn et al. 2012). In addition, the extension of the ACCESS1.3 simulations described here beyond the standard historical period of 1850–2005 through to 2020 or beyond would provide a useful context for understanding recent climatic trends and events. The extension of the existing suite of simulations is currently ongoing.

Acknowledgments

This research was supported by funding from the Australian Research Council Centre of Excellence for Climate System Science (grant CE 110001028). The computation for this work was performed at the NCI National Facility at the ANU. We particularly thank the Bureau of Meteorology and CSIRO for their development of ACCESS and their ongoing support in conducting CMIP5 simulations. We thank also Tony Hirst and an anonymous reviewer for their constructive comments. We also acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank climate modelling groups for producing and making available their model output. For CMIP, the U.S. Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals.

References Bi, D., Dix, M., Marsland, S. J., O’Farrell, S., Rashid, H., Uotila, P., Hirst,

A. C., Kowalczyk, E., Golebiewski, M. and Sullivan, A. 2013. The AC-CESS coupled model: Description, control climate and evaluation, Aust. Met. Oceanogr. J., 63, 9–32.

Bindoff, N.L., Stott, P. A., AchutaRao, K. M., Allen, M. R., Gillett, N., Gutzler, D., Hansingo, K., Hegerl, G., Hu, Y., Jain, S., Mokhov, I. I., Overland, J., Perlwitz, J., Sebbari, R. and Zhang, X. 2013. Detection and Attribution of Climate Change: from Global to Regional, in Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, edited by F. T. Stocker, D. Qin, G. K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P. M. Midgley, Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. pp. 1–86.

Brohan, P., Kennedy, J. J., Harris, I., Tett, S. F. B. and Jones, P. D. 2006. Uncertainty estimates in regional and global observed temperature changes: A new data set from 1850, J. Geophys. Res-Atmos., 111(D12), D12106, doi:10.1029/2005JD006548.

Dix, M., Vohralik, P., Bi, D., Rashid, H., Marsland, S., O’Farrell, S., Uo-tila, P., Hirst, T., Kowalczyk, E., Sullivan, A., Yan, H., Franklin, C., Sun, Z., Watterson, I., Collier, M., Noonan, J., Rotstayn, L., Stevens, L. and Puri, P. U. A. K. 2013. The ACCESS coupled model: documentation of core CMIP5 simulations and initial results, Aust. Met. Oceanogr. J., 63, 83–99.

Driscoll, S., Bozzo, A., Gray, L. J. and Robock, A. 2012. Coupled Model Inter-comparison Project 5 (CMIP5) simulations of climate following volcanic eruptions, J. Geophys. Res-Atmos., doi:10.1029/2012JD017607/full.

Drost, F. and Karoly, D. 2012. Evaluating global climate responses to dif-ferent forcings using simple indices, Geophys. Res. Lett., 39 (L16701), doi:10.1029/2012GL052667.

Gillett, N. P. 2004. Detection of volcanic influence on global precipitation, Geophys. Res. Lett., 31(12), L12217, doi:10.1029/2004GL020044.

Gregory, J. M., Bi, D., Collier, M. A., Dix, M. R., Hirst, A. C., Hu, A., Hu-ber, M., Knutti, R., Marsland, S. J., Meinshausen, M., Rashid, H. A., Rotstayn, L. D., Schurer, A. and Church, J. A. 2013. Climate models without preindustrial volcanic forcing underestimate historical ocean thermal expansion, Geophys. Res. Lett., 40 (8), 1600–04, doi:10.1002/grl.50339.

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. and Zadeh, N. 2011. The GFDL CM3 coupled climate model: Charac-teristics of the ocean and sea ice simulations, J. Clim., 24 (13), 3520–44, doi:10.1175/2011JCLI3964.1.


Gu, G. and Adler, R. F. 2011. Precipitation and Temperature Variations on the Interannual Time Scale. Assessing the Impact of ENSO and Volca-nic Eruptions, J. Clim., 24 (9), 2258–2270, doi:10.1175/2010JCLI3727.1.

Harris, I., Jones, P. D., Osborn, T. J. and Lister, D. H. 2013. Updated high-resolution grids of monthly climatic observations–the CRU TS3. 10 Dataset, Int. J. Climatol., doi:10.1002/joc.3711.

Hewitt, H. T., Copsey, D., Culverwell, I. D., Harris, C. M., Hill, R. S. R., Keen, A. B., McLaren, A. J. and Hunke, E. C. 2011. Design and imple-mentation of the infrastructure of HadGEM3: the next-generation Met Office climate modelling system, Geoscientific Model Development, 4 (2), 223–53, doi:10.5194/gmd-4-223-2011.

Hunke, E. C. and Lipscomb, W. H.: CICE. 2008. the Los Alamos Sea Ice Model Documentation and Software User’s Manual Version 4.0, LA-CC-06–012.

Iles, C. E., Hegerl, G. C., Schurer, A. P. and Zhang, X. 2013. The effect of volcanic eruptions on global precipitation, J. Geophys. Res. Atmos., 118, 8770–86, doi:10.1002/jgrd.50678.

Jones, C. D., Hughes, J. K., Bellouin, N., Hardiman, S. C., Jones, G. S., Knight, J., Liddicoat, S., O’Connor, F. M., Andres, R. J., Bell, C., Boo, K. O., Bozzo, A., Butchart, N., Cadule, P., Corbin, K. D., Doutriaux-Boucher, M., Friedlingstein, P., Gornall, J., Gray, L., Halloran, P. R., Hurtt, G., Ingram, W. J., Lamarque, J. F., Law, R. M., Meinshausen, M., Osprey, S., Palin, E. J., Parsons Chini, L., Raddatz, T., Sanderson, M. G., Sellar, A. A., Schurer, A., Valdes, P., Wood, N., Woodward, S., Yosh-ioka, M. and Zerroukat, M. 2011. The HadGEM2-ES implementation of CMIP5 centennial simulations, Geoscientific Model Development, 4 (3), 543–70, doi:10.5194/gmd-4-543–2011.

Jones, G. S., Stott, P. A. and Christidis, N. 2013. Attribution of observed historical near-surface temperature variations to anthropogenic and natural causes using CMIP5 simulations, J. Geophys. Res. Atmos., 118, 4001–24, doi:10.1002/jgrd.50239.

Knutson, T. R., Zeng, F. and Wittenberg, A. T. 2013. Multimodel Assess-ment of Regional Surface Temperature Trends: CMIP3 and CMIP5 Twentieth-Century Simulations, J. Clim., 26 (22), 8709–43, doi:10.1175/JCLI-D-12-00567.1.

Kowalczyk, E., 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, 1–42 [online] Available from: www.cmar.csiro.au/e-print/open/kowalczykea_2006a.pdf.

Kumar, S., Merwade, V., Kinter, J. L., III and Niyogi, D. 2013. Evaluation of Temperature and Precipitation Trends and Long-Term Persistence in CMIP5 Twentieth-Century Climate Simulations, J. Clim., 26 (12), 4168–4185, doi:10.1175/JCLI-D-12-00259.1.

Lean, J. 2009. Calculations of Solar Irradiance, sparcsolaris.geomar.de, 1–4 [online] Available from: http://sparcsolaris.geomar.de/Input_data/Calculations_of_Solar_Irradiance.pdf (Accessed 20 November 2013).

Lewis, S. C. and Karoly, D. J. 2013. Anthropogenic contributions to Aus-tralia’s record summer temperatures of 2013, Geophys. Res. Lett., doi:10.1002/grl.50673.

Marshall, A. G., Scaife, A. A. and Ineson, S. 2009. Enhanced Seasonal Pre-diction of European Winter Warming following Volcanic Eruptions, J. Clim., 22 (23), 6168–80, doi:10.1175/2009JCLI3145.1.

Martin, G. M., Bellouin, N., Collins, W. J., Culverwell, I. D., Halloran, P. R., Hardiman, S. C., Hinton, T. J., Jones, C. D., McDonald, R. E., McLaren, A. J., O’Connor, F. M., Roberts, M. J., Rodriguez, J. M., Woodward, S., Best, M. J., Brooks, M. E., Brown, A. R., Butchart, N., Dearden, C., Derbyshire, S. H., Dharssi, I., Doutriaux-Boucher, M., Edwards, J. M., Falloon, P. D., Gedney, N., Gray, L. J., Hewitt, H. T., Hobson, M., Huddleston, M. R., Hughes, J., Ineson, S., Ingram, W. J., James, P. M., Johns, T. C., Johnson, C. E., Jones, A., Jones, C. P., Joshi, M. M., Keen, A. B., Liddicoat, S., Lock, A. P., Maidens, A. V., Manners, J. C., Milton, S. F., Rae, J. G. L., Ridley, J. K., Sellar, A., Senior, C. A., Totterdell, I. J., Verhoef, A., Vidale, P. L. and Wiltshire, A. 2011. The HadGEM2 fam-ily of Met Office Unified Model climate configurations, Geoscientific Model Development, 4 (3), 723–57, doi:10.5194/gmd-4-723-2011.

Meinshausen, M., Smith, S. J., Calvin, K., Daniel, J. S., Kainuma, M. L. T., Lamarque, J. F., Matsumoto, K., Montzka, S. A., Raper, S. C. B., Riahi, K., Thomson, A., Velders, G. J. M. and van Vuuren, D. P. P. 2011. The RCP greenhouse gas concentrations and their extensions from 1765 to 2300, Climatic Change, 109 (1–2), 213–41, doi:10.1007/s10584-011-0156-z.

Morice, C. P., Kennedy, J. J., Rayner, N. A. and Jones, P. D. 2012. Quanti-fying uncertainties in global and regional temperature change using an ensemble of observational estimates: The HadCRUT4 data set, J. Geophys. Res. Atmos., 117 (D8), D08101, doi:10.1029/2011JD017187.

Perlwitz, J. and Graf, H.-F. 2001. Troposphere-stratosphere dynamic cou-pling under strong and weak polar vortex conditions, Geophys. Res. Lett., 28 (2), 271–74, doi:10.1029/2000GL012405.

Ren, L., Arkin, P. and Smith, T. M. 2013. Global precipitation trends in 1900–2005 from a reconstruction and coupled model simulations, J. Geophys. Res. Atmos., doi:10.1002/jgrd.50212.

Rotstayn, L. D., Jeffrey, S. J., Collier, M. A., Dravitzki, S. M., Hirst, A. C., Syktus, J. I. and Wong, K. K. 2012. Aerosol- and greenhouse gas-induced changes in summer rainfall and circulation in the Austral-asian region: a study using single-forcing climate simulations, Atmos. Chem. Phys., 12 (14), 6377–404, doi:10.5194/acp-12-6377-2012.

Rupp, D. E., Mote, P. W., Bindoff, N. L., Stott, P. A. and Robinson, D. A. 2013. Detection and Attribution of Observed Changes in North-ern Hemisphere Spring Snow Cover, J. Clim., 26 (18), 6904–6914, doi:10.1175/JCLI-D-12-00563.1.

Sato, M., Hansen, J. E., McCormick, M. P. and Pollack, J. B. 1993. Strato-spheric aerosol optical depths, 1850–1990, J. Geophys. Res. Atmos., 98 (D12), 22987–94.

Schneider, U., Becker, A., Finger, P., Meyer-Christoffer, A., Ziese, M. and Rudolf, B. 2013. GPCC’s new land surface precipitation climatology based on quality-controlled in situ data and its role in quantifying the global water cycle, Theoretical and Applied Climatology, doi:10.1007/s00704-013-0860-x.

Sen, P. K. 1968. Estimates of the Regression Coefficient Based on Kendall’s Tau, Journal of the American Statistical Association, 63 (324), 1379–89.

Shindell, D. T., Lamarque, J. F., Schulz, M., Flanner, M., Jiao, C., Chin, M., Young, P. J., Lee, Y. H., Rotstayn, L., Mahowald, N., Milly, G., Falu-vegi, G., Balkanski, Y., Collins, W. J., Conley, A. J., Dalsoren, S., Easter, R., Ghan, S., Horowitz, L., Liu, X., Myhre, G., Nagashima, T., Naik, V., Rumbold, S. T., Skeie, R., Sudo, K., Szopa, S., Takemura, T., Voulgara-kis, A., Yoon, J. H. and Lo, F. 2013. Radiative forcing in the ACCMIP historical and future climate simulations, Atmos. Chem. Phys., 13 (6), 2939–74, doi:10.5194/acp-13-2939-2013.

Shindell, D. T., Schmidt, G. A., Miller, R. L. and Rind, D. 2001. Northern Hemisphere winter climate response to greenhouse gas, ozone, solar, and volcanic forcing, J. Geophys. Res. Atmos., 106(D7), 7193–210.

Soden, B. J. 2002. Global Cooling After the Eruption of Mount Pinatubo: A Test of Climate Feedback by Water Vapor, Science, 296 (5568), 727–30, doi:10.1126/science.296.5568.727.

Stenchikov, G., Hamilton, K., Stouffer, R. J., Robock, A., Ramaswamy, V., Santer, B. and Graf, H.-F. 2006. Arctic Oscillation response to volcanic eruptions in the IPCC AR4 climate models, J. Geophys. Res. Atmos., 111 (D7), D07107, doi:10.1029/2005JD006286.

Stott, P. A. and Jones, G. S. 2012. Observed 21st century temperatures further constrain likely rates of future warming, Atmos. Science Lett., 13 (3), 151–56, doi:10.1002/asl.383.

Taylor, K. E., Stouffer, R. J. and Meehl, G. A. 2012. An overview of CMIP5 and the experiment design, Bull. Am. Meteorol. Soc., 93(4), 485, doi:10.1175/BAMS-D-11-00094.1.

Trewin, B. 2012. A daily homogenized temperature data set for Australia, Int. J. Climatol., doi:10.1002/joc.3530,.

Valcke, S., Craig, T. and Coquart, L. 2013. OASIS3-MCT User Guide OA-SIS3-MCT 2.0.

Xinyu, W., Guoli, T., Shaowu, W. and Jianbin, H. 2011. Comparison of Global Mean Temperature Series, Advances in Climate Change Re-search, 2 (4), 187–192, doi:10.3724/SP.J.1248.2011.00187.

Assessment of forced responses of the Australian · PDF filemodelling groups have contributed to the fifth phase ... 2013. As such, ‘tier 1’ and ‘tier 2’ CMIP5 ......

Documents