Future Australian Severe Thunderstorm Environments. Part I: A Novel Evaluation and Climatology of Convective Parameters from Two Climate Models for the Late Twentieth Century

John T. Allen School of Earth Sciences, University of Melbourne, Victoria, Australia, and International Research Institute for Climate and Society, The Earth Institute, Columbia University, Palisades, New York

Search for other papers by John T. Allen in
Current site
Google Scholar
PubMed
Close
,
David J. Karoly School of Earth Sciences, University of Melbourne, Victoria, Australia

Search for other papers by David J. Karoly in
Current site
Google Scholar
PubMed
Close
, and
Kevin J. Walsh School of Earth Sciences, University of Melbourne, Victoria, Australia

Search for other papers by Kevin J. Walsh in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

The influence of a warming climate on the occurrence of severe thunderstorms over Australia is, as yet, poorly understood. Based on methods used in the development of a climatology of observed severe thunderstorm environments over the continent, two climate models [Commonwealth Scientific and Industrial Research Organisation Mark, version 3.6 (CSIRO Mk3.6) and the Cubic-Conformal Atmospheric Model (CCAM)] have been used to produce simulated climatologies of ingredients and environments favorable to severe thunderstorms for the late twentieth century (1980–2000). A novel evaluation of these model climatologies against data from both the ECMWF Interim Re-Analysis (ERA-Interim) and reports of severe thunderstorms from observers is used to analyze the capability of the models to represent convective environments in the current climate. This evaluation examines the representation of thunderstorm-favorable environments in terms of their frequency, seasonal cycle, and spatial distribution, while presenting a framework for future evaluations of climate model convective parameters. Both models showed the capability to explain at least 75% of the spatial variance in both vertical wind shear and convective available potential energy (CAPE). CSIRO Mk3.6 struggled to either represent the diurnal cycle over a large portion of the continent or resolve the annual cycle, while in contrast CCAM showed a tendency to underestimate CAPE and 0–6-km bulk magnitude vertical wind shear (S06). While spatial resolution likely contributes to rendering of features such as coastal moisture and significant topography, the distribution of severe thunderstorm environments is found to have greater sensitivity to model biases. This highlights the need for a consistent approach to evaluating convective parameters and severe thunderstorm environments in present-day climate: an example of which is presented here.

Corresponding author address: John T. Allen, International Research Institute for Climate and Society, Lamont Doherty Earth Observatory, P.O. Box 1000, 61 Route 9W, Palisades, NY 10964-1000. E-mail: johnterrallen@gmail.com

Abstract

The influence of a warming climate on the occurrence of severe thunderstorms over Australia is, as yet, poorly understood. Based on methods used in the development of a climatology of observed severe thunderstorm environments over the continent, two climate models [Commonwealth Scientific and Industrial Research Organisation Mark, version 3.6 (CSIRO Mk3.6) and the Cubic-Conformal Atmospheric Model (CCAM)] have been used to produce simulated climatologies of ingredients and environments favorable to severe thunderstorms for the late twentieth century (1980–2000). A novel evaluation of these model climatologies against data from both the ECMWF Interim Re-Analysis (ERA-Interim) and reports of severe thunderstorms from observers is used to analyze the capability of the models to represent convective environments in the current climate. This evaluation examines the representation of thunderstorm-favorable environments in terms of their frequency, seasonal cycle, and spatial distribution, while presenting a framework for future evaluations of climate model convective parameters. Both models showed the capability to explain at least 75% of the spatial variance in both vertical wind shear and convective available potential energy (CAPE). CSIRO Mk3.6 struggled to either represent the diurnal cycle over a large portion of the continent or resolve the annual cycle, while in contrast CCAM showed a tendency to underestimate CAPE and 0–6-km bulk magnitude vertical wind shear (S06). While spatial resolution likely contributes to rendering of features such as coastal moisture and significant topography, the distribution of severe thunderstorm environments is found to have greater sensitivity to model biases. This highlights the need for a consistent approach to evaluating convective parameters and severe thunderstorm environments in present-day climate: an example of which is presented here.

Corresponding author address: John T. Allen, International Research Institute for Climate and Society, Lamont Doherty Earth Observatory, P.O. Box 1000, 61 Route 9W, Palisades, NY 10964-1000. E-mail: johnterrallen@gmail.com

1. Introduction

The response of severe convective storms to a warming climate is far from well understood. In both the public and scientific communities, there is increasing concern that the frequency and intensity of these events may be changing as a result of anthropogenic influences on the climate system (Bouwer 2011; Brooks 2013). However, climatological records have suggested that no significant trends in frequency or intensity have yet occurred (Balling and Cerveny 2003; Changnon 2003; Trenberth et al. 2007). Because of difficulties with observational records, caution must be taken when considering changes in severe thunderstorms and resulting extreme events (Doswell et al. 2005; Brooks 2013).

Despite the occurrence of particularly damaging severe thunderstorms in recent years (e.g., Allen 2012), the implications are unclear for a warming climate over Australia. Station-based observations have suggested little trend in days producing thunderstorms in either the United States or Australia for the last century (Kuleshov et al. 2002; Changnon 2003). Although the observational database of severe storms in Australia has improved substantially in the past 20 yr (Griffiths et al. 1993; Schuster et al. 2005; Allen et al. 2011), it is not sufficient to perform the trend analysis required to look at past changes in these relatively infrequent but high-impact events. It has been suggested that a decrease in the frequency of hail over Australia appears to negate the spurious positive observational trends in the dataset that result from increasing population and reporter density (Schuster et al. 2005; Nicholls 2008). However, the extent to which regional studies (e.g., Schuster et al. 2005) can be considered representative of the continent is questionable. Thus, lacking an adequate source of observational records in order to identify a trend, an alternative option is to consider changes to the occurrence of convective environments using model or reanalysis-derived soundings.

The applicability of derived pseudosoundings has been well established for both reanalysis and model data over the United States (J. Lee 2002, personal communication; Brooks et al. 2003; Brooks 2009) as well as Australia (Allen et al. 2011; Allen and Karoly 2014). Derived soundings have been found to be reasonably effective at simulating relevant convective quantities. These fall into two categories: ingredients refer to simple quantities such as temperature, moisture, or wind fields, while parameters describe a greater portion of the atmospheric state [e.g., convective available potential energy (CAPE) and vertical shear of the deep-layer horizontal wind (VWS)]. While both CAPE and VWS within the model and reanalysis data have limitations (e.g., Brooks et al. 2003; Allen et al. 2011; Allen and Karoly 2014), over the majority of continental areas there is sufficient evidence to suggest they are a useful proxy for rawinsonde data. Previous work has described the climatology of Australian severe thunderstorm environments and ingredients from the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Re-Analysis (ERA-Interim; Dee et al. 2011) and discussed how environments can be defined (Allen and Karoly 2014). These environments can also be determined from other sources such as climate models, particularly when information on the influence of climate warming on these environments is sought (e.g., Marsh et al. 2007; Trapp et al. 2007a; Diffenbaugh et al. 2013). While the ability of reanalysis datasets to simulate convective environments is reasonably well established, the capability of climate models to simulate these environments remains untested for most climate models. Over Australia, no climate models have been previously evaluated in this way.

To date, several limitations have been identified in climate models used to simulate convective environments. Convective inhibition is often poorly handled by limited vertical resolution, where the shallow inversion layer is poorly represented resulting from the lack of sufficient vertical levels in the large-scale model (Marsh et al. 2007, 2009; Diffenbaugh et al. 2013). There is also sensitivity to the convective parameterization scheme, particularly for CAPE, arising from the schemes eliminating energy from the atmosphere and the issues present in handling convective inhibition (Gettelman et al. 2002; Marsh et al. 2007). Additionally, subgrid-scale processes and topographic influences, both arising because of the relatively coarse horizontal grid scale, potentially contribute to problems with CAPE (Iorio et al. 2004; Niall and Walsh 2005). Despite these issues, global climate models are useful tools for the simulation of large-scale features and synoptic systems (Marsh et al. 2007, 2009; Van Klooster and Roebber 2009), suggesting the potential to resolve convective environments similar to reanalyses.

Whether climate models are capable of realistically simulating the ingredients and environments associated with severe convection and their climatology is a critical question when applying any dataset to simulate convective environments. Thus, before examining future simulations, the climatology produced by a climate model first requires evaluation to determine similarity to the observations and reanalysis climatology: in a sense, an evaluation of the model’s ability to represent convective environments. To this end, a novel, in-depth analysis of two climate models [the Commonwealth Scientific and Industrial Research Organisation Mark, version 3.6 (CSIRO Mk3.6) and Cubic-Conformal Atmospheric Model (CCAM) climate models] in their simulation of late twentieth century (1980–2000) severe thunderstorm environments over Australia was performed. As climate models do not actually represent the climate system identically to observed conditions, this is achieved by considering the mean seasonal cycle and comparing spatial distributions of convective ingredients and favorable environments. In this paper, the performance of the climate models in producing a climatology of severe thunderstorm environments has been evaluated against the observational climatology derived from the ERA-Interim (Allen and Karoly 2014) and a report climatology (Allen et al. 2011). This type of analysis is unique for the Australian region as no detailed evaluation exists of convective ingredients as produced by climate models, and how climate models simulate severe thunderstorm environments over the continent is not well understood. In the second part of this study, simulations with the same two models of the response to warmed future climate scenarios are used to assess potential changes and shifts to severe thunderstorm environment occurrences in the late twenty-first century (Allen et al. 2014, hereafter Part II).

The paper is organized as follows: The characteristics of the climate models and datasets are first discussed (section 2). The capability of the models to simulate the diurnal cycle is then explored (section 3). Consideration is then made of convective ingredients in comparison to the ERA-Interim climatology (section 4). Having compared these ingredients, the seasonal cycle and mean climatologies of severe thunderstorm environments are explored over both the east Australian region (EAReg; Fig. 1) and the full continent (sections 5 and 6). Finally, in section 7, the implications of this evaluation of CSIRO Mk3.6 and CCAM climate models for future climate projections of convection are discussed, and the case for considering changes resulting from warming as projected by these two models is examined.

Fig. 1.
Fig. 1.

Distribution of grid points over the Australian continent after filtering using the land–sea mask for (a) ERA-Interim, (b) CSIRO Mk3.6. and (c) CCAM. The dashed box outlines the EAReg, bound by 39°–20°S and 144°–154.5°E. Topographic height contours in terms of height above mean sea level are shaded over the continent, with light blue indicating the extent of the land–sea mask used to filter the environment climatology for grid points occurring over land only.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

2. Data and method

a. Verification datasets

To provide a baseline climatology for evaluation, ERA-Interim was used to produce a 20-warm-season (September–April) climatology of severe thunderstorm environments and ingredients for the period 1980–2000. The dataset used is of 1.5° horizontal resolution with 29 vertical levels and processed using the sounding analysis program developed by Allen and Karoly (2014). Over the Australian region, output was considered between 48° and 8°S and between 105° and 160°E, over land grid points. The variables evaluated include a 50-hPa mixed-layer CAPE (MLCAPE) with a most unstable equivalent potential temperature condition for parcels above the mixed layer and with the virtual temperature correction applied (CAPE; Doswell and Rasmussen 1994), 0–6-km bulk magnitude vertical wind shear (S06), and severe thunderstorm (SEV) environments as defined by the Australian discriminant covariate relationship (Allen et al. 2011),
e1
Details of the determination and suitability of this discriminant for Australian conditions are described in Allen et al. (2011) and Allen and Karoly (2014). As pointed out by Brooks (2013), this discriminant equation with S06 to the power of between 1.6 and 1.7 encompasses the conditions for severe thunderstorms for the United States, Europe, and Australia; thus, following earlier applications (e.g., Marsh et al. 2007, Diffenbaugh et al. 2013), the application of Eq. (1) to climate models is appropriate. In addition, this relationship is conditional on 700–500-hPa lapse rates being larger than 6.5 K km−1, with convective inhibition (CIN) less than 25 J kg−1. These two additional conditions are used to identify instances where the atmosphere is unstable in the midlatitudes and the inversion layer can be overcome, respectively.

Alongside the reanalysis climatology, the observational reports database developed by Allen et al. (2011) was used to produce a mean seasonal climatology to evaluate the models. The dataset is composed of 1550 severe and significant severe thunderstorm reports from the period March 2003–April 2010, producing a reasonable spatial sample of event occurrence over the continent, albeit within the limits of the population distribution (Allen and Karoly 2014). Because of the limitations of the reanalysis climatology, this dataset is used to examine the temporal distribution of environment occurrences and the spatial distribution of severe thunderstorm environments over the EAReg.

b. Climate models

Two climate models were chosen to represent convective environments and detect changes over Australia: the CSIRO Mk3.6 (Rotstayn et al. 2010; Jeffrey et al. 2013) and the selectively downscaled CCAM (McGregor 2005). CSIRO Mk3.6 is a fully coupled atmosphere–ocean global climate model (GCM) evolved from the CSIRO Mk3.0 and Mk3.5 models (Gordon et al. 2002, 2010), with interactive aerosols and minor improvements to the atmospheric physics. The historical, all-forcing simulation from phase 5 of the Coupled Model Intercomparison Project (CMIP5) was used for 20 warm seasons (September–April) for 1980–2000. CCAM is a specialized atmospheric global climate model incorporating atmospheric and oceanic forcing characteristics from a more coarsely resolved climate model (McGregor 2005; McGregor and Dix 2008). It uses these data to dynamically downscale to a higher-resolution quasi-uniform grid over focused areas to a desired resolution. The data driving CCAM utilize bias-corrected sea surface temperature (SST) and sea ice forcing from the CSIRO Mk3.5 coupled atmosphere–ocean global climate model (an earlier version of the CSIRO Mk3.6 described previously) but with no atmospheric forcing from the model. The bias-corrected SST inclusion makes the monthly climatology of SSTs the same as observations, removing inherent biases in GCMs relative to observed values (for further details, see Corney et al. 2013). This results in improved representation of spatial variations and extremes while maintaining interannual variability and seasonal cycle from the GCM data (Katzfey et al. 2009; Corney et al. 2010; Nguyen et al. 2012; Bennett et al. 2012; Corney et al. 2013; Grose et al. 2013). CCAM was chosen for its higher horizontal resolution (50 km) and is a superior atmospheric model when compared with the CMIP5 CSIRO Mk3.6 (175 km; Table 1). As for the reanalysis, severe convective environments were considered over land only, while constituent parameters were considered over the entire domain. Variables from both models included temperature, specific humidity, and the zonal and meridional wind fields on all model levels, with surface data added as the lowest layer for 6-hourly analyses. These were interpreted using the sounding analysis program, creating vertical profiles and analyzing necessary convective variables for comparison to the reanalysis data.

Table 1.

Description of model grid for the CSIRO Mk3.6 GCM, CCAM, and ERA-Interim used to produce the comparable climatologies for the period 1 Sep 1980–30 Apr 2000.

Table 1.

The difference in horizontal resolution strongly influences how the models represent the most significant feature of the Australian topography, the Great Dividing Range (Fig. 1). The Great Dividing Range stretches along the east coast of the continent, spanning from central Victoria to southeast Queensland, reaching heights between 1500 and 2228 m, with the highest peaks in the southeast. This feature is important to severe thunderstorm environments on the east coast of the continent, acting as a boundary between the coastal plains characterized by moist maritime airand the drier inland (Allen and Karoly 2014). Even in ERA-Interim, it is noticeable that the topography over the east coast is substantially lower than reality (Fig. 1a). Peak heights from ERA-Interim over this feature are 800–900 m, which, while enough to restrict advection, are unlikely to provide the blocking effect the feature produces. In comparison, CSIRO Mk3.6 barely resolves the feature at all (Fig. 1b), with heights no more than 400 m at any point over the divide. While this is a result of the coarser grid scale, it has the potential to greatly impact the distribution of moisture over the eastern half of the continent and the corresponding rainfall throughout southeast Australia (Randall et al. 2007; Rotstayn et al. 2010). In contrast, CCAM by virtue of substantially higher resolution shows an improved handling of the topographic features of the Great Dividing Range in general, as well as over the Australian Alps (Fig. 1c). This contrast in topographic representation has the potential to affect the results of the comparison of climatologies and thus must be taken into consideration in the evaluation of both ingredients and environments.

While the resolutions of the chosen datasets differ, both climate models apply a similar convective parameterization (mass-flux closure) with minor differences. The CSIRO Mk3.6 applies a scheme adapted from the Hadley Centre model with updrafts, downdrafts, and entrainment processes (Gregory and Rowntree 1990), with shallow convection also handled by this scheme. Cumulus convection in CCAM is handled by an Arakawa style mass-flux closure scheme that allows simultaneous plumes and includes downdrafts and detrainment, with shallow convection resolved by high detrainment of shallow clouds (McGregor 2003).

However, despite this similarity, both models have differing physics and corresponding characteristics over the continent. CSIRO Mk3.6 performs reasonably well in simulated ENSO-related SSTs, which is an important factor for severe environments (Allen and Karoly 2014), and has the highest skill of any of the CSIRO models in simulating both rainfall and temperature (Rotstayn et al. 2010; Jeffrey et al. 2013). Despite this, as with many models, CSIRO Mk3.6 has issues with the excessively westward displacement of ENSO-related SST anomalies (Cai et al. 2003; Randall et al. 2007). This can result in warmer temperatures over the northwest shelf of the Australian continent, influencing available moisture in the region (Rotstayn et al. 2010). There are also issues with precipitation, likely resulting from the overly moist continental mixing ratios previously identified in the CSIRO Mk3.5 (Willett et al. 2010) and the removal of the convective parameterization bias correction by cooling surface temperatures using increased cloud amount in convective columns (Rotstayn et al. 2010). These factors suggest that the model may exaggerate convective potential in the northern parts of Australia and over inland areas, where moisture during the historical runs is excessive compared to observational values.

CCAM has been applied for a range of CMIP3 models, with simulations using both CSIRO Mk3.0 and Mk3.5 producing a more accurate distribution of Australian rainfall and moisture compared to parent GCMs, particularly in the near-coast areas and near the complex terrain of the east coast (McGregor 2006). Validation against reanalysis data has suggested that the use of the “added value” of the higher-resolution CCAM produces a better spatial pattern and magnitude of precipitation, winds, and temperature compared to coarser global models (Katzfey et al. 2009; Nguyen et al. 2012; Grose et al. 2013; Corney et al. 2013; Smith et al. 2013). However, biases still exist in the location of midlatitude westerly winds that, while improved compared to many GCMs, are still displaced 3°–5° equatorward compared to reanalysis data (Grose et al. 2013). Simulations of the inland parts of the continent are also found to be overly dry relative to observations, with lower atmospheric moisture and reduced rainfall. This is partially related to the overly strong high pressure systems simulated by CCAM over the midlatitudes (Nguyen et al. 2012; Grose et al. 2013) and issues in the simulation of moisture advection within the model. Despite the limitations with inland moisture, this suggests that CCAM has the capability to simulate moisture and handle coastal topography, both of which are important aspects in identifying convective environments not necessarily well represented by CSIRO Mk3.6.

3. Diurnal distribution

An important aspect of evaluating the climate models is the choice of specific single or multiple times of day for analysis. In previous work (Allen and Karoly 2014), if any environment within a day was above the SEV threshold, the day was counted. However, GCMs are known to have issues associated with convective parameterization, resulting in the premature development of convective “environments” (Gettelman et al. 2002; Randall et al. 2007). Thus, the question is whether to consider all time steps in a day or a single time step at the diurnal peak in thunderstorm occurrence in the Australian region (0600 UTC analysis). The choice to neglect the remaining time steps is supported by analyzing how the models simulate the diurnal cycle over Australia.

There are noticeable differences between the diurnal cycle of reports and the normalized population of severe thunderstorm environments over the period 1980–2000. The environmental peak for both ERA-Interim and CCAM occurs at 0600 UTC (1700 local time; Figs. 2a,b), despite CCAM having a flatter diurnal cycle compared to observations and the reanalysis. Over the full continent, the relatively dry inland moisture field in CCAM contributes to the lower frequency of environments at the 0600 UTC time step (Fig. 3c). Examining the negligible difference between the CCAM at 0000 and 0600 UTC time-step frequency reveals that spatially nearly all of the environments occurring outside of 0600 UTC are found north of 20°S (Fig. 4c). While this is also found to a lesser extent in the reanalysis (Fig. 4a), this suggests less impact from incorrect representation of the diurnal cycle is found if the 0600 UTC time step is solely considered. Over the EAReg, this problem is lessened, with CCAM simulating a larger amplitude diurnal cycle that has a similar pattern to the reanalysis, with a reduced frequency of environments consistent with the contribution of a drier continental southeast (Figs. 2b, 3c).

Fig. 2.
Fig. 2.

Temporal distribution of mean SEV environments per season for the respective models and reanalysis for all grid points over (a) Australia and (b) EAReg as compared to severe thunderstorm reports over the comparable region for the period March 2003–April 2010. Climatology period corresponds to 20 warm seasons (September–April 1980–2000) for the ERA-Interim, CSIRO Mk3.6, and CCAM. Counts of environments at the respective time steps are normalized by the number of model grid points over the respective regions divided by 10. Black line and dots represent the frequency of reports by nearest hour in local time and equivalent temporal distribution of the favorable SEV environments produced from the ERA-Interim (black circles), CSIRO Mk3.6 (crossed circles), and CCAM (dotted circles) at the respective analysis times. Times correspond to Australian eastern daylight time or equivalently 1800, 0000, 0600, and 1200 UTC, respectively.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

Fig. 3.
Fig. 3.

Seasonal mean surface mixing ratio at 0600 UTC for warm seasons (September–April) over the years 1980–2000 for the Australian region from (a) ERA-Interim, (b) CSIRO Mk3.6, and (c) CCAM. Contour interval is 1 g kg−1 between 6 and 12 g kg−1, 2 g kg−1 between 12 and 20 g kg−1, and 4 g kg−1 between 20 and 28 g kg−1.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

Fig. 4.
Fig. 4.

Spatial distribution of the mean warm-season difference between SEV environments over Australia as determined by counting one SEV environment per day if an environment is detected for any time step vs counts for the single 0600 UTC diurnal peak time step over the years 1980–2000 for (a) ERA-Interim, (b) CSIRO Mk3.6, and (c) CCAM. Contour spacing between 0 and 10 is at 2 environment intervals, while contour spacing between 10 and 46 are at 4 environment intervals.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

In contrast, CSIRO Mk3.6 is unable to resolve the diurnal cycle of convection, with a greater frequency of environmental occurrences at both the 0000 and 1200 UTC time steps compared to both the observation and reanalysis peak (Figs. 2a,b). This difference arises from the coarser grid scale and the disparity in the surface moisture field of CSIRO Mk3.6 (Fig. 3b). Unsurprisingly, subgrid-scale processes in the boundary layer and topographical features are smoothed within this model (Fig. 1) and likely contribute to the poor timing of convective release (Iorio et al. 2004; Niall and Walsh 2005). While high moisture values in CSIRO Mk3.6 are coastally confined and elevated compared to both the reanalysis and CCAM, values over the continental interior differ by 2–3 g kg−1 (Fig. 3b). Considering where environments outside of 0600 UTC occur, for CSIRO Mk3.6 the contribution extends over the southeast, reflecting areas where the moisture is higher (Fig. 4b). The premature (and extended) occurrence of convection within the model indicates that the mass-flux scheme is activated too early in the day, with contributions from excessive boundary layer moisture (Fig. 3b) and poor resolving of convective inhibition (not shown). This poor performance for convective inhibition is not unexpected and is related to the limitations of convective parameterization schemes and model vertical resolution (Marsh et al. 2007). These limitations within the simulation of the diurnal cycle for CSIRO Mk3.6 and to a lesser extent ERA-Interim and CCAM suggest that the most suitable approach is to look at favorable environments in the time period at which the greatest potential for severe convection is expected: that is, the peak diurnal period for eastern Australia, 0600 UTC (1700 eastern daylight savings time).

4. Evaluation of convective ingredients

To evaluate the capability of the climate models to represent convective ingredients, comparison is made to the reanalysis climatology. Insomuch as reanalysis is a representation of the present climatology, it would be expected that climatological characteristics of CAPE and S06 occurrence and magnitude should be similar. Biases within the model climatologies are detectable following normalization by the number of grid points over Australia and the EAReg, respectively. To produce comparable spatial grids for statistical analysis, the data of the higher-resolution dataset are regridded to the coarser spatial scale of the two datasets. For CSIRO Mk3.6, this regridding is completed for the ERA-Interim (1.5° native resolution) using a procedure of local area averaging to interpolate from the higher-resolution rectilinear grid to a lower-resolution rectilinear grid (1.875°). For CCAM, the data from the climate model (0.5° native resolution) are regridded to the coarser ERA-Interim grid (1.5°). While the process is not strictly a conservative remapping, it yields a sufficiently accurate grid for calculation of pattern congruence and correlation. Performance in simulating the spatial variance of the mean climatology for each of the convective ingredients (CAPE, S06, and 700–500-hPa lapse rates) was considered using a centered pattern correlation (Richman 1986; Wilks 2006), with the tested levels of agreement between the reanalysis and the respective models classified by squaring this correlation as follows: less than 50% of spatial variance (correlation <0.71) explained is referred to as poor, greater than 50% (>0.71) is referred to as good, greater than 75% (>0.87) is referred to as very good, and greater than 90% (>0.95) is referred to as excellent. Pattern congruence (uncentered pattern correlation) of the spatial means between reanalysis and models was also calculated to evaluate spatial similarity. The pattern congruence (similarity) of two spatially equivalent temporal variables is then calculated by taking the dot product of the temporal mean (Wilks 2006): that is,
e2
However, given that pattern congruences of random fields would be expected to have values of 0.7, assuming 2 degrees of freedom (Richman 1986), this is less useful for evaluating the mean ingredient fields and more relevant for environmental occurrences (sections 5 and 6).

a. CAPE

Consideration was made of the CAPE distribution to evaluate the potential model bias (Figs. 5a,b). The CAPE distribution of the climate models is reasonably close to that of the reanalysis. However, CSIRO Mk3.6 has an unusually high number of environments between 2000 and 3600 J kg−1 compared to the ERA-Interim. This is not the case over the EAReg, suggesting the differences originate from grid points outside this region. Comparing the spatial distribution from CSIRO Mk3.6 to ERA-Interim (Fig. 6 and Table 3), the model produces from very good to excellent agreement for the spatial variance of mean CAPE, particularly during December–February (DJF; Fig. 6d) and March–April (MA; not shown). From September to November (SON), CSIRO Mk3.6 overestimates mean CAPE over the east coast, extending the mean CAPE field farther inland and south than ERA-Interim. There are also monthly biases within the seasons: pattern congruence is lower during September and October compared to SON (0.92), where CAPE is underestimated. Pattern correlations for CAPE over the EAReg are also decreased to good levels of agreement during this part of the season. During DJF and MA, mean CAPE values are much higher over the northern parts of the continent in CSIRO Mk3.6, extending farther south across the continent, particularly along both coastlines. This overestimation in the northern parts of the continent suggests that the majority of excessively high CAPE values occur in the northwest of Australia, which is supported by the steep model lapse rates in this region juxtaposed with higher moisture (Figs. 7c,d). The primary source of this moisture is the warmer SSTs in the oceans to the northwest of the continent (not shown). It is also likely that the higher values of CAPE over inland areas, particularly during the summer months, can be attributed to issues with the moisture field (Fig. 3; Willett et al. 2010).

Fig. 5.
Fig. 5.

Frequency distribution of nonzero MLCAPE environments as a fraction of the total occurrences of nonzero MLCAPE for the 20-season climatology for each of the CSIRO Mk3.6, CCAM, and ERA-Interim datasets for all grid points over (a) Australia and (b) the EAReg and the relative frequency distribution of S06 for nonzero MLCAPE environments as a ratio of total environments with nonzero MLCAPE for (c) Australia and (d) the EAReg.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

Fig. 6.
Fig. 6.

Comparison among ERA-Interim, CSIRO Mk3.6, and CCAM of seasonal mean MLCAPE for (a),(c),(e) SON and (b),(d),(f) DJF. Values in the bottom right of (c),(d) correspond to pattern congruences over the Australian region between CSIRO Mk3.6 and the reanalysis regridded to the models spatial resolution, while (e),(f) correspond to pattern congruence between the reanalysis and CCAM, with the model regridded to the ERA-Interim spatial resolution.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

Fig. 7.
Fig. 7.

As in Fig. 6, but for the comparison of mean seasonal 700–500-hPa vertical lapse rate among ERA-Interim, CSIRO Mk3.6, and CCAM.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

In contrast, low CCAM CAPE frequency is overrepresented, while the higher values of CAPE tend to be less frequent than in ERA-Interim (Fig. 5). Potential sources of this difference are horizontal mixing within the higher-resolution downscaled model and the relatively low moisture values identified for inland areas (Fig. 3c). Over the EAReg, CCAM better represents the frequency of CAPE between 400 and 1000 J kg−1 while slightly underestimating the high-end values and overestimating relative to the reanalysis by 20% between 10 and 200 J kg−1. The spatial distribution of CAPE throughout the season from CCAM is also closer to that of the ERA-Interim with near-excellent agreement in all seasons (Table 3). During SON (Fig. 6e), mean CAPE does not extend as far inland, with peak mean CAPE over northern Australia lower for CCAM than ERA-Interim, with maximum contours of 400 J kg−1 for CCAM, contrasting 800 J kg−1 in the reanalysis. Despite these differences, the east coast pattern of mean CAPE for CCAM visually appears close in peak magnitude and distribution to ERA-Interim, despite only good agreement for pattern correlations during SON and DJF. Peak magnitudes are similar and high pattern congruence is found for all seasons, and pattern correlation is higher in SON from CCAM than CSIRO Mk3.6. Minor differences between ERA-Interim and CCAM also exist in the southern extent of the nonzero CAPE field, particularly over the western half of the continent. Poor pattern correlation is found over the EAReg for MA, despite similar inland extent to the reanalysis. Based on this comparison, we can conclude that CCAM is excellent in representing the spatial variance of mean CAPE and its seasonal cycle well over the continent, with limitations over the EAReg, where the frequency of CAPE environments is heavily biased toward the lower end of the distribution, particularly during MA.

While the high levels of agreement suggest that the model climatologies are similar to ERA-Interim, there are a number of issues with modeled CAPE. We also point out that, while the reanalysis does offer a reasonable picture of the mean field, it cannot be viewed as a perfect observation of the distribution but rather an estimate of observations. Comparison to rawinsonde soundings (Allen and Karoly 2014) previously identified that the ERA-Interim tends to underestimate peak CAPE values and suffers from a number of issues relating to near-coastal areas and complex topography. There are also known differences in inland moisture because of sparse surface observations (Fig. 3; Simmons et al. 2010), which also impacts the mean CAPE as simulated by the ERA-Interim. Despite these limitations, high magnitude values over the northern part of Australia from CSIRO Mk3.6 appear to be unrealistic, and this model produces lower agreement than CCAM. The increased spatial resolution of CCAM does appear to improve the handling of coastal moisture boundaries in the mean over the east coast; however, the negative bias in CAPE magnitude in this region is pronounced.

CIN was also considered, but most climate models handle convective inhibition poorly because of issues with the limited vertical resolution (Marsh et al. 2007; Brooks 2013). As was noted by Allen and Karoly (2014), CIN is not particularly well represented by the reanalysis either, but how the climate models compare for nonzero CAPE environments can be considered. Over Australia, the CSIRO Mk3.6 underestimates CIN, with relative frequency of 20% of the lowest categories of nonzero ERA-Interim CIN (not shown). CCAM is slightly improved relative to the CSIRO Mk3.6, with the relative fraction of nonzero CIN environments less than half. Given this weakness, it is plausible that environments with appreciable CAPE may have been mixed prior to 0600 UTC in the models’ diurnal cycles, especially in CSIRO Mk3.6, which struggles to identify this aspect. Thus, some of the differences present in the CAPE distribution both in relative fraction and spatial distribution may be related to this issue, and the results must be considered carefully.

b. S06

In comparison to the distribution of CAPE, the distribution of S06 values for nonzero CAPE is closer to ERA-Interim for both climate models. CSIRO Mk3.6 slightly overestimates the occurrence of high S06 environments over both the full continent and the EAReg (Figs. 5c,d). This result corresponds with the comparisons made for other models (e.g., Marsh et al. 2007; Diffenbaugh et al. 2013) as well as comparisons made between ERA-Interim and observational rawinsonde data (Allen and Karoly 2014). The spatial distribution of S06 from the CSIRO Mk3.6 for SON has high pattern congruence and good agreement; however, pattern correlations for this period are very poor over the EAReg (Figs. 8c,d and Table 3). Peak mean values are up to 2.5 m s−1 higher than the reanalysis, particularly over the southern parts of the Great Australian Bight and over southeastern Australia. Very good agreement is found for both DJF and MA, with small differences where the minimum S06 contour is farther north than for ERA-Interim, particularly in the area that intersects the peak mean CAPE. Contrasting the poor agreement in SON, excellent agreement is found in spatial variance for DJF, and good agreement is found during MA.

Fig. 8.
Fig. 8.

As in Fig. 6, but for the comparison of mean seasonal S06 among ERA-Interim, CSIRO Mk3.6, and CCAM.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

The relative frequency of high S06 environments is underestimated over both Australia and the EAReg in CCAM, despite high pattern congruence throughout the season (Figs. 8e,f). CCAM has 50% fewer environments with S06 greater than 20 m s−1 over the EAReg, with a slightly higher proportion over the full continent. This low frequency of high S06 values may be related to mixing processes within model downscaling. While pattern correlations over the continent indicate very good to excellent agreement for mean S06, agreements over the EAReg are poor for both SON and MA, reflecting the 5 m s−1 lower values of peak mean shear over the continent. This also corresponds to a southward displacement of the field compared to the ERA-Interim. In DJF, CCAM has excellent agreement, with the equatorial bias in westerly flow compared to the reanalysis assisting in producing higher S06 over the domain. The minimum and 7.5 m s−1 S06 contours also extend farther north along the west coast. During both SON and DJF, there is also reduced shear (2.5 m s−1) over the central parts of Australia compared to ERA-Interim. The mean field remains southwardly displaced through the MA period. This southward displacement explains the noted differences in the distribution of S06 environments in nonzero CAPE environments over both regions.

As for the comparison of mean CAPE, it is difficult to be certain that the mean S06 of the ERA-Interim represents the actual occurrence of S06 over Australia given the potential for variations in boundary layer winds attributable to local effects and complex topography (Allen and Karoly 2014). Despite the differences being smaller by relative magnitude as compared to CAPE, mean S06 from CSIRO Mk3.6 is higher than the reanalysis, particularly over northwestern Australia. In contrast, mean S06 from CCAM is biased low despite excellent agreement with the ERA-Interim for the continent, particularly during DJF.

5. Warm-season convective environments over the east Australian region

Evaluating climate model capability to simulate the frequency and distribution of severe convective environments is complicated by reanalysis data potentially not being representative of the observed climatology. To address this, model environment pattern and magnitude are compared to the known climatology of severe thunderstorm reports over the EAReg (Fig. 9). However, the mean seasonal occurrence of reports for the EAReg has relatively low frequency because of population density and gaps in the frequency as compared to SEV environments (Allen and Karoly 2014). Despite these limitations, the ability of the models to simulate relative frequency peaks and the spatial pattern of the mean climatology of reports is a useful test to examine the characteristics of the model representations. In each of the climatologies, a relatively thin band of high frequency extends along the Great Dividing Range, particularly for the reports and environments from ERA-Interim and CCAM. The position of the maximum frequency is also similar for both the ERA-Interim and CCAM with higher pattern congruence, despite poor agreement in spatial variance reflecting the slight southward displacement in CCAM. Pattern congruence for the CSIRO Mk3.6 is also higher over the EAReg despite the higher number of SEV environments along the east coast of Australia for the comparable period (78 per season compared to 26 for ERA-Interim). However, this is related to the distribution having a similar spatial pattern (good agreement), particularly along the Great Dividing Range, rather than the mean peak magnitudes of the datasets being similar (Fig. 9c). In contrast, CCAM has a similar peak frequency to that of the ERA-Interim, displaced southward over the central east coast (Fig. 9d).

Fig. 9.
Fig. 9.

Comparison of the mean frequency of SEV environments or reports as determined using the (a) mean seasonal reports, based on the climatology spanning September 2003–April 2010; (b) ERA-Interim; (c) CSIRO Mk3.6; and (d) CCAM. Reports are gridded to the reanalysis grid and contoured as for the models using mean seasonal values. Frequency is based on the number of days per year where CAPE × S061.67 exceeds the severe discriminant [Eq. (1)] for pseudosoundings at each grid point of the respective datasets for the EAReg. Favorable environmental conditions are described in the text. Contours are at 2 environment intervals from 0 to 10, 4 environment intervals from 10 to 30, and 8 environment intervals from 30 to 78.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

In each of the model representations, it is encouraging to note spatial similarities to the reports in terms of relative frequency even though the magnitudes are not directly comparable. Higher frequencies in each climatology stretch along the Great Dividing Range of the east coast, with frequencies gradually decreasing as they extend southward. Despite this, pattern congruence to the reports climatology is low because of their relatively sparse distribution in comparison to environments (Table 2). While the highest pattern congruence to reports is identified for the CSIRO Mk3.6, it is difficult to identify the reason for this. An obvious trait of the CSIRO Mk3.6 data is poor handling of the environment distribution as compared to both ERA-Interim and reports. Two of the features that contribute are the positive bias in the CAPE distribution (Fig. 5) and unimpeded advection processes over the east coast because of the terrain smoothing (Figs. 1, 7).

Table 2.

Pattern congruence for land grid points over the EAReg using the 7-yr report mean as compared to each of ERA-Interim, CSIRO Mk3.6, and CCAM for the warm-season climatological period. For comparison, pattern congruence to the reanalysis for both Australia and the EAReg are also calculated. Reports are binned to the ERA-Interim grid and then pattern congruence and centered pattern correlations for the respective models are calculated by regridding (see text). Italics denote 75% of the spatial variance explained in the model as compared to reanalysis referred to as very good.

Table 2.

6. Evaluation of seasonal cycle and environments

a. Seasonal cycle of environments

To evaluate the mean seasonal cycle of environments, the monthly frequency of environments in the respective models is used to produce a normalized distribution of the mean seasonal cycle over the EAReg. This normalization was based on the number of grid points over the EAReg for the respective model outputs. The reports climatology and ERA-Interim (Fig. 10a) display a frequency distribution that rises from September to a peak in December before falling in subsequent months. The peak frequency is underestimated by the normalized SEV environment reanalysis climatology. While mean values are lower over the EAReg for the reanalysis, this is most likely related to the lack of a one-to-one relationship between a report and a favorable environment in the reanalysis (the nonconditionality where a favorable environment is present but does not initiate or produce severe weather). Furthermore, when a SEV environment (or environments) occurs within the coarse grid scale, it may result in more than one reportable thunderstorm event.

Fig. 10.
Fig. 10.

Mean seasonal cycle of the frequency of SEV environments over the EAReg as determined using (a) ERA-Interim, (b) CSIRO Mk3.6, and (c) CCAM. Reports are shown on the respective frequency distributions in gray, while the environmental mean, maximum, and minimum over the 20-yr period are shown by the black box and whiskers for the respective months. Frequencies are normalized by one-tenth of the total grid points over the EAReg, except in the case of the reports, which are climatological values. Note that the y-axis scale for CSIRO Mk3.6 was modified to encompass the frequencies of that model.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

Frequencies over the EAReg from CSIRO Mk3.6 are considerably positively biased relative to the ERA-Interim (Fig. 10b), with the closest agreement during the early part of the season (September–October). This, however, is followed by an apparent lack of response to the seasonal cycle, with significant overestimation of the mean and variance from November to April, reflecting the spatial overestimation (Fig. 11). In contrast, CCAM produces a seasonal cycle with a similar peak to the reanalysis and reports over the EAReg (Fig. 10c). Mean monthly occurrences of environments are lower than reports, suggesting that CCAM is capable of producing a more realistic seasonal cycle comparable to ERA-Interim.

Fig. 11.
Fig. 11.

Comparison among ERA-Interim, CSIRO Mk3.6, and CCAM of mean seasonal frequency of SEV environments over the 20 convective warm seasons of 1980–2000 per grid box for the respective model grids. Seasons correspond to (a),(c),(e) SON and (b),(d),(f) DJF. Contours are at 1 environment interval from 0 to 2, 2 environment intervals from 2 to 20, and 4 environment intervals from 20 to >36.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

b. Spatial distribution of severe thunderstorm environments

The ERA-Interim climatology for the late twentieth century shows a small number of SEV environments over the southeastern parts of the continent during September and October (Fig. 11a). The region of higher-frequency shifts inland and southward between November and January before retreating northward, with decreasing frequency between February and April. CSIRO Mk3.6 produces a larger number of SEV environments throughout the season, particularly November–April. Pattern congruence is high with very good spatial variance during DJF, despite large differences in magnitude from ERA-Interim (Fig. 11c and Tables 3, 4). However, during both SON and MA, pattern congruence decreases associated with a displaced maximum with substantially differing magnitudes that double those of the reanalysis. Two contributions to this difference are the greater moisture over the continent (Fig. 3b) combined with steeper lapse rates that extend over the east coast (Figs. 7c,d). Despite this, pattern congruence over the EAReg remains substantially higher than the values for the entire landmass.

Table 3.

Pattern correlations (centered) over both Australia and the EAReg for the mean season climatologies of SEV; conditional high MLCAPE and high S06 environments; and mean MLCAPE, S06, and 700–500-hPa lapse rates (LAPSE) from CSIRO Mk3.6 and CCAM as compared to ERA-Interim. Values where the model climatology explains greater than 75% of the spatial variance (very good) are italicized, while boldfaced values indicate 90% of the spatial variance being explained (excellent).

Table 3.
Table 4.

Pattern congruence for land grid points over both Australia and the EAReg for the mean season climatologies of SEV, conditional high MLCAPE, and high S06 environments from CSIRO Mk3.6 and CCAM as compared to the ERA-Interim.

Table 4.

As would be expected from the seasonal cycle, the CCAM distribution for bimonthly periods is closer to ERA-Interim with good agreement throughout (Figs. 11e,f). The northern and inland SEV environment distribution during SON is spatially limited compared with the reanalysis, resulting in lower pattern congruence. There is also poor agreement between the reanalysis and CCAM in MA, particularly over the EAReg (not shown), which is the result of a low frequency of environments. Despite these limitations, the positioning of the maximum frequency over the east coast is close to both the reanalysis and reports throughout the season reflected by the pattern congruence for the EAReg (Table 3). Maximum frequencies over the east coast are also similar to those of ERA-Interim, peaking between 8 and 10 SEV environments per season. The variations in the spatial distribution and magnitude relate to the influence of the grid scale on coastal moisture distributions (Figs. 3a,c) and lessened eastward extent of steep lapse rates (Figs. 7e,f).

Over the northwest of Australia, the peak of the seasonal cycle coincides with the development of the northern monsoon in December and January. This may explain the tendency of both the models and reanalysis to favor environment occurrence and high CAPE for this region during the summer months. Typically, this area is associated with relatively weak wind shear (Fig. 8) and strong convective inhibition that, despite appreciable moisture and steep lapse rates, rarely yields observed severe thunderstorms (Figs. 3, 7; Allen and Karoly 2014). The weaknesses in model vertical resolution and convective parameterization scheme also contribute to the poor simulation of convective inhibition, placing SEV environments in this area further into question.

c. Ingredient conditional environments

While mean climatologies of ingredients are useful to ascertain a model’s ability to represent the mean convective state, the conditional coincidence of ingredient occurrence (i.e., which ingredient contributes to the environment depending on location) is helpful in evaluating the model performance in simulating convective environments. Consideration was made of the mean distribution of CAPE environments exceeding 1000 J kg−1 in the presence of at least 5 m s−1 for S06 and the distribution of S06 environments exceeding 15 m s−1 in the presence of at least 10 J kg−1 (and thereby nonzero product of the two quantities). These values were chosen to reflect cases where high values of the chosen parameter were coincident with suppressed dependence on the covariate.

CSIRO Mk3.6 produces a larger number of CAPE environments with good agreement in spatial variance and pattern congruence for SON (Fig. 12c), though low-frequency values for these conditional environments extend farther southeast. Pattern congruences are high with good agreement in variance throughout the warm season; however, during DJF an excessively large area of conditional CAPE environments exceeding 10 per season occurs, particularly over the northwest (Fig. 12d). In contrast, CCAM tends to produce fewer conditional high CAPE environments (Figs. 12e,f) over a much smaller spatial scale than ERA-Interim, particularly during SON. The pattern over northern Australia has good agreement during DJF, producing high pattern congruence. This is despite the model having a lesser inland extent because of the moisture distribution and lower frequency of high CAPE over the east coast with less steep lapse rates.

Fig. 12.
Fig. 12.

As in Fig. 11, but for comparison among ERA-Interim, CSIRO Mk3.6, and CCAM climatologies of mean seasonal frequency of high MLCAPE environments exceeding 1000 J kg−1 conditional on the presence of S06 exceeding 5 m s−1.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

Conditional S06 environments occur preferentially along the east coast, with the greatest frequency during SON (Figs. 13a,b). CSIRO Mk3.6 overestimates the frequency of these environments, extending them farther north and west during SON and DJF with good to very good agreement in spatial variance as a response to positive biases in both the S06 and CAPE distributions. While CCAM remains closer to ERA-Interim for pattern and frequency (Figs. 13e,f), the spatial area tends to be more limited to areas coastward of the dividing range as a result of negative biases in CAPE and S06.

Fig. 13.
Fig. 13.

As in Fig. 11, but for comparison among ERA-Interim, CSIRO Mk3.6, and CCAM climatologies of mean seasonal frequency of high S06 environments exceeding 15 m s−1 conditional on the presence of MLCAPE exceeding 10 J kg−1.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00425.1

7. Discussion and conclusions

The evaluation of CSIRO Mk3.6 and CCAM has revealed for the period 1980–2000 that spatially the simulated distribution of environments favorable to the development of severe convection over Australia is comparable to ERA-Interim. However, while the models are of differing resolutions, a number of limitations and biases within the respective climatologies have been identified. CSIRO Mk3.6 overestimates the frequency of SEV environments by a significant margin as a response to positive biases in both CAPE and S06 along with steeper lapse rates and higher continental moisture. CCAM produces a distribution that is closer to ERA-Interim but is influenced by reduced inland moisture and negative biases in both CAPE and S06. While the peak values of SEV environments in the northwest of the continent are identified in both the models and reanalysis, these environments may not be realistic in any of the datasets because of the high climatological CAPE and inflated values of S06 occurring over the region.

CSIRO Mk3.6 simulates an overly moist inland area during the late twentieth century. The high frequency of SEV and CAPE environments over the northern parts of the continent also appears to coincide with the presence of excessive tropical rainfall in the model relative to observations (Rotstayn et al. 2010). This suggests that the convective parameterization scheme and subgrid-scale processes contribute to the model’s traits and biases and the overly moist simulation, along with overly steep continental lapse rates produced by the model is similar to the problems noted with CSIRO Mk3.5 (Fig. 3; Willett et al. 2010). This when combined with the higher SSTs north of the continent and resultant increases to moisture result in a positive bias in CAPE and hence the frequency of high CAPE environments. This problem is further aggravated by poor rendering of convective inhibition (while somewhat expected), resulting in an artificially high number of potential “environments” by circumventing the condition of CIN below 25 J kg−1. Small positive biases also exist in the distribution of the S06 parameter, inflating the frequency of SEV environments. Despite this difference, there is very good agreement between the CSIRO Mk3.6 and ERA-Interim for S06, with the mean latitude of midlatitude westerlies in both products being similar (Grose et al. 2013). The larger grid scale and topographic smoothing also result in an unrealistic representation of the complex topography of the east coast, producing a distribution that does not reflect environments in this area. While the climatology of environments is high for CSIRO Mk3.6, we also emphasize that ERA-Interim underestimates CAPE magnitude relative to observations (Allen and Karoly 2014) and that the actual population of SEV environments likely lies between the two climatologies rather than either being categorically incorrect.

In contrast to CSIRO Mk3.6, CCAM has a relatively dry continental interior in its simulation, artificially biasing the climatology toward coastal areas. This is most likely related to the moisture field and overly strong continental high pressure systems within the model (Nguyen et al. 2012). Both CAPE and S06 within the model are negatively biased, leading to fewer model-derived SEV environments, which is closer to the expectations of climate models from earlier studies (Niall and Walsh 2005; Marsh et al. 2007). While the distribution of SEV environments has a similar magnitude to ERA-Interim and has good agreement in spatial variance over the EAReg, it is also southwardly displaced within the model, following the extent of coastal moisture and the spatial distribution of high S06. This is somewhat surprising given that CCAM in this configuration has a northward bias of 3° in the mean latitude of the westerly belt compared to ERA-Interim (Grose et al. 2013). However, the model’s handling of the complex topography over the Great Dividing Range does appear to contribute to how moisture is distributed over the continent and the corresponding distribution of ingredients. It is also plausible that the influence of spatial resolution on the environmental climatology may mean that a larger number of environments would be identified where instability and inhibition occur over a smaller scale than the ERA-Interim grid or that these environments may correspond to noise.

This evaluation raises questions as to our confidence in all climate models for simulating convective environments, particularly those of a coarse horizontal resolution. While differences in parameterizations and model dynamics contribute to different simulations from CSIRO Mk3.6 compared to CCAM, the role of increased spatial resolution also appears important. This is particularly the case for areas where local topography influences surface flow such as the Great Dividing Range along the east coast, along land–ocean boundaries, and in handling dynamics that result from resolution increases (e.g., Niall and Walsh 2005; Randall et al. 2007; Grose et al. 2013; Smith et al. 2013). While this cannot be fully attributed based on the results here, further examination of the influence of spatial resolution on topography for convective ingredients over both Australia and the United States is needed, particularly with respect to the influence on midlevel lapse rates. Issues with vertical resolution are also a significant limitation to both models (e.g., Marsh et al. 2007; Diffenbaugh et al. 2013), and this influence on resolving convective inhibition will continue to limit the applicability of statistical downscaling in climate simulations from global climate models. These limitations would suggest that the way forward involves dynamical downscaling (e.g., Trapp et al. 2007b; Robinson et al. 2013), the use of higher spatially resolving models for ingredients-based climatologies, or ensemble-based approaches that are limited by the weakest member (Diffenbaugh et al. 2013; Part II). At the very least, detailed evaluation of climate models using the techniques presented here and perhaps multiple reanalyses is necessary to identify biases for any model before application to analysis of convective environments and ingredients in the future.

An overarching question of this evaluation has been how good a model needs to be in representing the twentieth-century climatology as represented by a single reanalysis. The mean spatial ingredients over Australia simulated by both climate models have either very good (>75%) or better agreement in spatial variance. Consequently, both models also appeared capable of representing at least 50% of spatial variance in convective environments, though CCAM appeared to be slightly superior in terms of pattern and magnitude of SEV environments when compared to both ERA-Interim and reports data, albeit with a negative bias in frequency. Mean seasonal and diurnal cycles presented a major problem for CSIRO Mk3.6, with the model struggling to identify an annual or diurnal cycle for much of the continent. However, CSIRO Mk3.6 does remain relatively consistent in its overestimation with generally good agreement for spatial variance, suggesting that any response to a warming climate in future simulations is likely to follow the correct direction as compared to the historical climatology. Given the apparent limitations to both models, before either can be used to examine changes to convective climates under a warming climate, the relative changes to the atmospheric variables that contribute to the ingredients and parameters (i.e., temperature, moisture, and wind fields) must first also be examined. This can then be used to establish whether any changes are less to do with model bias, and more to do with a real change in convective conditions over the continent. The implications of this warming for Australia are discussed in Part II.

Acknowledgments

We thank the ECMWF for providing the reanalysis data used in this study. We are also grateful to CSIRO Marine and Atmospheric Research, including the CSIRO Mk3.6 modelling group and J. Katzfey for providing access to the CCAM data. We thank the three anonymous reviewers, whose contributions assisted in improving the manuscript. This research was supported in part by funding from the Australian Research Council Centre of Excellence for Climate System Science (Grant CE110001028) and the NCI National Facility at the ANU. The Office of Naval Research (Grant N00014-12-1-0911) and a Columbia University Research Initiatives for Science and Engineering (RISE) award also supported the writing of this paper.

REFERENCES

  • Allen, J. T., 2012: Supercell storms: Melbourne’s white Christmas 2011. Bull. Aust. Meteor. Oceanogr. Soc., 25, 4751.

  • Allen, J. T., and D. J. Karoly, 2014: A climatology of Australian severe thunderstorm environments 1979–2011: Inter-annual variability and ENSO influence. Int. J. Climatol., 34, 8197, doi:10.1002/joc.3667.

    • Search Google Scholar
    • Export Citation
  • Allen, J. T., D. J. Karoly, and G. A. Mills, 2011: A severe thunderstorm climatology for Australia and associated thunderstorm environments. Aust. Meteor. Oceanogr. J., 61, 143158.

    • Search Google Scholar
    • Export Citation
  • Allen, J. T., D. J. Karoly, and K. J. Walsh, 2014: Future Australian severe thunderstorm environments. Part II: The influence of a strongly warming climate on convective environments. J. Climate, 27, 38483868, doi:10.1175/JCLI-D-13-00426.1.

    • Search Google Scholar
    • Export Citation
  • Balling, R. C., Jr., and R. S. Cerveny, 2003: Compilation and discussion of trends in severe storms in the United States: Popular perception vs. climate reality. Nat. Hazards, 29, 103112, doi:10.1023/A:1023674722282.

    • Search Google Scholar
    • Export Citation
  • Bennett, J. C., and Coauthors, 2012: High-resolution projections of surface water availability for Tasmania, Australia. Hydrol. Earth Syst. Sci., 16, 12871303, doi:10.5194/hess-16-1287-2012.

    • Search Google Scholar
    • Export Citation
  • Bouwer, L. M., 2011: Have disaster losses increased due to anthropogenic climate change? Bull. Amer. Meteor. Soc., 92, 3946, doi:10.1175/2010BAMS3092.1.

    • Search Google Scholar
    • Export Citation
  • Brooks, H. E., 2009: Proximity soundings for severe convection for Europe and the United States from reanalysis data. Atmos. Res., 93, 546553, doi:10.1016/j.atmosres.2008.10.005.

    • Search Google Scholar
    • Export Citation
  • Brooks, H. E., 2013: Severe thunderstorms and climate change. Atmos. Res., 123, 129138, doi:10.1016/j.atmosres.2012.04.002.

  • Brooks, H. E., J. W. Lee, and J. P. Craven, 2003: The spatial distribution of severe thunderstorm and tornado environments from global reanalysis data. Atmos. Res., 67–68, 7394, doi:10.1016/S0169-8095(03)00045-0.

    • Search Google Scholar
    • Export Citation
  • Cai, W. J., P. H. Whetton, and D. J. Karoly, 2003: The response of the Antarctic Oscillation to increasing and stabilized atmospheric CO2. J. Climate, 16, 15251538, doi:10.1175/1520-0442-16.10.1525.

    • Search Google Scholar
    • Export Citation
  • Changnon, S. A., 2003: Shifting economic impacts from weather extremes in the United States: A result of societal changes, not global warming. Nat. Hazards, 29, 273290, doi:10.1023/A:1023642131794.

    • Search Google Scholar
    • Export Citation
  • Corney, S. P., and Coauthors, 2010: Climate futures for Tasmania: Methods and results on climate modelling. Antarctic Climate and Ecosystems Cooperative Research Centre Rep., 88 pp.

  • Corney, S. P., M. Grose, J. C. Bennett, C. White, J. Katzfey, J. McGregor, G. Holz, and N. L. Bindoff, 2013: Performance of downscaled regional climate simulations using a variable-resolution regional climate model: Tasmania as a test case. J. Geophys. Res., 118, 11 93611 950, doi:10.1002/2013JD020087.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Diffenbaugh, N. S., M. Scherer, and R. J. Trapp, 2013: Robust increases in severe thunderstorm environments in response to greenhouse forcing. Proc. Nat. Acad. Sci.,100, 16 361–16 366, doi:10.1073/pnas.1307758110.

  • Doswell, C. A., III, and E. N. Rasmussen, 1994: The effect of neglecting the virtual temperature correction on CAPE calculations. Wea. Forecasting, 9, 625629, doi:10.1175/1520-0434(1994)009<0625:TEONTV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, H. E. Brooks, and M. P. Kay, 2005: Climatological estimates of daily local nontornadic severe thunderstorm probability for the United States. Wea. Forecasting, 20, 577595, doi:10.1175/WAF866.1.

    • Search Google Scholar
    • Export Citation
  • Gettelman, A., D. Seidel, M. Wheeler, and R. Ross, 2002: Multi-decadal trends in tropical convective available potential energy. J. Geophys. Res., 107, 46064614, doi:10.1029/2001JD001082.

    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and Coauthors, 2002: The CSIRO Mk3 climate system model. CSIRO Atmospheric Research Tech. Rep., 130 pp.

  • Gordon, H. B., S. P. O’Farrell, M. A. Collier, M. R. Dix, L. D. Rotstayn, E. A. Kowalczyk, A. C. Hirst, and I. G. Watterson, 2010: The CSIRO Mk3.5 climate model. Centre for Australia Weather and Climate Research Tech. Rep. 21, 62 pp.

  • Gregory, D., and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability-dependent closure. Mon. Wea. Rev., 118, 14831506, doi:10.1175/1520-0493(1990)118<1483:AMFCSW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Griffiths, D., J. Colquhoun, K. Batt, and T. Casinader, 1993: Severe thunderstorms in New South Wales: Climatology and means of assessing the impact of climate change. Climatic Change, 25, 369388, doi:10.1007/BF01098382.

    • Search Google Scholar
    • Export Citation
  • Grose, M. R., S. P. Corney, J. J. Katzfey, J. C. Bennett, G. K. Holz, C. J. White, and N. L. Bindoff, 2013: A regional response in mean westerly circulation and rainfall to projected climate warming over Tasmania, Australia. Climate Dyn., 40, 20352048, doi:10.1007/s00382-012-1405-1.

    • Search Google Scholar
    • Export Citation
  • Iorio, J. P., P. B. Duffy, B. Govindasamy, S. L. Thompson, M. Khairoutdinov, and D. Randall, 2004: Effects of model resolution and sub-grid scale physics on the simulation of precipitation in the continental United States. Climate Dyn., 23, 243258, doi:10.1007/s00382-004-0440-y.

    • Search Google Scholar
    • Export Citation
  • Jeffrey, S. J., L. D. Rotstayn, M. A. Collier, S. M. Dravistzki, C. Hamalainen, K. K. Wong, and J. J. Syktus, 2013: Australia’s CMIP5 submission using the CSIRO Mk3.6 model. Aust. Meteor. Oceanogr. J., 63, 113.

    • Search Google Scholar
    • Export Citation
  • Katzfey, J. J., J. L. McGregor, K. C. Nguyen, and M. Thatcher, 2009: Dynamical downscaling techniques: Impacts on regional climate change signals. Proc. 18th World IMACS Congress and MODSIM09 Int. Congress on Modelling and Simulation, Cairns, Australia, Modelling and Simulation Society of Australia and New Zealand and International Association for Mathematics and Computers in Simulation, 2377–2383. [Available online at http://www.mssanz.org.au/modsim09/I13/katzfey_I13.pdf.]

  • Kuleshov, Y., G. D. Hoedt, W. Wright, and A. Brewster, 2002: Thunderstorm distribution and frequency in Australia. Aust. Meteor. Mag., 51, 145154.

    • Search Google Scholar
    • Export Citation
  • Marsh, P. T., H. E. Brooks, and D. J. Karoly, 2007: Assessment of the severe weather environment in North America simulated by a global climate model. Atmos. Sci. Lett., 8, 100106, doi:10.1002/asl.159.

    • Search Google Scholar
    • Export Citation
  • Marsh, P. T., H. E. Brooks, and D. J. Karoly, 2009: Preliminary investigation into the severe thunderstorm environment of Europe simulated by the Community Climate Systems Model 3. Atmos. Res., 93, 607618, doi:10.1016/j.atmosres.2008.09.014.

    • Search Google Scholar
    • Export Citation
  • McGregor, J. L., 2003: A new convection scheme using a simple closure. Extended Abstracts, 15th Annual BMRC Modelling Workshop, Melbourne, Australia, Bureau of Meteorology Research Centre, 3336.

    • Search Google Scholar
    • Export Citation
  • McGregor, J. L., 2005: CCAM: Geometic aspects and dynamical formulation. CSIRO Atmospheric Research Tech. Rep. 70, 43 pp.

  • McGregor, J. L., 2006: Regional climate modelling using CCAM. BMRC Research Rep.123, 118122.

  • McGregor, J. L., and M. R. Dix, 2008: An updated description of the Conformal-Cubic Atmospheric Model. High Resolution Simulation of the Atmosphere and Ocean, K. Hamilton and W. Ohfuchi, Eds., Springer, 51–76.

    • Search Google Scholar
    • Export Citation
  • Nguyen, K. C., J. J. Katzfey, and J. L. McGregor, 2012: Global 60 km simulations with CCAM: Evaluation over the tropics. Climate Dyn., 39, 637654, doi:10.1007/s00382-011-1197-8.

    • Search Google Scholar
    • Export Citation
  • Niall, S., and K. Walsh, 2005: The impact of climate change on hailstorms in southeastern Australia. Int. J. Climatol., 25, 19331952, doi:10.1002/joc.1233.

    • Search Google Scholar
    • Export Citation
  • Nicholls, N., 2008: Australian climate and weather extremes: Past, present and future. Commonwealth of Australia Department of Climate Change Rep., 32 pp.

  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 591–662.

  • Richman, M. B., 1986: Rotation of principal components. J. Climatol., 6, 293335, doi:10.1002/joc.3370060305.

  • Robinson, E. D., R. J. Trapp, and M. E. Baldwin, 2013: The geospatial and temporal distributions of severe thunderstorms from high-resolution dynamical downscaling. J. Appl. Meteor. Climatol.,52, 2147–2161, doi:10.1175/JAMC-D-12-0131.1.

  • Rotstayn, L. D., M. A. Collier, M. R. Dix, Y. Feng, H. B. Gordon, S. P. O’ Farrell, I. N. Smith, and J. Syktus, 2010: Improved simulation of Australian climate and ENSO-related rainfall variability in a global climate model with an interactive aerosol treatment. Int. J. Climatol., 30, 10671088, doi:10.1002/joc.1952.

    • Search Google Scholar
    • Export Citation
  • Schuster, S. S., R. J. Blong, and M. S. Speer, 2005: A hail climatology of the greater Sydney area and New South Wales. Int. J. Climatol., 25, 16331650, doi:10.1002/joc.1199.

    • Search Google Scholar
    • Export Citation
  • Simmons, A. J., K. M. Willett, P. D. Jones, P. W. Thorne, and D. Dee, 2010: Low-frequency variations in surface atmospheric humidity, temperature and precipitation: Inferences from reanalyses and monthly gridded observational datasets. J. Geophys. Res., 115, D01110, doi:10.1029/2009JD012442.

    • Search Google Scholar
    • Export Citation
  • Smith, I., A. Moise, J. Katzfey, K. Nguyen, and R. Colman, 2013: Regional-scale rainfall projections: Simulations for the New Guinea region using the CCAM model. J. Geophys. Res. Atmos.,118, 1271–1280, doi:10.1002/jgrd.50139.

  • Trapp, R. J., N. S. Diffenbaugh, H. E. Brooks, M. E. Baldwin, E. D. Robinson, and J. S. Pal, 2007a: Changes in severe thunderstorm environment frequency during the 21st century caused by anthropogenically enhanced global radiative forcing. Proc. Natl. Acad. Sci. USA, 104, 19 719–19 723, doi:10.1073/pnas.0705494104.

    • Search Google Scholar
    • Export Citation
  • Trapp, R. J., B. A. Halvorson, and N. S. Diffenbaugh, 2007b: Telescoping multimodel approaches to evaluate extreme convective weather under future climates. J. Geophys. Res., 112, D20109, doi:10.1029/2006JD008345.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

  • Van Klooster, S. L., and P. J. Roebber, 2009: Surface-based convective potential in the contiguous United States in a business-as-usual future climate. J. Climate, 22, 33173330, doi:10.1175/2009JCLI2697.1.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences.2nd ed. International Geophysics Series, Vol. 59, Academic Press, 627 pp.

  • Willett, K. M., P. D. Phillip, P. W. Thorne, and N. P. Gillett, 2010: A comparison of large scale changes in surface humidity over land in observations and CMIP3 general circulation models. Environ. Res. Lett., 5, 025210, doi:10.1088/1748-9326/5/2/025210.

    • Search Google Scholar
    • Export Citation
Save
  • Allen, J. T., 2012: Supercell storms: Melbourne’s white Christmas 2011. Bull. Aust. Meteor. Oceanogr. Soc., 25, 4751.

  • Allen, J. T., and D. J. Karoly, 2014: A climatology of Australian severe thunderstorm environments 1979–2011: Inter-annual variability and ENSO influence. Int. J. Climatol., 34, 8197, doi:10.1002/joc.3667.

    • Search Google Scholar
    • Export Citation
  • Allen, J. T., D. J. Karoly, and G. A. Mills, 2011: A severe thunderstorm climatology for Australia and associated thunderstorm environments. Aust. Meteor. Oceanogr. J., 61, 143158.

    • Search Google Scholar
    • Export Citation
  • Allen, J. T., D. J. Karoly, and K. J. Walsh, 2014: Future Australian severe thunderstorm environments. Part II: The influence of a strongly warming climate on convective environments. J. Climate, 27, 38483868, doi:10.1175/JCLI-D-13-00426.1.

    • Search Google Scholar
    • Export Citation
  • Balling, R. C., Jr., and R. S. Cerveny, 2003: Compilation and discussion of trends in severe storms in the United States: Popular perception vs. climate reality. Nat. Hazards, 29, 103112, doi:10.1023/A:1023674722282.

    • Search Google Scholar
    • Export Citation
  • Bennett, J. C., and Coauthors, 2012: High-resolution projections of surface water availability for Tasmania, Australia. Hydrol. Earth Syst. Sci., 16, 12871303, doi:10.5194/hess-16-1287-2012.

    • Search Google Scholar
    • Export Citation
  • Bouwer, L. M., 2011: Have disaster losses increased due to anthropogenic climate change? Bull. Amer. Meteor. Soc., 92, 3946, doi:10.1175/2010BAMS3092.1.

    • Search Google Scholar
    • Export Citation
  • Brooks, H. E., 2009: Proximity soundings for severe convection for Europe and the United States from reanalysis data. Atmos. Res., 93, 546553, doi:10.1016/j.atmosres.2008.10.005.

    • Search Google Scholar
    • Export Citation
  • Brooks, H. E., 2013: Severe thunderstorms and climate change. Atmos. Res., 123, 129138, doi:10.1016/j.atmosres.2012.04.002.

  • Brooks, H. E., J. W. Lee, and J. P. Craven, 2003: The spatial distribution of severe thunderstorm and tornado environments from global reanalysis data. Atmos. Res., 67–68, 7394, doi:10.1016/S0169-8095(03)00045-0.

    • Search Google Scholar
    • Export Citation
  • Cai, W. J., P. H. Whetton, and D. J. Karoly, 2003: The response of the Antarctic Oscillation to increasing and stabilized atmospheric CO2. J. Climate, 16, 15251538, doi:10.1175/1520-0442-16.10.1525.

    • Search Google Scholar
    • Export Citation
  • Changnon, S. A., 2003: Shifting economic impacts from weather extremes in the United States: A result of societal changes, not global warming. Nat. Hazards, 29, 273290, doi:10.1023/A:1023642131794.

    • Search Google Scholar
    • Export Citation
  • Corney, S. P., and Coauthors, 2010: Climate futures for Tasmania: Methods and results on climate modelling. Antarctic Climate and Ecosystems Cooperative Research Centre Rep., 88 pp.

  • Corney, S. P., M. Grose, J. C. Bennett, C. White, J. Katzfey, J. McGregor, G. Holz, and N. L. Bindoff, 2013: Performance of downscaled regional climate simulations using a variable-resolution regional climate model: Tasmania as a test case. J. Geophys. Res., 118, 11 93611 950, doi:10.1002/2013JD020087.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Diffenbaugh, N. S., M. Scherer, and R. J. Trapp, 2013: Robust increases in severe thunderstorm environments in response to greenhouse forcing. Proc. Nat. Acad. Sci.,100, 16 361–16 366, doi:10.1073/pnas.1307758110.

  • Doswell, C. A., III, and E. N. Rasmussen, 1994: The effect of neglecting the virtual temperature correction on CAPE calculations. Wea. Forecasting, 9, 625629, doi:10.1175/1520-0434(1994)009<0625:TEONTV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Doswell, C. A., III, H. E. Brooks, and M. P. Kay, 2005: Climatological estimates of daily local nontornadic severe thunderstorm probability for the United States. Wea. Forecasting, 20, 577595, doi:10.1175/WAF866.1.

    • Search Google Scholar
    • Export Citation
  • Gettelman, A., D. Seidel, M. Wheeler, and R. Ross, 2002: Multi-decadal trends in tropical convective available potential energy. J. Geophys. Res., 107, 46064614, doi:10.1029/2001JD001082.

    • Search Google Scholar
    • Export Citation
  • Gordon, H. B., and Coauthors, 2002: The CSIRO Mk3 climate system model. CSIRO Atmospheric Research Tech. Rep., 130 pp.

  • Gordon, H. B., S. P. O’Farrell, M. A. Collier, M. R. Dix, L. D. Rotstayn, E. A. Kowalczyk, A. C. Hirst, and I. G. Watterson, 2010: The CSIRO Mk3.5 climate model. Centre for Australia Weather and Climate Research Tech. Rep. 21, 62 pp.

  • Gregory, D., and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability-dependent closure. Mon. Wea. Rev., 118, 14831506, doi:10.1175/1520-0493(1990)118<1483:AMFCSW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Griffiths, D., J. Colquhoun, K. Batt, and T. Casinader, 1993: Severe thunderstorms in New South Wales: Climatology and means of assessing the impact of climate change. Climatic Change, 25, 369388, doi:10.1007/BF01098382.

    • Search Google Scholar
    • Export Citation
  • Grose, M. R., S. P. Corney, J. J. Katzfey, J. C. Bennett, G. K. Holz, C. J. White, and N. L. Bindoff, 2013: A regional response in mean westerly circulation and rainfall to projected climate warming over Tasmania, Australia. Climate Dyn., 40, 20352048, doi:10.1007/s00382-012-1405-1.

    • Search Google Scholar
    • Export Citation
  • Iorio, J. P., P. B. Duffy, B. Govindasamy, S. L. Thompson, M. Khairoutdinov, and D. Randall, 2004: Effects of model resolution and sub-grid scale physics on the simulation of precipitation in the continental United States. Climate Dyn., 23, 243258, doi:10.1007/s00382-004-0440-y.

    • Search Google Scholar
    • Export Citation
  • Jeffrey, S. J., L. D. Rotstayn, M. A. Collier, S. M. Dravistzki, C. Hamalainen, K. K. Wong, and J. J. Syktus, 2013: Australia’s CMIP5 submission using the CSIRO Mk3.6 model. Aust. Meteor. Oceanogr. J., 63, 113.

    • Search Google Scholar
    • Export Citation
  • Katzfey, J. J., J. L. McGregor, K. C. Nguyen, and M. Thatcher, 2009: Dynamical downscaling techniques: Impacts on regional climate change signals. Proc. 18th World IMACS Congress and MODSIM09 Int. Congress on Modelling and Simulation, Cairns, Australia, Modelling and Simulation Society of Australia and New Zealand and International Association for Mathematics and Computers in Simulation, 2377–2383. [Available online at http://www.mssanz.org.au/modsim09/I13/katzfey_I13.pdf.]

  • Kuleshov, Y., G. D. Hoedt, W. Wright, and A. Brewster, 2002: Thunderstorm distribution and frequency in Australia. Aust. Meteor. Mag., 51, 145154.

    • Search Google Scholar
    • Export Citation
  • Marsh, P. T., H. E. Brooks, and D. J. Karoly, 2007: Assessment of the severe weather environment in North America simulated by a global climate model. Atmos. Sci. Lett., 8, 100106, doi:10.1002/asl.159.

    • Search Google Scholar
    • Export Citation
  • Marsh, P. T., H. E. Brooks, and D. J. Karoly, 2009: Preliminary investigation into the severe thunderstorm environment of Europe simulated by the Community Climate Systems Model 3. Atmos. Res., 93, 607618, doi:10.1016/j.atmosres.2008.09.014.

    • Search Google Scholar
    • Export Citation
  • McGregor, J. L., 2003: A new convection scheme using a simple closure. Extended Abstracts, 15th Annual BMRC Modelling Workshop, Melbourne, Australia, Bureau of Meteorology Research Centre, 3336.

    • Search Google Scholar
    • Export Citation
  • McGregor, J. L., 2005: CCAM: Geometic aspects and dynamical formulation. CSIRO Atmospheric Research Tech. Rep. 70, 43 pp.

  • McGregor, J. L., 2006: Regional climate modelling using CCAM. BMRC Research Rep.123, 118122.

  • McGregor, J. L., and M. R. Dix, 2008: An updated description of the Conformal-Cubic Atmospheric Model. High Resolution Simulation of the Atmosphere and Ocean, K. Hamilton and W. Ohfuchi, Eds., Springer, 51–76.

    • Search Google Scholar
    • Export Citation
  • Nguyen, K. C., J. J. Katzfey, and J. L. McGregor, 2012: Global 60 km simulations with CCAM: Evaluation over the tropics. Climate Dyn., 39, 637654, doi:10.1007/s00382-011-1197-8.

    • Search Google Scholar
    • Export Citation
  • Niall, S., and K. Walsh, 2005: The impact of climate change on hailstorms in southeastern Australia. Int. J. Climatol., 25, 19331952, doi:10.1002/joc.1233.

    • Search Google Scholar
    • Export Citation
  • Nicholls, N., 2008: Australian climate and weather extremes: Past, present and future. Commonwealth of Australia Department of Climate Change Rep., 32 pp.

  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 591–662.

  • Richman, M. B., 1986: Rotation of principal components. J. Climatol., 6, 293335, doi:10.1002/joc.3370060305.

  • Robinson, E. D., R. J. Trapp, and M. E. Baldwin, 2013: The geospatial and temporal distributions of severe thunderstorms from high-resolution dynamical downscaling. J. Appl. Meteor. Climatol.,52, 2147–2161, doi:10.1175/JAMC-D-12-0131.1.

  • Rotstayn, L. D., M. A. Collier, M. R. Dix, Y. Feng, H. B. Gordon, S. P. O’ Farrell, I. N. Smith, and J. Syktus, 2010: Improved simulation of Australian climate and ENSO-related rainfall variability in a global climate model with an interactive aerosol treatment. Int. J. Climatol., 30, 10671088, doi:10.1002/joc.1952.

    • Search Google Scholar
    • Export Citation
  • Schuster, S. S., R. J. Blong, and M. S. Speer, 2005: A hail climatology of the greater Sydney area and New South Wales. Int. J. Climatol., 25, 16331650, doi:10.1002/joc.1199.

    • Search Google Scholar
    • Export Citation
  • Simmons, A. J., K. M. Willett, P. D. Jones, P. W. Thorne, and D. Dee, 2010: Low-frequency variations in surface atmospheric humidity, temperature and precipitation: Inferences from reanalyses and monthly gridded observational datasets. J. Geophys. Res., 115, D01110, doi:10.1029/2009JD012442.

    • Search Google Scholar
    • Export Citation
  • Smith, I., A. Moise, J. Katzfey, K. Nguyen, and R. Colman, 2013: Regional-scale rainfall projections: Simulations for the New Guinea region using the CCAM model. J. Geophys. Res. Atmos.,118, 1271–1280, doi:10.1002/jgrd.50139.

  • Trapp, R. J., N. S. Diffenbaugh, H. E. Brooks, M. E. Baldwin, E. D. Robinson, and J. S. Pal, 2007a: Changes in severe thunderstorm environment frequency during the 21st century caused by anthropogenically enhanced global radiative forcing. Proc. Natl. Acad. Sci. USA, 104, 19 719–19 723, doi:10.1073/pnas.0705494104.

    • Search Google Scholar
    • Export Citation
  • Trapp, R. J., B. A. Halvorson, and N. S. Diffenbaugh, 2007b: Telescoping multimodel approaches to evaluate extreme convective weather under future climates. J. Geophys. Res., 112, D20109, doi:10.1029/2006JD008345.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and Coauthors, 2007: Observations: Surface and atmospheric climate change. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 235–336.

  • Van Klooster, S. L., and P. J. Roebber, 2009: Surface-based convective potential in the contiguous United States in a business-as-usual future climate. J. Climate, 22, 33173330, doi:10.1175/2009JCLI2697.1.

    • Search Google Scholar
    • Export Citation
  • Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences.2nd ed. International Geophysics Series, Vol. 59, Academic Press, 627 pp.

  • Willett, K. M., P. D. Phillip, P. W. Thorne, and N. P. Gillett, 2010: A comparison of large scale changes in surface humidity over land in observations and CMIP3 general circulation models. Environ. Res. Lett., 5, 025210, doi:10.1088/1748-9326/5/2/025210.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    Distribution of grid points over the Australian continent after filtering using the land–sea mask for (a) ERA-Interim, (b) CSIRO Mk3.6. and (c) CCAM. The dashed box outlines the EAReg, bound by 39°–20°S and 144°–154.5°E. Topographic height contours in terms of height above mean sea level are shaded over the continent, with light blue indicating the extent of the land–sea mask used to filter the environment climatology for grid points occurring over land only.

  • Fig. 2.

    Temporal distribution of mean SEV environments per season for the respective models and reanalysis for all grid points over (a) Australia and (b) EAReg as compared to severe thunderstorm reports over the comparable region for the period March 2003–April 2010. Climatology period corresponds to 20 warm seasons (September–April 1980–2000) for the ERA-Interim, CSIRO Mk3.6, and CCAM. Counts of environments at the respective time steps are normalized by the number of model grid points over the respective regions divided by 10. Black line and dots represent the frequency of reports by nearest hour in local time and equivalent temporal distribution of the favorable SEV environments produced from the ERA-Interim (black circles), CSIRO Mk3.6 (crossed circles), and CCAM (dotted circles) at the respective analysis times. Times correspond to Australian eastern daylight time or equivalently 1800, 0000, 0600, and 1200 UTC, respectively.

  • Fig. 3.

    Seasonal mean surface mixing ratio at 0600 UTC for warm seasons (September–April) over the years 1980–2000 for the Australian region from (a) ERA-Interim, (b) CSIRO Mk3.6, and (c) CCAM. Contour interval is 1 g kg−1 between 6 and 12 g kg−1, 2 g kg−1 between 12 and 20 g kg−1, and 4 g kg−1 between 20 and 28 g kg−1.

  • Fig. 4.

    Spatial distribution of the mean warm-season difference between SEV environments over Australia as determined by counting one SEV environment per day if an environment is detected for any time step vs counts for the single 0600 UTC diurnal peak time step over the years 1980–2000 for (a) ERA-Interim, (b) CSIRO Mk3.6, and (c) CCAM. Contour spacing between 0 and 10 is at 2 environment intervals, while contour spacing between 10 and 46 are at 4 environment intervals.

  • Fig. 5.

    Frequency distribution of nonzero MLCAPE environments as a fraction of the total occurrences of nonzero MLCAPE for the 20-season climatology for each of the CSIRO Mk3.6, CCAM, and ERA-Interim datasets for all grid points over (a) Australia and (b) the EAReg and the relative frequency distribution of S06 for nonzero MLCAPE environments as a ratio of total environments with nonzero MLCAPE for (c) Australia and (d) the EAReg.

  • Fig. 6.

    Comparison among ERA-Interim, CSIRO Mk3.6, and CCAM of seasonal mean MLCAPE for (a),(c),(e) SON and (b),(d),(f) DJF. Values in the bottom right of (c),(d) correspond to pattern congruences over the Australian region between CSIRO Mk3.6 and the reanalysis regridded to the models spatial resolution, while (e),(f) correspond to pattern congruence between the reanalysis and CCAM, with the model regridded to the ERA-Interim spatial resolution.

  • Fig. 7.

    As in Fig. 6, but for the comparison of mean seasonal 700–500-hPa vertical lapse rate among ERA-Interim, CSIRO Mk3.6, and CCAM.

  • Fig. 8.

    As in Fig. 6, but for the comparison of mean seasonal S06 among ERA-Interim, CSIRO Mk3.6, and CCAM.

  • Fig. 9.

    Comparison of the mean frequency of SEV environments or reports as determined using the (a) mean seasonal reports, based on the climatology spanning September 2003–April 2010; (b) ERA-Interim; (c) CSIRO Mk3.6; and (d) CCAM. Reports are gridded to the reanalysis grid and contoured as for the models using mean seasonal values. Frequency is based on the number of days per year where CAPE × S061.67 exceeds the severe discriminant [Eq. (1)] for pseudosoundings at each grid point of the respective datasets for the EAReg. Favorable environmental conditions are described in the text. Contours are at 2 environment intervals from 0 to 10, 4 environment intervals from 10 to 30, and 8 environment intervals from 30 to 78.

  • Fig. 10.

    Mean seasonal cycle of the frequency of SEV environments over the EAReg as determined using (a) ERA-Interim, (b) CSIRO Mk3.6, and (c) CCAM. Reports are shown on the respective frequency distributions in gray, while the environmental mean, maximum, and minimum over the 20-yr period are shown by the black box and whiskers for the respective months. Frequencies are normalized by one-tenth of the total grid points over the EAReg, except in the case of the reports, which are climatological values. Note that the y-axis scale for CSIRO Mk3.6 was modified to encompass the frequencies of that model.

  • Fig. 11.

    Comparison among ERA-Interim, CSIRO Mk3.6, and CCAM of mean seasonal frequency of SEV environments over the 20 convective warm seasons of 1980–2000 per grid box for the respective model grids. Seasons correspond to (a),(c),(e) SON and (b),(d),(f) DJF. Contours are at 1 environment interval from 0 to 2, 2 environment intervals from 2 to 20, and 4 environment intervals from 20 to >36.

  • Fig. 12.

    As in Fig. 11, but for comparison among ERA-Interim, CSIRO Mk3.6, and CCAM climatologies of mean seasonal frequency of high MLCAPE environments exceeding 1000 J kg−1 conditional on the presence of S06 exceeding 5 m s−1.

  • Fig. 13.

    As in Fig. 11, but for comparison among ERA-Interim, CSIRO Mk3.6, and CCAM climatologies of mean seasonal frequency of high S06 environments exceeding 15 m s−1 conditional on the presence of MLCAPE exceeding 10 J kg−1.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 529 118 5
PDF Downloads 362 76 3